Frank Wuttke, Sebastian Bauer Marcelo Sanchez - Energy Geotechnics-Crc Press (2016) PDF

COPYRIGHT © 2016 TAYLOR and FRANCIS GROUP, LONDON, UK CRC PRESS / BALKEMA - PROCEEDINGS AND MONOGRAPHS IN ENGINEERING, WATER
TER AND EARTH SCIENCES

WWW.CRCPRESS.COM, WWW.TAYLORANDFRANCIS.COM
ISBN 978-1-138-03299-6
Energy Geotechnics
Editors: Frank Wuttke, Sebastian Bauer and Marcelo Sánchez
help open
Energy Geotechnics – Wuttke, Bauer & Sánchez (Eds)
© 2016 Taylor & Francis Group, London, ISBN 978-1-138-03299-6
Committees
INTERNATIONAL ADVISORY BOARD
Tony Amis (UK) Fernando Pardo (Spain)

Malek Bouazza (Australia) Jean-Michel Pereira (France)
Leonardo Cabral (Brazil) Enrique Romero (Spain)
Gye-Chun Cho (Korea) Marcelo Sánchez (US)
Yu-Jun Cui (France-China) Tom Schanz (Germany)
Pierre Delage (France) Kenichi Soga (UK)
Antonio Gens (Spain) J. Carlos Santamarina (US)
Leonardo Guimaraes (Brazil) Anh Minh Tang (France)
Jacco Haasnoot (Netherlands) Hywel Thomas (UK)
Tomasz Hueckel (US) Ivan Verdugo (Colombia)
Lyesse Laloui (Switzerland) Thomas Vienken (Germany)
M. R. Lakshmikantha (Spain) Peter Bourne-Webb (Portugal)
John McCartney (US) Frank Wuttke (Germany)
Guillermo Narsilio (Australia) Tae Sup Yun (Korea)
Duncan Nicholson (UK) Dietmar Adam (Austria)
Guney Olgun (US)
NATIONAL TECHNICAL PROGRAM COMMITTEE
Frank Wuttke, Kiel University, Germany

Sebastian Bauer, Kiel University, Germany
Andreas Dahmke, Kiel University, Germany
Tom Schanz, RU Bochum, Germany
Rolf Katzenbach, TU Darmstadt, Germany
Ernst Huenges, GFZ Potsdam, Germany
Olaf Kolditz, UFZ Leipzig, Germany
Stavros Savidis, TU Berlin, Germany
Dietmar Adam, TU Wien, Austria
MINISYMPOSIA ORGANIZERS
Dietmar Adam (TU Wien, Austria) and Malek Bouazza (Monash University, Australia) – Minisymposium
thermo-active foundations, tunnels and earth-coupled structures
Marcelo Sánchez (Texas A & M University, USA) and Christian Deusner (GEOMAR, Germany) –
Minisymposium geomechanical characterization and modeling of hydrate bearing sediments
David M.J. Smeulders (Eindhoven University of Technology, The Netherlands) and Sebastian Bauer & Frank
Wuttke (Kiel University, Germany) – Minisymposium trends and challenges in energy geotechnical storage
systems and materials
Robert Charlier (Université de Liège, Belgium) and Bertrand François (Université Libre de Bruxelles, Belgium) –
Minisymposium shallow geothermal systems
Pierre Duffaut (French Committee on Rock Mechanics) – Minisymposium geotechnics risk and items for
underground nuclear power plants
Enrique Romero (Universitat Politècnica de Catalunya UPC, Spain), Xiangling Li (European Underground
Research Infrastructure for Disposal of Nuclear Waste In Clay Environment EIG EURIDICE, Belgium)
and Paul Marschall (Nationale Genossenschaft für die Lagerung Radioaktiver Abfälle NAGRA, Switzerland) –
Minisymposium geotechnics for nuclear waste disposal
XIII
CONFERENCE ORGANIZATION AND TECHNICAL ENQUIRIES
Henok Hailemariam, Hem Motra and Frank Wuttke

Geomechanics and Geotechnics
Kiel University, Germany
LIST OF REVIEWERS
Malek Bouazza Enrique Romero

Dietmar Adam Pierre Duffaut
Christian Deusner Paul Marschall
Marcelo Sánchez Xiangling Li
Sebastian Bauer Henok Hailemariam
Frank Wuttke Zarghaam H. Rizvi
Robert Charlier Dinesh Shrestha
Bertrand François Amir S. Sattari
David M.J. Smeulders
XIV
Preface
The 1st International Conference on Energy Geotechnics (ICEGT 2016) was held from 29th to 31st August 2016,
at the University of Kiel, Kiel, Germany. This was the first conference under the guidance of the International
Society for Soil Mechanics and Geotechnical Engineering (ISSMGE) TC308 on Energy Geotechnics.
With the increasing energy demand and climate change implications, the development of sustainable energy
systems based on integrated schemes of energy production, transport, and transfer as well as energy storage is
of the utmost importance. This issue is of increasing interest to the research field of geotechnical engineering.
The focus of this relatively new research area is new developments and solutions for civil, environmental, and
industrial applications.
The behaviour of geomaterials (i.e. soils and rocks) when subjected to thermo-hydro-mechanical and chemical
solicitations is generally very complex. Their response is also strongly influenced by, shape, size and mineralogy
of the aggregates, as well as, by the type of loading and porosity. The ability of these materials to store and
dissipate energy is of great importance to energy transportation and storage systems. Beside the understanding
of the material behaviour, the study and development of energy geo-structure as well the multiphysics interaction
behaviour becomes of highest importance too.
The aim of the conference was to provide a wide platform for the interaction among colleagues from different
countries working in different subjects, and playing different roles in the energy sector. The discussions focused on
past, present and future investigations & practices in the area of energy geotechnics, covering from experimental
up to numerical and fundamental studies.
This proceedings consists of 97 peer reviewed papers from engineers and researchers, investigations and
reporting from their findings. The editors wish that this proceedings acts as a stack of knowledge for the frontier
of Energy Geotechnics.
Frank Wuttke
Chairman of the ICEGT 2016
Head of Geomechanics and Geotechnics
Kiel University
XI
Sponsors
ORGANIZING INSTITUTION
Kiel University, Germany
ORGANIZING ASSOCIATION
ISSMGE TC 308 on Energy Geotechnics
SPONSORS
Fugro Consult GmbH, Germany
(Platin-Sponsor)
APS Antriebs-, Prüf- und Steuertechnik

GmbH, Germany (Gold-Sponsor)
SUPPORTING AUTHORITIES
Business Development and Technology
Transfer Corporation of Schleswig-
Holstein, Germany
XVII
Acknowledgements
We would like to thank GEOMAR Helmholtz Centre for Ocean Research (Kiel, Germany), Andre Lindhorst
(Pur.pur, Kiel, Germany), Dietmar Adam (TU Wien, Austria) and Stiftung OFFSHORE-WINDENERGIE
(Berlin, Germany) for providing us with the photos used on the conference webpage as well as on the front
book cover.
We would also like to thank Henok Hailemariam (Kiel University, Kiel, Germany) for his outstanding
organizational work in the ICEGT 2016.
XV
Potential of district-scale geothermal energy in urban cities
K. Soga, Y. Zhang & R. Choudhary

University of Cambridge, Cambridge, UK
ABSTRACT: Ground source heat pumps (GSHPs) is a shallow geothermal system of pumping heat from or
to the ground to supply low carbon heating or cooling to buildings. The underground becomes a heart storage
and combination of heating and cooling can be beneficial to system efficiency improvement and temperature
balance underground. In this study, a GIS based simulation tool was developed to quantify how many GSHPs
could be installed at a district or a city scale without losing control of ground thermal capacity, and to identify its
contribution to both heating and cooling demands of buildings. The City of Westminster, London was selected
as a case study. Results show that many buildings (more than 50%) can install enough boreholes to support their
own heating and cooling demands. For high-rise buildings and the high people density infrastructures, the limited
space for geothermal energy extraction is a concern for geothermal applications. As domestic buildings have
more space to install boreholes and thus to obtain more energy from underground, residential houses can share
their surplus geothermal energy with commercial buildings and urban infrastructure at city scale. Parametric
analysis was carried out to investigate the influence of district size for sharing a GSHP network at city scale.
1 INTRODUCTION utilise the heat convection mechanism of groundwa-

ter flow by extracting heated or cooled water. Vertical
In the UK, £32 billion is spent on heating every year GSHPs are normally constructed by placing two small-
and the heat consumption accounts for approximately a diameter polyethylene tubes in a vertical borehole
third of total greenhouse gas emissions. As space heat- (typically 150 m deep), which horizontal GSHPs are
ing accounts for about 66% of the domestic energy placed in narrow trenches and this design requires
bills (DECC, 2013a) and delivers approximate 74% greater amount of ground area. The vertical GSHPs are
of the carbon dioxide emissions, the majority of heat commonly chosen for urban areas because they need
related activities are from fossil fuel consumption in relatively small plots of space; contact with the soil
the domestic sector (DECC, 2012). Many countries that varies little in temperature and thermal properties;
which have similar situation encourage the adaptation consume the smallest amount of pipe and pumping
of renewable energy technologies. As part of 2009 EU- energy; and can yield the most efficient performance.
wide action to increase the use of renewable energy, For the existing buildings in urban areas, closed ver-
the UK government has committed to set 15% renew- tical loops are used due to space limitations and their
able energy target by 2020. This is a significant rise system efficiency.
compared with approximate 2% in 2008. Additional requirement for heat network is bet-
It is traditionally considered that heat is generated ter to provide cooling as well. Cooling is playing an
on-site in individual buildings and the most com- increasing part in comfort satisfactory in buildings,
mon sources are electric heaters, gas boilers, and oil especially in the offices and other commercial build-
boilers. Ground source heat pumps (GSHPs) is a shal- ings. The cooling demand is projected to grow further.
low geothermal system of pumping heat from or to The combination of heating and cooling can make
the ground to supply low carbon heating or cool- district energy more economically attractive in the
ing to buildings. Shallow geothermal systems require areas with a mix of land use buildings (DECC, 2012;
no specific geological condition or high temperature DECC, 2013a). GSHP systems can match with this
gradient, so they are increasingly popular worldwide requirement. Not only is cooling supply a basic func-
as an environmental friendly alternative to traditional tion for the GSHP system, but combination of heating
technologies such as gas fired boilers (Haehnlein and cooling can be also greatly beneficial to sys-
et al., 2010). tem efficiency improvement and temperature balance
GSHPs can be mainly grouped into two types, underground. The underground essentially becomes a
closed loops and open loops. In closed GSHP, borehole heart storage.
system achieves heat exchange through the circula- The benefits of GSHP systems also include low
tion of fluid in a closed U-loop embedded within an running cost, minimal maintenance cost and long life
infill medium. In contrast, open loop GSHP systems expectancy. Heat pump can be expected to use for
3
20–25 years and the borehole can be used as long peak and an office building with day-time peaks, so
as 100 years (Kensa Heat Pumps, 2014). In order to that the heating load can be allowed to get balanced.
encourage the growth of this technology in the UK, According to the heat strategy of the UK government,
Renewable Heat Incentive (RHI) classified the GSHP provided they can be used to district heat from low car-
district heating into non-domestic tariff stream. The bon sources, heat networks can be the core and have
system is eligible to receive 8.7 pence per kWh heating great potential to play a key role in achieving targeted
consumption for as long as 20 years (DECC, 2013c). low carbon heat supply by 2050 (DECC, 2012).
Although GSHP has been available for long time, Although district heating has started since the
their applications are generally limited to a small scale. 1950’s, the market penetration is still quite low. The
If such ground source energy systems are employed typical district heating are university campuses, new
at a larger scale to provide low carbon heating solu- inner city commercial buildings and urban infrastruc-
tions to buildings and infrastructures, a low carbon tures (DECC, 2013b). It is estimated that networks
city would be developed. According to the UK Envi- currently provide less than 2% of UK heat demand,
ronment Agency (2009), the total number of installed including 1600 networks (DECC, 2013b) serving
GSHP systems in the UK at the time of year 2009 around 210,000 dwellings and 1700 commercial and
was 8000, of which, there were only 300 open loops public buildings across the UK (The Association for
(3.75%) and the rest are closed loops (96.75%). There Decentralised Energy, 2015). This percentage is much
has only a very few works on evaluating the poten- lower than the average value of 10% in Europe. In
tial capacity and sustainability of shallow geothermal some other parts of Europe, the district heating appli-
energy at a large scale. cations are much wider spread. In Denmark, Finland
In this study, a GIS based simulation tool was devel- and Sweden, for example, the market shares are high
oped to quantify the number of GSHPs that can be to around 70%, 49% and 50%, respectively (DECC,
installed at a district or a city scale without losing 2012). In some countries, although the district heat-
control of ground thermal capacity, and to identify its ing contributes a low percentage at the nation scale,
contribution to both heating and cooling demands of it makes a major supply in large cities. For exam-
buildings. Previously Zhang et al. (2014; 2015) used ple, district heating provides only 18% of total heat
this model to the city of Westminster, London. Results in Austria, but the percentage is doubled to be 36%
show that more than 50% of the buildings can install for Vienna (Poyry, 2009). The potential estimated by
enough boreholes to support their own heating and DECC is to supply 14% of all homes in the UK by
cooling demands. Although such large scale GSHP 2050 (DECC, 2012).
installations can approach long-term heating cost sav- There are currently many types of heat sources
ing of up to 70% (Committee on Climate Change, available. However, in order to achieve the Carbon
2013), it still requires more than 20 years for a general Plan by 2050, low carbon intensive ways over long
domestic home to get full cost-recovery. Hence more term are required by the government to heat the build-
proactive effort to increase the adoption rate of GSHP ings relying on coal, oil, and even the natural gas
is required from both commercial and regulatory sides. (HM Government, 2011). In addition, DECC have also
A small scale district heating/cooling system is given further support to heat networks in the Renew-
an attractive solution to implement the low carbon able Heat Incentive Policy (DECC, 2013c). Moreover,
energy utilization. For high-rise buildings and the high Energy Company Obligation (ECO), a subsidy from
people density infrastructures, the limited space for energy suppliers, works alongside the Green Deal
geothermal energy extraction is a concern for geother- to provide energy-saving improvement for vulnerable
mal applications. As domestic buildings have more households. The district heating has been considered
space to install boreholes and thus to obtain more as a primary measure (DECC, 2015). The Office of
energy from underground, residential houses can share Gas and Electricity Markets (Ofgem) has recently
their surplus geothermal energy with commercial confirmed the lifetime of GSHP district heating con-
buildings and urban infrastructure at city scale. This nections are 40 years under ECO (DECC, 2014).
paper extends the work of the authors (Zhang et al., Unlike the single system, the heat pump used in the
2014; 2015) by considering the potential of a shallow district system can be either a centralised system with a
geothermal technology for district scale heating. larger heat pump or a de-centralised system with mul-
tiples of smaller controlled heat pumps (DECC, 2012).
In both of cases, the heat pumps should be connected to
2 DISTRICT SCALE HEATING AND COOLING a communal ground array including one or more bore-
hole heat exchangers (BHE) to collect energy from
A district heating network is either two or more dis- ground. In the design process of BHE, the coverage
tinct buildings connected to a single heat source or on of the targeted area that receives energy supply should
building in which there are more than ten individual be firstly confirmed. The next step is to investigate
customers connected to a single heat source (DECC, all the geological site information and the building
2013b). Heat networks are to transport heat to con- load information for design. With all these prepared
sumers through a network of insulated pipes. They are work, the total required borehole length can be finally
able to deliver heat to the areas with a mix of sources calculated. After determining the total required BHE
of demand such as a residential house with night-time length, the most important task before installation is
4
to consider a borehole array for the reasonable district is the building design heating block load (W), qlc is
sharing. The array arrangement will decide what is the the building design cooling block load (W), Rga is the
proportion of the total required length can be satisfied. effective thermal resistance of the ground in annual
Therefore, the borehole array and the size of the tar- pulse (mK/W), Rgd is the effective thermal resistance
geted district are the two key parameters to influence of the ground in daily pulse (mK/W), Rgm is the effec-
the capacity of a GSHP district system. tive thermal resistance of the ground in monthly pulse
(mK/W), Rb is the thermal resistance of borehole
(mK/W), tg is the undisturbed ground temperature
3 CITY-SCALE GSHP SIMULATOR (K), tp is the temperature penalty for interference of
adjacent boreholes (K), twi is the liquid temperature
3.1 Background at heat pump inlet (K), two is the liquid temperature
at heat pump outlet (K), Wh is the power input at
The efficiency of a vertical GSHP system is highly design heating load (W), and Wc is the power input
dependent on correctly sizing the ground heat at design cooling load (W).
exchangers according to energy demand (Shonder The GIS tool gives the output of the total required
and Hughes, 1998). In order to estimate the numbers borehole length of GHE per building, which is the
of GSHPs that can be installed in specific areas of larger one of the results from Equations (1) and (2).The
cities or districts, and the numbers that are required individual borehole length is set typically as 150 m,
to satisfy heating demands, it is necessary to quantify which is the common-used value in practice for verti-
the geothermal capacity using high-resolution infor- cal closed loop GSHP installations. The number of
mation about land-use, underground conditions and boreholes per building is then calculated based on
heating/cooling demand. these two length values.
Zhang et al. (2014; 2015) developed a simulation The land area may not be enough for the required
tool that integrates, within a geographic informa- number of boreholes for some buildings. In such cases,
tion system (GIS), high-resolution land use datasets, the maximum possible borehole length (the maximum
heating/cooling demands of buildings, ground prop- borehole number × 150 m) for a land area is provided
erties, and the ground heat exchanger (GHE) design as input into the model. The model inversely calcu-
calculations. A PYTHON code that estimates the size lates the maximum heating and cooling demands that
of GSHP required for a given heat demand of a can be provided by a GSHP system for that build-
building was developed and embedded into ArcGIS ing. In this study, the GSHP capacity of a building is
software, which is a widely used platform for spatial defined as the ratio of capacity to demand (C/D), and
design and analysis. This integrated simulation tool calculated by dividing the maximum possible num-
has been used to quantify the exact geothermal capac- ber of boreholes within the building’s land area by the
ity of specific areas. A brief overview of the tool is required borehole number. If the C/D ratio of a build-
given in this section, but further details can be found ing is equal to or greater than 100%, both heating and
in Zhang et al. (2014; 2015). cooling demands of this building could be satisfied by
Most current GHE design software packages refer its own GSHP system.
to the Cylinder and Line Source Method, which has
been found to be the most accurate model through
comparisons with calibrated data from actual instal- 3.2 Case Study: The city of Westminster, London
lations (Shonder and Hughes, 1998). The tool utilises The influence of switching the single system to the
the following heat transfer equation that determines district network system on the GSHP satisfactory
the required vertical borehole length Lh for heating capacity at the city scale was investigated for the city
(Kavanaugh and Rafferty, 1997). of Westminster, London. The site specific or spatial
inputs in the design calculation were firstly prepared
for the case study, which were heating demand per
building, cooling demand per building, thermal con-
ductivity, thermal diffusivity, and ground temperature.
In addition to the spatial data, related conditions and
assumptions for the GSHP and the borehole are listed
The required vertical borehole length Lc for cooling is in Table 1.
found using the following equation. For the calculation of heating and cooling demands,
UKMap, a GIS database, was used to obtain spatial
information about buildings within the City of West-
minster including building type, floor area and height.
According to this database (see Figure 1), there are
95,817 buildings within this district. 83% of the floor
area is made up of residences, offices, and retail.
where Fsc is the short-circuit heat loss factor, PLF m The remaining 17% includes hotels, schools, hospitals
is the part-load factor during design month, qa is the and leisure facilities (Choudhary, 2012). The inten-
net annual average heat transfer to the ground (W), qlh sity of heating and cooling demand per building type
5
Table 1. Conditions and assumptions in bhe design for
planning parametric study.
Parameter Unit Value Justification
Coefficient of / 3.3 Kavanaugh and

Performance (COP) Rafferty, 1997
Energy Efficiency / 4.2 Kavanaugh and
Ratio (EER) Rafferty, 1997
Short-circuit Heat / 1.04 Kavanaugh and
Loss Factor (Fsc ) Rafferty, 1997
Liquid Temperature K 278.5 Chosen Design
at heat pump inlet value
for Heating (twi )
Liquid Temperature K 275.0 Estimate based
at heat pump outlet on typical
for Heating (two ) temperature drop
Liquid Temperature K 300.0 Chosen Design Figure 1. Land use type distributions of buildings in west-
at heat pump inlet value minster.
for Cooling (twi )
Liquid Temperature K 308.0 Estimate based
at heat pump outlet on typical in east-west and north-south directions and 1 m in the
for Cooling (two ) temperature drop vertical direction. In the case study, the average ther-
Minimum Borehole m 6 MIS (DECC, mal property value within the depth of 150 m of each
Spacing 2008)a grid was estimated to develop the thermal conductivity
Borehole Diameter mm 130 MIS (DECC, and the thermal diffusivity maps of Westminster.
2008)a For the ground temperature, the ground temperature
Pipe Diameter mm 32 mm OD MIS (DECC,
of London was used. Headon et al. (2009) gave the
SDR-11 2008)a
Thermal W/m.K 0.420 MIS (DECC, underground temperature information of London city
Conductivity (PE 100) 2008)a based on the well data as 12.3◦ C at 60 m depth, 12.8◦ C
of Pipe at 80 m depth and 13.1◦ C at 100 m depth. Accordingly,
Pipe Centre-Pipe mm 52 MIS (DECC, the ground temperature in the design was set to be
Centre Shank 2008)a 12.8◦ C, which was considered as the average ground
Spacing temperature value within the depth of 150 m.
Thermal Transfer / 25% Mono MIS (DECC, Sizing Heat pump is a key intermediate step in the
Fluid Ethylene 2008)a GHE design calculation. A database of heat pumps
Gylcol
within capacity range of 5–75 kW is therefore also
a. MIS (Microgeneration Installation Standard), DECC
included within the model. If the required capacity is
(Department of Energy and Climate Change, UK) (2008). outside of this range, a combination of two or more
heat pumps is used. In the design calculation, the
temperature penalty tp is used as a parameter to con-
sider thermal interference of adjacent boreholes with
(in kWh/m2 per year) was gathered from DECC cer- reasonable borehole arrangements. Hence, within the
tificates (compiled and released by the UK Centre for run time, there is temperature decrease surrounding
Sustainable Energy), Chartered Institution of Building the borehole, and the temperature reduction decreases
Services Engineers Guide F and TM 46 (CIBSE, 2004, with time as transient state. It is assumed that no heat
2008) and 2011 energy distribution charts (EDCs).The is diffused out of a square cylinder with sides equal
design heating block load per building (qlh in kW) to the borehole separation distance (Kavanaugh and
was estimated by multiplying the heating demand per Rafferty, 1997).
building type in kWh/m2 per year with the floor area
of a building and dividing by the number of heating
hours in a year (2160 h in this case, assuming 12 h of
3.3 Borehole location scenarios
heating per day for half the year). Cooling estimation
is performed in the same way. In this case study, the required number of boreholes for
The thermal conductivity and thermal diffusivity each building was allocated at a spatial position. Two
maps were developed based on the geological map and scenarios were considered: (a) under building (Fig-
the thermal property look-up table. The soil distribu- ure 2a) – within the land-area of the existing building,
tion map of the City of Westminster was obtained from and (b) around the building (Figure 2b) – on the buffer
the British Geological Survey (BGS) geological map area with the building boundary as the midline. The
of London. For each type of soil, its thermal conductiv- shaped polygons and the points stand for the buildings
ity and thermal diffusivity were assigned according to and the installed boreholes, respectively. In order to
the logs from site investigation work by BGS. A ther- avoid thermal interference, the spacing between any
mal conductivity map and a thermal diffusivity map two boreholes was fixed at 6 metres, as per the MIS
were then developed with grids of size 50 m × 50 m (DECC, 2008). This means that no heat is diffused
6
Figure 3. Map of ratio of capacity to demand of westminster
for heating and cooling (Scenario 1: Under buildings).
Figure 2. Borehole allocation map.
out of a square cylinder with sides equal to 6 metres

with a reasonable temperature penalty. Outside this
square cylinder, the ground temperature is assumed to
be undisturbed.
Scenario 1 is more suitable for new and refur-
bished buildings, but scenario 2 is more practical for
existing buildings and can be achieved by using direc-
tional drilling at a shallow depth to the target borehole
location followed by drilling vertically downwards. In
these two cases, the permitted areas for locating the
boreholes are different, and thus the maximum num- Figure 4. Map of ratio of capacity to demand of westminster
ber of boreholes that can be installed for each building for heating and cooling (Scenario 2: Around buildings).
also varies. In scenario 2, boreholes are installed in
a buffer area that is within 3 meters of the edges of a high proportion of buildings are capable of meeting
building, both away from and under it. The installation their own demands is that, for the total 95,817 build-
area can be changed in the model to correspond to more ings in Westminster, 77,355 buildings have 5 floors
restrictions, such as pavements and parking areas. or fewer (80.7%), 17,638 buildings have 6–10 floors
The ratio of capacity to demand (C/D) is cal- (18.4%) and only 824 buildings have more than 10
culated by dividing the maximum possible number floors (0.87%). Most of the low rise (5 floors or fewer)
of boreholes within the building’s land area by the buildings can meet their own heating demands by
required borehole number for the heating and cool- GSHP only since the typical heating demand for these
ing demand. Figure 3 shows the map of the C/D ratio buildings is only around 40 W/m2 .
for Scenario 1, whereas Figure 4 shows the map for The difference between the two scenarios is mainly
Scenario 2. Green colour represents the buildings that the result of the presence of many long, narrow build-
have a large enough land area to support their own ings in Westminster. For these buildings, more land
heating and cooling demands. Red represents build- area is available for borehole installation around them
ings that can support less than 50% heating demand than underneath them. In contrast, buildings in the cen-
due to their small land area. The rest buildings with tral area cover larger areas, and so they could have more
ratio in the range from 50% to 100% are indicated by boreholes installed under the buildings.
the yellow colour. The detailed discussion of this particular result can
In Scenario 1, 51% of buildings can meet their be found in Zhang et al. (2014; 2015). Under both
own heating and cooling requirements. Such build- scenarios, not all the buildings have GSHP capacity
ings are generally found at the edges of built-up areas. greater than 100%. However, if boreholes can be
In Scenario 2, 67% of the buildings have enough land shared with neighbours who have surplus geothermal
area for a GSHP system to cover their heating and capacity, then such district scale GSHP system has
cooling demands. Buildings that can support less than the potential to meet the heating and cooling demands
50% of their heating demand are concentrated in the of all buildings within this area, which is the main
centre of Westminster. The main reason that such a objective of the study presented in this paper.
7
4 DISTRICT SCALE NETWORK
4.1 Background
A district heating network is either two or more dis-
tinct buildings connected to a single heat source or
on building in which there are more than ten indi-
vidual customers connected to a single heat source.
The network size is also defined on the basis of the
number of domestic and non-domestic customers con-
nected to the heat source. These are large networks
(500 or more residential properties and/or more than
10 non-domestic users); Medium networks (between
100 and 500 residential properties and/or between
3 and 10 non-domestic users; and Small networks
Figure 5. Grids for dividing westminster into districts.
(less than 100 residential properties and/or less than
3 non-domestic users). Currently in the UK, the small
size networks take the predominant position, occupy-
ing 75% of the total 1765 individual district heating
networks. The average number of the dwellings per
network is 35.The medium and the large sizes share the
rest proportions, which are 20% and 5%, respectively
(DECC, 2013b).
For the GSHP district heating, the size can range
from micro to macro scale. With this linking together, Figure 6. C/D ratio frequency distribution with various
the capital cost can be reduced due to smaller num- district sizes for scenario 1.
ber of deeper boreholes and the borehole location can
be more flexible. Moreover, residential houses can
share their surplus geothermal energy with commer- one district by itself. For all the cases, it was assumed
cial buildings and urban infrastructure at a large scale that, no matter how large the district was, the network
to achieve a low carbon city. In addition, for the resi- was able to achieve the perfect energy sharing and sup-
dential houses, more tariff benefits from switching to plying. The C/D ratio of one district (or one grid) was
a district network can be received because the eligi- calculated by dividing the total maximum borehole
ble incentive scheme can change from the domestic to number allowed by the whole available installation
the non-domestic. The domestic scheme offers 19.1 space by the total required borehole number satisfy-
pence per kWh renewable energy consumption for ing all the heating and cooling demands within this
7 years (DECC, 2013d) and the non-domestic one district. The C/D ratio frequency distributions for all
provides 8.84 pence per kWh energy consumption for the cases with various district sizes were estimated and
20 years (DECC, 2013c). compared for the parametric analysis.
According to the results of investigating the GSHP
capacity at city-scale for Westminster, the C/D ratios
4.3 Results
are not evenly distributed. If district planning can be
introduced to play a role, the buildings with high C/D Figure 6 shows the C/D ratio frequency distributions
ratios will achieve to share their surplus capacities with with various district sizes under Scenario 1 (Bore-
neighbours, so that more heating and cooling demands hole under Buildings). The values on the horizontal
in this area will be satisfied. An analysis needs here to axis demonstrate the district size. For example, GS50
prove if district sharing can truly increase the propor- means grid size of 50 m × 50 m. Figure 7 shows spa-
tion of high C/D ratios and also to investigate what is tial distributions of C/D ratio with district size of 50 m
the influence of district size on the GSHP capacity at under Scenario 1. In Figures 6 and 7, green colour
city scale. indicates the ratio of 100% or more, which means the
heating and cooling demand can be fully satisfied by
its own capacity. Yellow colour and red colour stand
4.2 Parametric study on district network scale for 50%–100% and 0%–50%, respectively.
In this analysis, the whole Westminster was divided In the case of GS0, the largest proportion of around
into grids, which were uniform in the size (see 50% is taken by the green colour and the smallest per-
Figure 5). Each grid was considered to be one district. centage of approximately 15% is yellow. In the case
There were 13 cases with various grid sizes for compar- of Whole, the C/D ratio value representing the GSHP
ison analysis. The grid sizes were set to be 0 m, 50 m, capacity of the one whole Westminster district network
100 m, 200 m, 300 m, 400 m, 500 m, 600 m, 700 m, is in the yellow colour. It can be seen that C/D ratio
1000 m, 1500 m, 2000 m and a Whole. Among these, distribution changes gradually from GS0 to Whole.
‘0 m’ stood for that there was no district sharing, and ‘a The most significant difference in the C/D ratio distri-
Whole’stood for an extreme case that Westminster was bution occurs at the beginning, where the grid size is
8
Figure 7. Spatial distributions of C/D ratio with district size Figure 9. Spatial distributions of C/D ratio with district size
of 50 m for scenario 1. of 50 m for scenario 2.
red colour proportion, which stands for the C/D ratio

less than 50%, is slowly reduced. Therefore, GS50 is
considered to be an optimal network size for enhanc-
ing GSHP capacity at the city scale for Westminster.
The GSHP district network of this size, 50 m × 50 m,
can greatly improve the C/D ratio distribution and
also practical to achieve with considerations of the
construction feasibility and the capital cost.
Figure 8. C/D ratio frequency distribution with various Around 20% of buildings in Westminster are high-
district sizes for scenario 2.
rise with more than five floors and it is highly difficult
to satisfy them with GSHP. With building district net-
enlarged from GS0 to GS50. After this, the C/D ratio works, more and more low-rise buildings can share
distribution changes more gently. their capacities with high-rise ones. However, for the
Figure 8 shows the C/D ratio frequency distri- gathering place of high-rise buildings, no surplus
butions with various district sizes under Scenario 2 energy from surrounding low-rise buildings can be
(Borehole around Buildings). In the case of GS0, provided any more. Therefore, within a certain range
green, yellow and red take account for 67%, 15% and of district size, enlarging the size has quite slight effect
14%, respectively. Followed by a gradual change, all on the C/D ratio distribution. This can be an issue for
the 98% turns to green in the case of Whole.Among the the megacities, for example, the city of Westminster in
changing procedure, the most significant step is at the this study. Compared with continuing increasing the
very beginning, where the size increases from GS0 to district size, keeping the district size of GS50 and intro-
GS50. According to Figure 9, which shows spatial dis- ducing supplementary energy sources are considered
tributions of C/D ratio with district size of 50 m under to be more cost-effective.
Scenario 2, very small percentages of grids are in the
red colour and most of the districts can be fully satis-
fied as they have enough space to install the required 5 CONCLUSIONS
boreholes. In addition, the change from GS2000 to
Whole is also great as the green proportion increases Vertical closed loop GSHP system can be an envi-
from around 75% to 98%. However, the improvement ronmental friendly application for supplying heating
from GS50 to GS2000 is very gentle and every time and cooling in the urban areas. In this paper, a city-
of size growth can only lead to a slight change. scale simulation tool developed by Zhang et al. (2014;
For the city of Westminster, it can be found that C/D 2015) was used to identify how many GSHPs could
ratio distributions under Scenarios 1 and 2 have sim- be installed in a district without overusing geothermal
ilar variation trends with the grid size. Both of them energy and to calculate the ratio of its contribution to
change gradually from the distribution in the condi- the heating and cooling demands of buildings.The City
tion of no district sharing (GS0) to the C/D ratio in of Westminster was selected as a case study. Two sce-
the condition that Westminster is one large district narios for borehole installations (‘under buildings’and
network (Whole). However, the influence of district ‘around buildings’) were examined and the borehole
size variation is greater under Scenario 1 than Sce- allocation maps and C/D ratio distribution maps were
nario 2 because the difference between GS0 and Whole generated. Results show that many buildings (51% for
is larger under Scenario 1. In addition, under both sce- Scenario 1 and 67% for Scenario 2) can install enough
narios, the most significant change due to increasing boreholes to support their own heating and cooling
the grid size occurs from GS0 to GS50.After GS50, the demands. There is a high proportion of C/D ratios (the
9
ratio of capacity to demand) over 100%, which means Department of Energy and Climate Change (DECC). 2013b.
a great number of houses and buildings have enough Summary evidence on District Heating Networks in the
space to install GSHP and thus to fully satisfy their UK. London, UK.
own heating demands. However, the C/D ratios are not Department of Energy and Climate Change (DECC). 2013c.
Renewable Heat Incentive: Non-Domestic Scheme Early
evenly distributed due to various heating demands and Tariff Review. London, UK.
available spaces. If district heating can be introduced Department of Energy and Climate Change (DECC). 2013d.
to play a role, the buildings with high C/D ratios will Domestic Renewable Heat Incentive: the first step towards
achieve to share their surplus capacities with neigh- transforming the way we heat our homes. London, UK.
bours, so that more heating demands in this area will Department of Energy and Climate Change (DECC). 2012.
be satisfied. The Future of Heating: A strategic framework for low
Following the work of Zhang et al. (2014; 2105), carbon heat in the UK. London, UK.
parametric analysis was carried out to investigate the Department of Energy and Climate Change (DECC). 2008.
influence of planning parameters on the application of Microgeneration Installation Standard:MIS 3005 Issue
3.0. London,UK.
GSHP systems for heating and cooling at city scale Environment Agency. 2009. Ground Source heating and
for a case study of Westminster, London. The district cooling pumps-state of play and future trends. Bristol, UK.
network is encouraged by the government for improv- Haehnlein, S., Bayer, P. & Blum, P. 2010. International legal
ing energy utilization and reducing carbon emissions. status of the use of shallow geothermal energy. Renewable
The influence of district size for sharing a GSHP net- and Sustainable Energy Reviews. 14(9), 2611–2625.
work on the C/D ratio distribution at city scale was Headon J., Banks D., Waters A. & Robinson V.K. 2009.
investigated. It was found that C/D ratio distributions Regional distribution of ground temperature in the Chalk
under Scenarios 1 and 2 have similar variation trends aquifer of London, UK. Quarterly Journal of Engineering
with the grid size. Both of them change gradually from Geology and Hydrogeology, 42, 313–323.
HM Government. 2011.The Carbon Plan: Delivering our low
the distribution in the condition of no district sharing carbon future. Presented to Parliament pursuant to Section
(GS0) to the C/D ratio in the condition that West- 12 and 14 of the Climate Change Act 2008. London, UK.
minster is one large district network (Whole). Under Kavanaugh S.P. & Rafferty K. 1997. Ground-Source Heat
both of scenarios, the most significant change due to Pumps Design of Geothermal Systems for Commer-
increasing the grid size occurs from GS0 to GS50 cial and Industrial Buildings. 1997 American Society of
(50 m × 50 m grid). After GS50, the red colour pro- Heating. USA.
portion is slowly reduced. GS50 is considered to be an Puttagunta S., Aldrich R.A., Owens D. & Mantha P. 2010.
optimal network size for enhancing GSHP capacity at Residential Ground-Source Heat Pumps: In-Field System
the city scale for Westminster. Compared with contin- Performance and Energy Modeling. GRC Transaction, 34,
941–948.
uing increasing the district size, taking the district size Kensa Heat Pumps. 2014. Ground Source District Heating.
of GS50 and using supplementary energy sources for www.kensaengineering.com.
the unsatisfied districts are considered to be more cost- The Association for Decentralised Energy. 2015. District
effective. heating sector: A step ahead on protecting heat cus-
tomers. http://www.theade.co.uk/district-heating-sector–
a-step-ahead-on-protecting-heat-customers_3016.html
POURY. 2009. The potential and costs of district heating net-
REFERENCES works, A report to the Department of Energy and Climate
Change.
Banks D. 2012. An Introduction to Thermogeology, 2nd ed., Schwencke A.B. 2013.Analysis and modelling of deep energy
Wiley-Blackwell. wells. Master’s thesis. Norwegian University of Sci-
British Geological Survey. 2011. Temperature and Thermal ence and Technology. Department of Energy and Process
Properties (Detailed). GeoReports. London, UK. Engineering.
CIBSE. 2008. Energy benchmarks: Technical memorandum Svendson P. 2008. A heat exchange module, in particular
46, Technical Report, 2008. for a ground source heat pump. EP. WO2008009289A1.
CIBSE. 2004. Cibse guide F: Energy efficiency in buildings, 24/01/2008.
Technical Report, CIBSE; 2004. Synergy boreholes and systems Ltd. 2015. Geothermal Bore-
Choudhary, R. 2012. Energy analysis of the non-domestic holes. http://www.synergyboreholes.co.uk/geothermal_
building stock of Greater London. Building and Environ- boreholes/
ment, 51, 243–254. Shonder, J.A. & Hughes, P.J. 1998. Increasing confidence
Committee on Climate Change. 2013. Fourth Carbon Budget in geothermal heat pump design methods. In: Stiles L,
Review-technical report. Chapter 3: Reducing emissions editors. 2nd Stocjton Geothermal Conference, pp. 41–57.
from buildings. London, UK The GeoInformation Group Ltd. 2010. UK Map USER
Department of Energy and Climate Change (DECC). 2015. GUIDE Version 4.0 February 2010.UK.
Policy paper 2010 to 2015 government policy: household ZhangY., Soga K. & Choudhary R. 2014. Shallow geothermal
energy. London, UK. energy application with GSHPs at city scale: study on the
Department of Energy and Climate Change (DECC). City of Westminster. Geotechnique Letters, 4, 125–131.
2014. The Future of the Energy Company Obligation. Zhang, Y., Choudhary, R. and Soga, K. (2015) “Influence
London, UK. of GSHP System Design Parameters on the Geothermal
Department of Energy and Climate Change (DECC). Application Capacity and Electricity Consumption at City
2013a. The Future of Heating: Meeting the challenge. Scale for Westminster, London,” Energy and Buildings,
London, UK. 106, 3–12.
10
Multiphysical phenomena and mechanisms involved with energy piles
Lyesse Laloui & Alessandro F. Rotta Loria

Swiss Federal Institute of Technology in Lausanne, EPFL, Laboratory of Soil Mechanics,
LMS, Lausanne, Switzerland
ABSTRACT: This study proposes an analysis of the multiphysical phenomena and mechanisms governing
the thermo-mechanical behaviour of energy piles. The analysis is based on the results of a series of full-scale
in-situ tests, laboratory experiments and numerical analyses. First, the thermo-mechanical behaviour of energy
piles is considered. Attention is given to both single and groups of energy piles. Next, the response of soils
and concrete-soil interfaces subjected to temperature changes is reviewed. The behaviours of clayey soils in
different overconsolidation states as well as of both concrete-sand and concrete-clay interfaces are analysed.
Finally, aspects considered of paramount importance for the analysis and design (e.g., geotechnical, structural
and energy) of energy piles are presented. Both floating and end-bearing energy piles are investigated. The goal
of this paper is to increase the confidence of civil engineers on the performance of energy piles.
1 INTRODUCTION obtained through laboratory experiments dealing with

oedometric tests (Di Donna and Laloui, 2015) and
Geothermal heat exchanger piles, often referred to as concrete-soil interface shear tests (Di Donna et al.,
energy piles, are foundations equipped with absorber 2015) under non-isothermal cyclic conditions are pro-
pipes with a heat carried fluid circulating into them posed for this aim. Finally, aspects considered of
that exchange heat with the surrounding ground. The paramount importance for the analysis and design of
temperature variations associated to the geothermal energy piles are presented. Results of finite element
operations of these foundations bring new challenges analyses devoted to analyse the geotechnical and struc-
to geotechnical and structural engineers. Thermal tural performance of single (Rotta Loria et al., 2015)
expansions and contractions of these elements and and groups (Rotta Loria and Laloui, 2016) of energy
the surrounding soil, with associated variations of piles as well as to investigate the energy performance
the stress and displacement fields in these media, are (Batini et al., 2015) of these foundations are presented
examples of such challenges. These phenomena must for this purpose.
be kept within acceptable limits to ensure an adequate
serviceability performance (e.g., mechanical) of the
energy piles. 2 THERMO-MECHANICAL BEHAVIOUR OF
This paper proposes insights into the analysis of the ENERGY PILES
various challenges associated to the geothermal and
structural support operations of energy piles. In par- 2.1 Single energy piles
ticular, it proposes an analysis of the multiphysical
phenomena and mechanisms governing the thermo- A full-scale in-situ test was performed by Laloui
mechanical behaviour of these foundations. The analy- et al. (2003) to investigate the thermo-mechanical
sis is based on the results of a series of full-scale in-situ behaviour of single energy piles.
tests, laboratory experiments and numerical analyses
performed at the Swiss Federal Institute of Technol- 2.1.1 The foundation and site
ogy in Lausanne (EPFL), Switzerland. The goal of this A four-storey building under construction at the EPFL
paper is to increase the confidence of civil engineers was chosen for this in-situ test. The structure (cf., Fig-
on the performance of energy piles, looking at their ure 1) was founded on piles approximately 25 m long.
increasing worldwide diffusion. The tested pile was 0.88 m in diameter and 25.8 m long
First, the thermo-mechanical behaviour of energy and was drilled in soil whose stratigraphic profile is
piles is considered. The results of in-situ tests car- presented in Figure 2. The groundwater table in this
ried out on single (Laloui et al., 2003) and groups zone is located at the ground surface. Polyethylene
(Rotta Loria and Laloui, 2016) of real-scale energy tubes were attached vertically (U-shaped configura-
piles are summarised for the considered purpose. tion) to the reinforcing cage of the pile to allow for
Next, the response of soils and concrete-soil interfaces the heating and cooling of this element for experi-
subjected to temperature changes is reviewed. Data mental purposes. The instrumentation chosen for the
11
Figure 1. The single energy pile test at the EPFL.
Figure 3. Evolution of temperature in the energy pile during
heating.
Figure 4. Evolution of vertical strains in the energy pile

during heating.
Figure 2. Stratigraphic profile and instrumentation of the
tested single energy pile.
Notable temperature changes can develop in energy
piles. In one of the considered tests, starting with initial
measurement of strain, temperature and load in this values of temperature that ranged from 13 to 14◦ C
test was made up of 58 gauges placed as indicated in along the energy pile length, it was reached an average
Figure 2. maximum value of 35◦ C (cf., Figure 3).
More detailed information on the site and founda- The thermal loads that are applied to energy piles
tion investigated as well as on instrumentation used generate expansive strains upon heating and contrac-
can be found in Laloui et al. (2003). tive strains upon cooling. These strains are generally
not uniform and develop primarily depending on the
friction characterising the pile-soil interface. In one
2.1.2 Features of the experimental test of the considered tests, average expansive vertical
The tested energy pile was subjected to two types strain variations caused by the temperature change
of loading: mechanical and thermal. The mechani- in the energy pile up to εv = −200 µε were mea-
cal load was applied through the dead weight of the sured (cf., Figure 4). These strains showed a reversible,
building under construction. The thermal load was i.e., thermo-elastic, character (cf., Figure 5). The
applied through a heat pump. The solicitations were radial strain measurements also showed a reversible
alternated in order to show the thermo-mechanical thermo-elastic character.
coupling clearly. At the end of the construction of each Because a proportion of the strain that is induced
storey, a thermal loading cycle was applied to the pile. by the thermal loads that are applied to energy piles
Nine different tests were performed. is generally prevented by the friction with the sur-
rounding soil, the presence of the superstructure at the
2.1.3 Observed behaviour pile head and the end-bearing at the pile toe, thermal
The result of this study demonstrated that the stresses develop in energy piles. Thermally-induced
behaviour of single energy piles is significantly char- compressive stresses comparable to those induced
acterised by the applied thermal loads associated to by the applied superstructure mechanical loads were
the geothermal operation of these foundations. observed in this study.
12
Figure 7. (a) The EPFL Swiss Tech Convention Centre
Figure 5. Evolution of vertical strains in the energy pile (http://www.tstcc.ch/, author: Frédéric Rauss); (b) plan view
during heating and cooling at a depth of z = −6.5 m. of the foundation; (c) schematic of the soil stratigraphy.
piles (labelled P1-16 in Figure 7 (b)) below a heav-

ily reinforced 0.9 m-thick slab. The energy piles are
28 m long and 0.9 m in diameter. All of the piles were
bored, cast onsite and are made of reinforced con-
crete. Vertical loads of 0, 800, 2200 and 2100 kN are
applied to energy piles EP1, 2, 3 and 4, respectively.
The energy piles were equipped with four 24-m-long
high-density polyethylene U-loops that are connected
in series. These loops were thermally insulated at the
inflow and outflow for 4 m to limit the affection of the
heat exchange process characterising the energy piles
by the surface thermal conditions.
All of the energy piles were instrumented with strain
gauges, optical fibres and thermocouples along their
Figure 6. Evolution of vertical pile head displacements
throughout the considered test.
lengths as well as with pressure cells at their toes.
Piezometers and thermistors were installed in two
boreholes in the soil.
Head heaves occur in energy piles upon heating The soil stratigraphy of the site (cf., Figure 7 (c))
whereas head settlements occur upon cooling as a con- was extrapolated based on information that was
sequence of the thermal deformation. An example of obtained during the construction of the foundation and
the head heaves (negative displacement values) that data from Laloui et al. (2003; 2006) collected for the
characterised the energy pile in one of the considered single energy pile test described above.
tests is presented in Figure 6. More detailed information on the site and founda-
tion investigated as well as on instrumentation used
2.2 Energy pile groups can be found in Rotta Loria and Laloui (2016).
A number of in-situ tests were recently performed
by Mimouni and Laloui (2015) and Rotta Loria and 2.2.2 Features of the experimental test
Laloui (2016) to investigate the thermally-induced The experimental test involved the application of a
group effects and interactions characterising closely heating-passive cooling cycle to energy pile EP1 (for
spaced energy pile groups. A short series of results of approximately 5 and 10 months, respectively), which
the in-situ test performed by Rotta Loria and Laloui was the only energy pile of the group that operated
(2016), which involved the analysis of closely spaced as a geothermal heat exchanger (cf., Figure 7 (b)).
energy piles that partially operate as geothermal heat Throughout the test, the mechanisms and phenomena
exchangers over a time-scale that is typical of practical occurring in the operating energy pile EP1, in the three
applications, will be considered in the following. surrounding non-operating energy piles EP2, 3 and 4,
and in the soil were recorded.
2.2.1 The foundation and site
The pile foundation that was considered for the exper- 2.2.3 Observed behaviour
imental test is located under the recently built Swiss The result of this study demonstrated that the
Tech Convention Centre, Lausanne, Switzerland (cf., behaviour of groups of closely spaced energy piles that
Figure 7 (a)). The foundation supports a 9 × 25 m2 operate partially as geothermal heat exchangers over
water retention tank and comprises a group of four end- time-scales that are typical of practical applications is
bearing energy piles (labelled EP1, EP2, EP3 and EP4 characterised by significant thermally-induced group
in Figure 7 (b)) and sixteen semi-floating conventional effects. These group effects are evidenced through
13
Figure 9. Experimental setup: (a) global view and (b) detail
(1: tubes with circulating water at the desired temperature, 2:
LVDTs, 3: thermocouples, 4: water supplier, 5: insulation, 6:
acquisition system, 7: heaters).
than the expansion that this pile would have undergone

under free thermal expansion conditions.
As a proportion of the thermally-induced defor-
mation of the energy piles was restrained due to the
presence of the soil and slab, vertical stress varia-
tions were observed in such elements. At the end of
the heating phase of the test, the operating energy pile
EP1 was subjected to a maximum compressive vertical
stress variation of σv = 5500 kPa. The non-operating
Figure 8. Temperature and vertical strain variations in energy pile EP2 was characterised in its upper and
operating and non-operating energy piles in a group. lower portions by maximum vertical stress varia-
tions of σv =1370 kPa (compressive) and −1419 kPa
(tensile), respectively.
thermal and thermally-induced mechanical interac-
Particular attention was devoted to the analysis of
tions between the operating and non-operating energy
these interactions by Rotta Loria and Laloui (2016)
piles.
because they were demonstrated to play an impor-
Thermal interactions between operating and non-
tant role in the geotechnical, structural and energy
operating energy piles occur during successive stages
performance (e.g., serviceability) of energy piles.
of geothermal operations. In this case study, thermal
interactions between a single operating energy pile
and the surrounding non-operating energy piles (3 m 3 RESPONSE OF SOILS AND
away; i.e., centre-to-centre spacing of approximately CONCRETE-SOIL INTERFACES SUBJECTED
3·D) occurred after 10 days of geothermal operation TO TEMPERATURE CHANGES
(constant thermal power of 3 kW) over 156 days. At
the end of the heating phase of the test (t = 156 days 3.1 Response of soils subjected to monotonic and
of testing), the uninsulated portion of the operating cyclic temperature changes
energy pile EP1 was subjected to an average tempera- The response of soils subjected to monotonic and
ture change of T = 20◦ C (cf. Figure 8). Because of cyclic temperature changes was recently investigated
the aforementioned thermal interactions, the region of by Di Donna and Laloui (2015).
the non-operating energy pile EP2 in correspondence
with the considered setting was subjected to an aver-
age temperature change of T = 5.3◦ C (heat diffused 3.1.1 Experimental setup, material and methods
in the soil with time). The devices that were employed for the experiments
Thermally-induced mechanical interactions are are four oedometric cells that were adapted to include
always present throughout geothermal operations. the control of temperature (cf., Figure 9).
These interactions are governed by the interplay Four natural silty-clay samples were collected near
between the responses of the piles, slab and soil to Geneva, Switzerland. The in-situ soil conditions were
temperature changes. Heating the operating energy normally consolidated (NC).
pile EP1 resulted in an expansion (negative strain The experimental program was divided into two
variation) of its thermally active portion and a com- parts, one devoted to characterise the material from
pression (positive strain variation) of its thermally a thermal point of view and the other to study the
inactive top portion because of the entrapment with effects of thermal cyclic loading. This corresponded
the slab. The aforementioned interactions eventually to two different stress-temperature paths: (i) oedomet-
induced an expansion of the non-operating energy pile ric tests at different constant temperatures (20, 40 and
EP2 that was comparable in its lower portion to the 60◦ C) and (ii) thermal cycles under constant vertical
expansion of EP1. This deformation was also greater effective stress.
14
Figure 10. Soil response at different temperature changes.
Figure 11. Soil response during thermal cycles.
More detailed information on the experimental
setup, materials and methods can be found in Di Donna
and Laloui (2015).
3.1.2 Observed behaviour: monotonic temperature

changes
In accordance to data available in the literature, it was
shown that clayey soils are characterised by thermal
softening when subjected to increasing temperature
changes (cf., Figure 10). The results also confirmed
the known effects of temperature on the primary con-
solidation coefficient and the hydraulic conductivity:
they increase with temperature leading to a faster
consolidation. This is mainly linked to the effects of
temperature on water viscosity. Figure 12. Development of the shear box for soil-concrete
investigation at different temperatures: (a) electrical heating
3.1.3 Observed behaviour: cyclic temperature tissue, (b) installation of the tissue in the lower part of the
changes shear box, (c) support for the concrete specimen, (d) position
The results confirmed that clayey soils show thermo- of the concrete specimen, (e) initial position of the upper part
elastic response if overconsolidated (OC) and thermo- of the box containing the soil and (f) electrical power supplier
elastic thermo-plastic response if normally consoli- and insulation system.
dated (NC) (cf., Figure 11). In the latter case, it was
shown that these materials undergo most of the thermal
3.2.1 Experimental setup, material and methods
plastic deformation during the first heating-cooling
The device that was employed for the experiments is a
cycle, followed by an accommodative behaviour dur-
direct shear box produced by GDS Instruments™. This
ing the subsequent ones. Increments of irreversible
apparatus was modified for reproducing the in-situ
deformation are observed in the thermal cycles succes-
behaviour of the energy pile-soil interface conditions
sive to the first one, which become smaller and smaller
at the laboratory scale (cf., Figure 12).
cycle after cycle until stabilisation. In the end, the
The experimental campaign included tests on sand-
material’s stress state tends to remain inside the elas-
concrete and clay-concrete interfaces at different tem-
tic domain showing a thermo-elastic expansion and
peratures, as well as tests on soil-soil specimens. For
contraction during heating and cooling, respectively.
both types of interfaces, different degrees of concrete
roughness were tested as well as monotonic and cyclic
3.2 Response of concrete-soil interfaces subjected
stress paths under constant normal loads (CNL) and
to temperature changes
constant normal stiffness (CNS) conditions.
The response of soils subjected to monotonic and More detailed information on the experimental
cyclic temperature changes was recently investigated setup, materials and methods can be found in Di Donna
by Di Donna et al. (2015). et al. (2015).
15
the most significant thermal effect resulted from an
increase in the adhesion between the two tested materi-
als. This result was related to the thermal consolidation
of the clay, which results in an increase of the contact
surface between the two materials. The same effect
was shown for the high and medium rough inter-
face, with the second interface having lower adhesion
at ambient and high temperatures due to its smaller
asperities. In addition, the same response obtained at
high temperature was also qualitatively shown at ambi-
ent temperature under the same testing conditions but
applied to the soil sample over a consolidated state.
4 ANALYSIS AND DESIGN OF ENERGY PILES
4.1 Geotechnical and structural performance

Several numerical analyses were recently performed
for investigating the geotechnical and structural per-
formance of energy piles at both short- and long-
Figure 13. Sand-concrete interface (high roughness): com- term periods. Those include finite difference and
parison between results at 20 and 50◦ C. finite element analyses. The former technique was
applied to consider the mechanical response of sin-
gle energy piles when subjected to the mechanical
and thermal loads associated to their structural sup-
port and geothermal operations (Knellwolf et al., 2011;
Mimouni and Laloui, 2014). The latter technique was
applied for the same purpose to single energy piles
(Laloui et al., 2006; Rotta Loria et al., 2015) but has
shown a remarkable potential for the analysis of energy
pile groups (Di Donna et al., 2016; Rotta Loria and
Laloui, 2016).
Different mathematical formulations and extents of
coupling between the variables governing the consid-
ered problem were used. The choice of these tools
was primarily based on both the goals of the study
and the aspects that wanted to be considered in the
investigation.
Rotta Loria et al. (2015) remarked that the mechan-
ical behaviour of energy piles can be crucially affected
by significant magnitudes of thermal and mechanical
loads. Plastic strain can occur in these conditions at
the pile-soil interface, involving a redistribution and/or
variation of forces in this setting that causes a shift of
Figure 14. Clay-concrete interface (medium roughness): the null point, i.e., the setting where zero thermally-
comparison between results at 20 and 50◦ C. induced displacements occur in the energy pile when
this element is subjected to a temperature change (cf.,
3.2.2 Observed behaviour Figure 15). The location of null point is key for the
The sand-concrete interface behaviour did not appear mechanical analysis of energy pile-related problems
to be affected by temperature changes based on both because allows the end-restraint conditions and the dis-
monotonic and cyclic (cf., Figure 13) interface shear placement and stress fields of the pile-soil system to
tests results. This result was expected because sandy be analysed in detail.
soils are known to behave thermo-elastically, with- In most cases of practical interest, the mechanical
out showing any particular relationship between their behaviour of energy piles subjected to the serviceabil-
strength and temperature. ity mechanical and thermal loads appears however to
Conversely, the response of the clay-concrete inter- be well captured by linear thermo-elasticity theory.
face changed at different temperatures, indicating an This theory, in particular, is considered a sufficiently
increase in strength with increasing temperature based accurate and expedient tool for both research- and
on both monotonic and cyclic (cf., Figure 14) interface engineering-oriented analyses of energy pile-related
shear test results. Interestingly, the interface friction problems. The successfulness of numerical analyses
angle slightly decreased at higher temperatures and exploiting linear thermo-elasticity has been recently
16
Figure 16. Thermal power extracted from the ground for
different pipe configurations.
from the configuration of the pipes. However, a length-

ening or shortening of the energy pile resulted in
markedly diverse responses of the foundation to the
Figure 15. Null point movements in a semi-floating energy thermo-mechanical loads, depending on the impact
pile subjected to different combinations and magnitudes of that the different mechanical and thermal properties
thermal and mechanical loads. of the surrounding soil layers may have had on the
bearing response of the pile.
demonstrated for energy pile groups by Rotta Loria The mass flow rate of the fluid circulating in the
and Laloui (2016). pipes remarkably influenced only the energy perfor-
mance of the foundation.
4.2 Energy performance Usual mixtures of a water-antifreeze liquid circu-
lating in the pipes did not markedly affect both the
The energy performance of energy piles was recently energy and the geotechnical performance of the pile.
analysed by Batini et al. (2015) in a coupled thermo-
mechanical study that also considered the geotechnical
performance of these foundations. The investigation 5 CONCLUDING REMARKS
was based on finite element sensitivity analyses that
were performed with reference to a base-case design This paper proposed a number of insights on the mul-
configuration characterising one of the energy piles of typhisical phenomena and mechanisms involved with
the EPFL. This study outlined the impacts of different energy piles. Some aspects of the thermo-mechanical
design solutions such as the (i) pipe configurations, behaviour of both single and groups of energy piles
(ii) foundation aspect ratios, (iii) mass flow rates of were presented and discussed. Comments on the
the fluid circulating in the pipes and (iv) fluid mix- response of soils and concrete-soil interfaces sub-
ture compositions on the energy and geotechnical jected to temperature changes were also proposed.
behaviour of the energy piles. Insights on some concepts that appear key for the
It was observed that the pipe configuration strongly analysis and design of energy piles were finally pre-
influenced both the energy (cf., Figure 16) and the sented. This work highlighted the significant impact
geotechnical performance of the energy piles. It was that the thermal and mechanical loads associated to
noted that the W-shaped pipe configuration resulted the geothermal and structural support operations of
in an increase of up to 54% in the heat transfer rate energy pile foundations have on the performance of
compared with the single U-shaped configuration at these ground structures.
the same flow rate. The double U-shaped pipe con-
figuration, which possessed a double flow rate with
respect to the other configurations, resulted in the REFERENCES
highest cooling of the concrete with the greatest related
stress and displacement distributions. Therefore, it was Batini, N., Rotta Loria, A. F., Conti, P., Testi, D., Grassi, W. &
considered to be a less advantageous solution with Laloui, L. (2015) Energy and geotechnical behaviour of
respect to the W-shaped pipe configuration both from energy piles for different design solutions. Computers and
Geotechnics, 86, 199–213.
a thermo-hydraulic and a geotechnical point of view. Di Donna, A., Ferrari, A. & Laloui, L. (2015) Experimen-
The foundation aspect ratio also played an important tal investigations of the soil-concrete interface: physi-
role in this context. The increase of the founda- cal mechanisms, cyclic mobilisation and behaviour at
tion aspect ratio resulted in an approximately linear different temperatures. Canadian Geotechnical Journal,
increase of the exchanged heat that was independent 10.1139/cgj-2015-0294.
17
Di Donna, A. & Laloui, L. (2015) Response of soil subjected exchanger pile. International Journal for Numerical and
to thermal cyclic loading: experimental and constitutive Analytical Methods in Geomechanics, 30, 763–781.
study. Engineering Geology, 190, 65-76. Mimouni, T. & Laloui, L. (2014) Towards a secure basis
Di Donna, A., Rotta Loria, A. F. & Laloui, L. (2016) Numer- for the design of geothermal piles. Acta Geotechnica, 9,
ical study on the response of a group of energy piles 355–366.
under different combinations of thermo-mechanical loads. Mimouni, T. & Laloui, L. (2015) Behaviour of a group of
Computers and Geotechnics, 72, 126–142. energy piles. Canadian Geotechnical Journal, 52, 1913–
Knellwolf, C., Peron, H. & Laloui, L. (2011) Geotechnical 1929.
analysis of heat exchanger piles. Journal of Geotechnical Rotta Loria, A. F., Gunawan, A., Shi, C., Laloui, L. &
and Geoenvironmental Engineering, 137, 890–902. Ng, C. W. (2015) Numerical modelling of energy piles
Laloui, L., Moreni, M. & Vulliet, L. (2003) Comportement in saturated sand subjected to thermo-mechanical loads.
d’un pieu bi-fonction, fondation et échangeur de chaleur. Geomechanics for Energy and the Environment, 1, 1–15.
Canadian Geotechnical Journal, 40, 388–402. Rotta Loria, A. F. & Laloui, L. (2016) Thermally-
Laloui, L., Nuth, M. & Vulliet, L. (2006) Experimental induced group effects among energy piles. Géotechnique,
and numerical investigations of the behaviour of a heat Submitted.
18
Geomechanical and numerical modeling of gas hydrate sediments
M. Sánchez & X. Gai

Texas A&M University, College Station, Texas, USA
ABSTRACT: Gas hydrates sediments are generally found in sub-marine sediments and in permafrost regions.
They are recognized as a huge potential energy resource. Methane hydrate deposits can lead to large-scale
submarine slope failures, blowouts, platform foundation failures, and borehole instability. Hydrate formation,
dissociation and methane production from hydrate bearing sediments are coupled Thermo-Hydro-Chemical and
Mechanical (THCM) processes that involve, amongst other, exothermic formation and endothermic dissociation
of hydrate and ice phases, mixed fluid flow and large changes in fluid pressure. The behavior of Hydrate Bearing
Sediments (HBS) is very complex and the mechanical modeling poses great challenges. The presence of hydrates
has a huge impact on the mechanical behavior of soils, affecting stiffenss, strength and dilatancy.A comprehensive
THM formulation for HBS is briefly presented in this paper. Special attention is paid to the mechanical behavior
of HBS. The model performance was very satisfactory in all the cases studied. It managed to capture very well
the main features of HBS behavior and it also assisted to interpret the behavior of this type of sediment under
different loading and hydrate conditions.
1 INTRODUCTION and geomechanical properties of hydrate sediments.

Depressurization, thermal injection and chemical
1.1 Background stimulation are the three basic mechanisms that trigger
hydrate dissociation. Current methods accepted as fea-
Hydrate bearing sediments (HBS) are natural soil sible for extracting methane gas from hydrate bearing
deposits that contain ice like methane hydrates in sediments are based on these three processes mainly.
its pore space. Methane hydrates are solid com- A proper modelling of this challenging problem will
pounds made of water molecules clustered around low assist to study optimal methane production strategies
molecular weight methane molecules. The stability of and also to prevent the multiple hazards associated
methane depends on pressure (P) and temperature (T) with uncontrolled hydrate dissociation and gas release
conditions. Methane hydrates forms are common in from hydrate sediments.
sub-permafrost layers and in deep marine sediments, Considering the complex multiphysics phenom-
this is because of the necessary conditions of low ena involved in gas hydrate formation/dissociation a
temperature and high pressure required for hydrate fully coupled Thermo-Hydro-Chemical and Mechani-
stability (e.g. Collett 2002). If changes in tempera- cal formulation is an indispensable component to con-
ture and/or pressure are such that the methane hydrate duct realistic engineering problems involving HBS.
shift from the ‘P-T’ stability zone, it will dissociate The mechanical modeling of gas hydrate sediments
producing methane and water. Hydrate dissociation is particularly challenging. In this paper an elasto-
will induce in turn profound changes in the sediment plastic model based on the strain partition concept
structure and physical properties. According to the (Pinyol et al. 2007) and the HIerarchical Single Sur-
U.S. Geological Survey, the estimated global natural face (HISS) framework (e.g. Desai 1996) was adopted
gas hydrate reserves are in the range from 100,000 to in an effort to provide a more general and versatile
about 300,000,000 trillion cubic feet (Mahajan et al. constitutive model for HBS. The model is very well
2007). Given the sheer magnitude of this resource, suited to simulate problems involving hydrate dissoci-
sediments with (a relatively) high concentration of ation. The proposed framework was widely validated
methane hydrate in the pore space are considered to be against recently published experiments involving both,
a significant energy resource for future exploitation. synthetic and natural hydrate soils, as well as different
However, hydrates dissociation is associated with sediments conditions (i.e., different Sh , and differ-
engineering problems, such as: blowouts; platform ent hydrates morphologies) and confinements. A brief
foundation failures; pipeline related issues; large sub- description of the THM formulation adopted in this
marine landslides; and borehole instability. Hydrate work is presented in the next section. Code validation
dissociation is a complex phenomenon involv- are reported elsewhere (e.g. Sanchez et al. 2015). Then
ing changes in fluid pressure, effective stresses, the mechanical model is briefly presented.
19
where, the gas constant R = 8.314 J/(molo K) and
the molecular mass of methane Mm = 16.042 g/mol
(example: ρg = 86 g/m3 at T = 280◦ K and Pg =
10 MPa).
The hydrate phase is made of water and methane.
The mass fraction of water in hydrate α = mw /mh
depends on the hydration number ξ for methane
hydrates. The ice transformation may take place dur-
ing fast depressurization. The densities of the hydrate,
the ice and the mineral phases are assumed constant.
2.3 Main equations

Figure 1. a) HBS schematic representation; b) HBS phase
diagram. The total volume is the sum of the partial volume of
each β-phase Vβ , where the subscript β is related to
2 COUPLED THCM FORMULATION the solid ‘s’, liquid ‘’, gas ‘g’, hydrate ‘h’ and ice
‘I ’ phases. Assuming that the solid mineral is a non-
2.1 Basics reactive phase, the total porosity (φ) is defined as the
ratio of the volume of voids to the total volume
The THM phenomena that take place in hydrate-
bearing sediments include the main following
phenomena: i) heat transport through conduc-
tion, liquid and gas phase advection; ii) heat of
formation-dissociation; iii) water flux as liquid phase; The following relationship can be written:
iv) methaneflux in gas phase and as dissolved
methane(diffusion in liquid phase); v) heat of ice
formation/thaw; vi) fluid transport of chemical
species; vii) mechanical behavior depending on effec- where S , Sg , Sh and Si are the liquid, gas, hydrate and
tive stress and hydrate concentration/morphology. ice saturations, respectively.
Those phenomena have been implemented in the finite The mathematical formulation is composed of three
element program CODE-BRIGHT (Olivella et al. main types of equations: i) balance equations (i.e.
1996). Some components of the coupled formulation mass balance of species, internal energy balance and
are presented in the subsequent sections. momentum balance equations); ii) constitutive equa-
tions; and iii) equilibrium restrictions. As an example
2.2 Phases and Species of the mass balance equations the mass balance of
The pores in the granular skeleton of HBS are filled water is presented below:
with gas, hydrate, water or ice (Fig. 1a). The three main
species mineral, water, and methane are found in five
phases: solid mineral particles, liquid, gas, hydrates
and ice as shown in the phase diagram in Figure 1b.
The liquid phase is made of water and dissolved gas.
The density of the liquid ρ depends on temperature T where ρh is the density of the hydrates, and qg is flux
[◦ K] and pressure P [MPa]. The asymptotic solution of methane.
for small volumetric changes is: Constitutive equations and equilibrium restrictions
relate the main unknowns (i.e. u, Pl ; Pg , Ph T ) with
the dependent variables (i.e. stresses; Sl ; Sg ; Sh fluxes).
For example, the retention curve dictates the rela-
tionship between the interfacial tension sustained by
where, ρo = 0.9998 g/m3 is the mass density of water the difference in liquid and gas pressures. Because of
at atmospheric pressure T is temperature in ◦ K and space limitations the full set of equations are not pre-
B = 2000 MPa is the maximum bulk stiffness of sented here, more details can be found elsewhere (e.g.
water (at 277◦ K), and βT = 0.0002◦ K−1 is the thermal Sanchez et al. 2015, 2016).
expansion coefficient.
The mass density of the gas phase is pressure Pg
[MPa] and temperature T [◦ K] dependent and it can 3 MECHNICAL BEHAVIOR OF HBS
be estimated using the ideal gas law modified for
methane gas. 3.1 Overview
The presence of hydrates strongly affects key mechan-
ical properties of soils. The samples containing
hydrates exhibit, higher shear strength, more dila-
tion under shearing, and they also soften more after
20
Figure 3. a) Schematic representation of the hydrate
Figure 2. Main types of hydrate morphology: a) cementa- damaged during shearing; b) rearrangement of the HBS
tion; b) pore filling; and c) supporting matrix. structure upon dissociation.
yielding when compared against the free hydrate sam- components of the proposed geomechanical model is
ples (Masui et al. 2008, Miyazaki et al. 2011). briefly introduced below.
Identical sediments but with different hydrate satura-
tions, generally shown an increases in HBS stiffness, 3.2 Constitutive model
pre-consolidation pressure, and sediment strength with
the increase of Sh . A degradation of the tangent stiff- The elasto-plastic framework contemplates the pres-
ness of hydrate-bearing soils during shearing has also ence of two basics components: sediment skeleton
been reported (e.g. Masui et al. 2005; Hyodo et al. and hydrates. The strain-partition concept proposed
2013). by Pinyol et al. (2007) was adapted for the case of
Hydrates morphology also impacts on HBS behav- HBS. Through this concept it is possible to account for
ior. Hydrates can be present in soils in three main the role of these two different structures on the global
types of pore habits, namely (e.g. Waite et al. 2009): response of HBS under different loading and hydrate
a) cementation (Fig. 2a); b) pore-filling (Fig. 2b); and saturation conditions, particularly during hydrate dis-
c) supporting matrix (Fig. 2c). sociation. Specific constitutive equations for these two
In the cementation mode the hydrates act as a bond- basic structural components can be proposed. For the
ing material at mainly on sediment grain boundaries sediment skeleton, a model based on critical state soil
and grow freely into the pore space without bridging mechanics was adopted. The particular constitutive
two or more particles together. The hydrates present in equation adopted was based on a modification of the
this morphology assists to the mechanical stability of HISS elasto-plastic model (Desai 1986). The proposed
the granular skeleton contributing to the load-bearing framework also incorporates sub-loading and dilation
framework of the sediment.The presence of hydrates in enhancement concepts. As for the hydrates, a damage
this case can strongly affects the sediment permeability model that considers the material degradation due to
and water storage capacity. loading and dissociation was suggested. Only some
The behavior of HBS upon dissociation is very com- basics components of the model are introduced below,
plex because their response not only depend on the a detailed description can be found elsewhere (e.g.
amount of hydrate, but also on the type of pore habit, Sanchez et al., 2015, 2016, and Gai & Sanchez, 2016).
and the stress level at which hydrate dissociation is The total volumetric strain (εv ) accounting for
induced. For example, when hydrate dissociation takes both, sediment skeleton and hydrate deformations (i.e.
place at a low deviatoric stress (i.e. lower than the subscript ss and h, respectively) can be calculated as:
strength of the pure sediment), the tendency of the
sediment after dissociation is to harden. An opposite
behavior was observed when dissociation occurs at a where Ch is the volumetric concentration of methane
higher deviatoric stress. Significant volumetric com- hydrate; which in turns is equal to the porosity times
pression deformations are observed when hydrate dis- the hydrate saturation (i.e., Ch = φSh ). The relation-
sociation is induced under constant effective stresses ships that link hydrates and soil skeleton strains are
(Santamaria et al. 2015). proposed following an approach similar to Pinyol et al.
It was also suggested that hydrate bonding effects (2007):
can be damaged during shearing (Uchida et al. 2012).
The progressive stiffness degradation in tests involv-
ing HBS is generally very evident. Figure 3a illustrates
the phenomenon of hydrate damage during shear- where χ is the strain partition variable that evolves
ing. Hydrate dissociation is also accompanied by during loading.
profound changes in the sediment structure. Figure As for the hydrates, previous studies suggested
3b shows schematically the expected changes in the that hydrate effects can be damaged during shear-
soil structure that lead to the collapse compression ing (Uchida et al. 2012). It is assumed that loading
deformations observed during dissociation under nor- degradation occurs when the stress state arrives to
mally consolidated conditions (as discussed later on, a predefined threshold value ‘r0 ’. When the stresses
Fig. 6b). are below a pre-established threshold, a linear elastic
In summary, the mechanical response of HBS is response of the material is assumed via the following
highly non-linear and complex, controlled by multiple relationships:
inelastic phenomena that depends on hydrate satura-
tion, sediment structure, and stress level. The main
21
where σh corresponds to the stresses taken by the
hydrate and Dh0 is the methane hydrate elastic consti-
tutive matrix of the intact material. Loading damage
takes place when the changes in the stress state is such
that the secant elastic energy reaches r0 . In this case the
damage variable L (i.e. +∞ > L ≥ 0) increases and the
stiffness reduces. The damage evolution is determined
by means of the function below (Carol et al. 2001):
Figure 4. Yield surfaces considered in the model.

where r1 controls the damage rate. The evolution law
for the partition variable is defined by: Table 1. Soil parameters adopted in the modeling of HBS
Test Test Test Test

Properties Sh = 0 Sh = 24.2% Sh = 35.1% Sh = 53.1%
M 1.30 1.30 1.30 1.30

where χ0 is an initial reference value assumed for the λ .16 .16 .16 16
partition variable. κ .004 .004 0.004 0.004
As for the sediment skeleton, a critical state con- pc (MPa) 10. 10 10 10
stitutive model based on a modified HISS framework n 3 3 3 3
was adopted. The model incorporates sub-loading con- a 1 1 1 1
cepts, as well as hardening and dilation enhancement γ −1/9 −1/9 −1/9 −1/9
mechanisms associated with the presence of hydrates Ch 0 .096 .138 .213
in the sediments. The modified HISS model involves α − 32 32 32
β − 1.0 1.0 1.0
a single and continuous yield surface that can adopt
r1 − 4.1 4.1 4.1
different shapes depending on the selected parameters r0 − 1e–5 1e–5 1e–5
(Desai 1986). The HISS yield surface (F) is given by: η 42 42 42 42
χ0 − 1 1 1
Kh (MPa) − 9600 9600 9600
Gh (MPa) − 4300 4300 4300
where a and γ are model constants; n is the parameter

related to the transition from compressive to dilative 3.3 Model applications
behavior; pss and qss are the mean effective and devi- The tests reported by Hyodo et al. (2013) were selected
atoric stresses, respectively, both associated with the to study the effect of hydrate saturation on the behav-
sediment skeleton; M is the slope of critical line in the ior of HBS. A series of triaxial compression tests
qss −pss space; and pc is the effective pre-consolidation on synthetic methane hydrate soils were conducted
pressure. The Modified Cam-Clay yield surface cor- at different (constant) hydrate saturations, namely:
responds to a particular case of this model. The yield Sh = 0; Sh = 24.2%; Sh = 35.1%; and Sh = 53.1%. All
function incorporating the strength enhancement (pd ) the samples were prepared at a similar porosity (i.e.
associated with the presence of methane hydrate can φ ∼ 40%). The effective confining pressure for all
be expressed as the tests was 5 MPa. The samples where isotropi-
cally consolidated first and then subjected to shear-
ing. The model parameters were determined using
back-analysis based on two tests, the one involv-
ing sediments without hydrates (i.e., Sh = 0) and the
test related to the highest hydrate saturation (i.e.,
Sub-loading concepts are incorporated in the formu- Sh = 53.1%). Then, this model (without modifying
lation to account for any irrecoverable strain that may the parameters adopted before) was used to pre-
occur in HBS when stresses are inside the yield sur- dict the behavior of the samples with Sh = 24.2 and
face, and also for having a smooth transition between Sh = 35.1%. Table 1 presents the parameters adopted
elastic and plastic states. The three yield surfaces con- in the analyses.
sidered in this model are presented schematically in Figures 5a & b show the comparisons between
Figure 4. experimental and model results for the different
The principle of virtual work was advocated to hydrate saturations in terms of deviatoric stress and
obtain the final expressions relating the external effec- volumetric strain versus axial strains. The compres-
tive stress σ with the total strain increment. (Sanchez sion behavior was dominant in all the samples, but
et al., 2015, 2016). the one with Sh = 53.1% showed a dilatant response
22
Figure 5. Comparisons between model and experimental Figure 6. Behavior during dissociation of natural HBS spec-
results for synthetic samples of HBS prepared at different imens under oedometric conditions: a) core 8P; and b) core
Sh : a) stress-strain behavior; and b) volumetric responses 10P, (experimental data from Santamarina et al. 2015).
(experimental data from (Hyodo et al. 2013).
structure is huge during dissociation. The main rea-

with a slight stress-softening behavior. The relatively sons behind this different behavior can be related to:
high confining pressure at which the tests were per- i) hydrate saturation is much smaller in core-8P than
formed (i.e. σc = 5 MPa) could be one reason for the core-10P (i.e., Sh = 18% for core-8P, and Sh = 74%
predominant hardening behavior with positive volu- for core-10P); ii) the vertical stress at which hydrates
metric strains observed in these tests. In all the tests are dissociated is lower and the dissociation took
the initial stiffness and shear strength increase with Sh . place under over-consolidated conditions, therefore
The model was able to match very well the stress-strain the effect of confinement on the re-accommodation
curves for all the experiments under study. Quite good of the sediments particles is less significant; and iii)
agreements were also observed in terms of volumetric this sample was previously loaded up to a very high
behavior (Fig. 4b). effective vertical stress (i.e. σv = 6 MPa) that degraded
The ability of the model to simulate tests involv- the bonding effect of the hydrate and induced impor-
ing different hydrate morphology was also checked tant changes in the sediment structure previous to
using the triaxial experiments conducted by Masui dissociation.
et al. (2008), more details can be found elsewhere
(Gai & Sanchez, 2016; Sanchez et al., 2016).
The tests conducted by Santamarina et al. (2015)
were selected to study the effect of hydrate disso- 4 CONCLUSIONS
ciation under loading conditions. Two natural core
samples were extracted from the Nankai Trough using The interest for gas hydrates have increased in the
Pressure Core Characterization Tools (Santamarina last few years. They represent huge opportunities in
et al. 2012) and tested under oedometric conditions. terms of energy resource and they are also associ-
Figure 6a shows the comparison between experimen- ated with some drawbacks because of the metastable
tal and modeling results for the sample coded as character of the HBS structure and its impact on sedi-
‘core 8P’ with Sh = 18%. Prior to hydrate dissocia- ment stability. A coupled THCM formulation for HBS
tion the specimen was loaded up to a vertical stress was briefly presented in this paper. A key compo-
σv = 6 MPa and then unloaded back to σv = 3 MPa nent of the proposed approach is the geomechanical
to study the stress-volume response of HBS. Under model. An advanced constitutive model for HBS was
this over-consolidated conditions, hydrate dissocia- also presented and applied to reproduce the mechani-
tion was induced. Once the hydrate fully dissociated, cal behavior observed in recent experiments involving
the sample was subjected to loading-unloading cycles both synthetic and natural HBS specimens. The model
with a maximum σv = 9 MPa. Figure 6b presents the was able to reproduce quite satisfactorily the main fea-
results related to specimen coded as ‘core 10P’, initial tures of soil behavior observed in these tests as, for
Sh = 74%. This sample was loaded until σv = 3 MPa, example: the enhancement in stiffness and strength
at this normally-consolidated conditions, the effec- induced by the presence of the hydrate, stiffenss
tive stress was hold constant and hydrate dissociation degradation during shearing, soil dilatancy, and the
was induced. After hydrate dissociation, the sample volumetric soil collapse compression observed during
was loaded up to σv = 9 MPa and then unloaded. hydrate dissociation at constant stresses.
The model managed to capture very satisfactorily the
main trends observed in both tests. The yield stress
and unloading-reloading behavior are quite well mod- ACKNOWLEDGEMENTS
eled in both specimens. It is worth to highlight the
model ability to reproduce the difference in volumetric The authors would like to acknowledge the finan-
strains observed during dissociation at constant stress cial support from NETL (National Energy Technol-
in these two tests. In the case of core-8P, the collapse- ogy Laboratory), DOE (Department of Energy, US)
compression behavior is significantly less marked than through Award No.: DE-FE0013889. The authors
in core-10P, where the rearrangement of the HBS would also like to acknowledge the other researcher
23
involved in this project, amongst others: J. Carlos artificial gas hydrate bearing sediments. Proceedings 6th
Santamarina; Ajay Shastri and Mehdi Teymouri. International Conference on Gas Hydrates ICGH 2008,
Vancouver, British Columbia, Canada
Miyazaki, K., Masui, A., Sakamoto,Y., Aoki, K., Tenma, N. &
REFERENCES Yamaguchi, T. 2011. Triaxial compressive properties of
artificial methane-hydrate-bearing sediment. Journal of
Carol, I., Rizzi, E. & Willam, K. 2001On the formulation Geophysical Research: Solid Earth 1978–2012 116(B6).
of anisotropic elastic degradation. I. Theory based on Olivella, S., Gens, A., Carrera, J. & Alonso E.E. 1996.
a pseudo-logarithmic damage tensor rate. International Numerical formulation for a simulator (CODE-BRIGHT)
Journal of Solids and Structures 38(4): 491–518 for the coupled analysis of saline media. Engineering
Collett, T.S. 2002. Energy resource potential of natural gas Computations, 13(7): 87–112.
hydrates. AAPG bulletin 86(11):1971–92. Pinyol, N., Vaunat, J. & Alonso, E.E. 2007. A constitu-
Desai, C.S., Somasundaram, S. & Frantziskonis, G. 1986. A tive model for soft clayey rocks that includes weathering
hierarchical approach for constitutive modelling of geo- effects. Géotechnique; 57(2): 137–151.
logic materials. International Journal for Numerical and Sánchez, M., Santamarina, J.C., Gai X. & Sun Z. 2015.
Analytical Methods in Geomechanics 10(3): 225–57. Quarterly Research Performance Progress Report (Period
Gai, X. & Sánchez, M. 2016. Mechanical Modeling of ending 03/31/2015)
Gas Hydrate Bearing Sediments Using an Elasto-Plastic Sánchez, M., Santamarina, J.C. & Shastri, A. (2016). “Cou-
Framework. Environmental Geotechnics (accepted). pled THM Analysis of Gas Hydrate Bearing Sediments”
Hyodo, M., Yoneda, J., Yoshimoto, N. & Nakata, Y. under review.
2013. Mechanical and dissociation properties of methane Santamarina, J.C., Dai, S., Terzariol, M., Jang, J., Waite, W.F.,
hydrate-bearing sand in deep seabed. Soils and founda- Winters, W.J., Nagao, J., Yoneda, J., Konno, Y., Fujii, T. &
tions. 53(2): 299–314. Suzuki, K. 2015. Hydro-bio-geomechanical properties of
Mahajan, D., Taylor, C.E. & Mansoori, G.A. 2007. An intro- hydrate-bearing sediments from Nankai Trough. Marine
duction to natural gas hydrate/clathrate: The major organic and Petroleum Geology.
carbon reserve of the Earth. Journal of Petroleum Science Uchida, S., Soga, K. &Yamamoto, K. 2012. Critical state soil
and Engineering 56(1): 1–8. constitutive model for methane hydrate soil. Journal of
Masui, A., Haneda, H., Ogata, Y. & Aoki, K. 2005. Effects of Geophysical Research: Solid Earth 1978–2012, 117(B3).
methane hydrate formation on shear strength of synthetic Waite, W.F., Santamarina, J.C., Cortes, D.D., Dugan, B.,
methane hydrate sediments. 15th International Offshore Espinoza, D.N., Germaine, J., Jang, J., Jung, J.W., Kneaf-
and Polar Engineering Conference. sey, T.J., Shin, H. & Soga, K. 2009. Physical properties
Masui, A., Miyazaki, K., Haneda, H., Ogata, Y. & Aoki, of hydrate-bearing sediments. Reviews of Geophysics 1
K. 2008. Mechanical characteristics of natural and 47(4).
24
Coupled chemo-mechanics: A comprehensive process modeling

for Energy Geotechnics
T. Hueckel
Duke University, Durham, NC, USA
L.B. Hu
University of Toledo, Toledo, OH, USA
M.M. Hu
University of Sydney, Sydney, Australia
ABSTRACT: Geo-chemistry and geo-mechanics have had a fruitful period of independent development over
last 50 years, and are now conceptually matured having produced useful and sophisticated modeling tools used in
the energy industry. As the understanding of the processes becomes more in-depth, interdependence of chemical
and mechanical phenomena in geo-materials are more and more appreciated. The primary variables/properties
that are involved in the interdependence are mineral, ion and proton mass transfer between porous medium
phases and mechanical stiffness and deformation, and soil permeability.
The paper outlines the efforts in modeling of coupling of geochemical reaction laws and laws of deforma-
tion and fluid flow. It identifies changes of mass of ion concentration, electrical charge, or individual solid
minerals as primary chemical variables affecting the constitutive mechanical laws of behavior of rocks and soils,
as well as of the fluid or species transport in the porous media. Examples of chemo-plasticity and chemo-elasticity
are discussed. A particular case of chemical softening or hardening as a function of mineral mass removed or
added is focused on. Additional chemo-mechanical coupling comes via micro-fracturation, which generates an
extra specific surface area of fluid/solid interface, at which an enhanced dissolution, and/or precipitation takes
place. Formulations of permeability evolution due to chemo-plastic process are also reviewed. Experimental
data and simulation results are discussed for processes exemplified by reservoir sediment aging/pressure solu-
tion and compaction, acidization enhanced subcritical crack propagation in hydraulic fracturing, or due to CO2
injection. Importance of multi-scale considerations (both for time and space scales), as well as of innovation in
chemo-mechanical experimentation are emphasized.
1 INTRODUCTION three decades of experience in thermo-mechanics

mainly in the context of nuclear waste disposal and
Simultaneous emergence of several new energy tech- geothermal energy, there is virtually nothing in terms
nologies that are related to energy extraction either as of chemo-mechanics as applied to energy geotechnics.
fluids or minerals from the earth, or long- or short-term Geo-chemistry and geo-mechanics have had a fruit-
storing or retrieving energy resources or by-products ful period of independent development over last 50
in or from the ground, gave a very strong impetus years, in which they both conceptually matured and
to development in geotechnics, including some basic produced useful and sophisticated modeling tools
areas. Most of the new technologies require rigor- of prediction of reservoir compaction or diagenesis
ous production predictions, as well as need to meet in petroleum engineering, contaminant isolation in
demanding safety requirements, together with long- nuclear waste disposal design, landfill settlement pre-
term performance assessment. All of the above vastly diction, clay swelling, coastal landform stability or
rely on experiments based numerical simulations. CO2 mineralization, to mention a few areas of rele-
However, basic science and often knowledge data base vance. However, as the understanding of the processes
is by far insufficient to provide confident products. became more in-depth, it was realized that the interde-
Traditional geotechnics only sporadically dealt with pendence of chemical and mechanical phenomena in
problems resulting from energy production issues, geo-materials is often of the first order, both in nature
which pose problems related to heating/cooling of and in engineering and goes beyond a p-V -T frame-
soil/rock and may involve chemical changes affecting work (Bayly, 1992). The primary variables/properties
soil/rock mechanical properties. While there is nearly that are involved in the interdependence are chemically
25
affected mineral, ion and proton mass transfer between 3 CONSTITUTIVE EQUATIONS
porous medium phases and mechanical stiffness and
deformation, as well as rock/soil permeability. To account for the role of chemical changes in mechan-
The lecture outlines the efforts in modeling of cou- ical material response one needs to formulate a set of
pling of geochemical reactions and processes and laws constitutive hypotheses on how to couple the mechan-
of deformation and fluid flow. It identifies changes ics of the soil solids to their chemistry, and in particular
of mass of ion concentration, electrical charge, or to the evolution and possibly transport of chemical
individual solid minerals as primary chemical vari- species. To start with let us note that to describe total
ables affecting the constitutive mechanical laws of free energy (isothermal process) in two-phase, multi-
behavior of rocks and soils, as well as of the fluid species reactive porous medium, the work by total
or species transport in the porous media. Reference stress during deformation process must be supple-
to thermodynamic principles is made. Examples of mented by the work of chemical potential µkK (mass
chemo-plasticity and chemo-elasticity are discussed. based, [J/g]) during addition of mass of species of both
A particular case of chemical softening or hardening phases, (k are indices for the fluid species, K are for
as a function of mineral mass removed or added is the solid species)
focused on. Additional chemo-mechanical coupling
comes via micro-fracturation, which generates an extra
specific surface area of fluid/solid interface, at which
an enhanced dissolution, and/or precipitation takes
place. Formulations of permeability evolution due Species may be entities of any kind, pore water,
to chemo-plastic process are also reviewed. Experi- adsorbed water, other fluid species, minerals, ions,
mental data and simulation results are discussed for even entire rock. Mass of species added to or removed
processes exemplified by aging/pressure solution in from the system and energy associated with the addi-
reservoir sediment subsidence, acidization enhanced tion/removal (i.e. chemical potential δµkK = δp K
ρkK
+
subcritical crack propagation in hydraulic fractur- RT
(M ) δ(lnx kK ) are the only variables related to the chem-
ing, or due to CO2 injection. The importance of mk
multi-scale considerations (both for time and space ical processes. In this definition, R = 8.31451 J/molK
scales), as well as of innovation in chemo-mechanical is the universal gas constant, T [K] the absolute tem-
(M )
experimentation. perature and mk is the molar mass of the species
(M )
k, e.g. for free pore water mw = 18 g, pK is pressure
or mean stress in phase K, assumed as equal in all
the species of the given phase, whereas xkW are molar
2 PHENOMENOLOGICAL AND fractions of the species k in phase W , which is actu-
ENGINEERING BACKGROUND ally the mass variable. We will use both the changes in
mass and or chemical potentials in reaction equations,
There are innumerous chemical processes and reac- as well as to quantify the effect that reactions make on
tions that affect the mechanical properties of geoma- the mechanical properties of soil/rock.
terials. Chief among them are removal or accretion To derive constitutive properties of the solid phase,
of mass of minerals within the pore space. This may inclusive of the adsorbed water, we shall adopt a strat-
occur through dissolution of precipitation. The result egy proposed originally by Heidug and Wong (1996).
of such mass loss/gain for the mechanical proper- The free energy of the solid phase is represented by the
ties is a strength increase or decrease, direct chemical difference between the total energy of the whole sys-
strain, usually volumetric expansion or compaction. tem of porous medium and that of free pore water, the
Also other properties of relevance may be affected, latter classically defined per unit volume of the fluid
such as thermal conductivity. phase as
There may be a change in ionic content of pore fluid
that is in contact with the materials solids. In materi-
als that are electrically charged, as clays, it produces
swelling or shrinkage, resulting often in a change in
strength and compressibility, as well as in hydraulic
and re-scaled with respect to unit volume of REV to
conductivity.
yield
The causes for the mentioned processes are diverse.
They may result directly from the technology, as injec-
tion of water or vapor, as in fracking, or in geothermal
technologies, injection of CO2 in technologies related
to its geological sequestration, or specific chemicals, where pW and vW are pressure and volume content,
for instance acids, as in fracking. respectively, of the fluid phase. Gibbs-Duhem relation-
Some processes that are of interest are natural pro- ship for fluid was used in the above derivation, which
cesses as a part of diagenesis, like pressure solution. restricts changes of all intensive variables, and for
Similar processes are involved in subsidence due to oil isothermal processes of pressure and chemical poten-
or gas extraction, such as resulting compaction. tial of species being exchanged in reactions. Recalling
26
the Heidug –Wong strategy, the free energy of the point of view of material failure and factor of safety it
reactive solid phase reads may be of critical importance.
Considering the plastic part of dissipation, with
a chemical component to it, and assuming for inde-
pendent reasons that diffusional dissipation, chemical
reaction dissipation and thermal dissipation, if any,
The elastic part of the free energy are independently each non-negative, its inevitable to
conclude that
requires that a reversible part of mass change, µel

kK is
identified, so that which to ensure positiveness of eq. 10 may require the
positivesess of each term separately, which implies a
sort of generalized normality rule,
Or through a Legendre transform one can employ

mixed form energy as a function of effective stress
and mass changes If we assume that the yield locus evolution includes
a chemical softening, when mineral mass is removed,
or hardening when it is added (but with no compensa-
tion rule for accretion).
As a result the chemo-elasticity equations can be

expressed alternatively via energy gradients
where q = 1/2 ∼s : s∼ , is an invariant of stress devia-
tor, while pc is an apparent preconsolidation pressure,
which is a value of the maximum past isotropic effec-
The modeling practice is to identify experimentally tive stress, but amplified or de-amplified by possible
main features of the chemo-elastic behavior and then chemical processes. M critical state parameter related
propose a corresponding expression consistent with to the internal friction angle, possibly also dependent
the equations (7) and (8). on the chemical processes (Hueckel, 1997). Hence,
A typical chemo-elastic law would contain two
particular features: a chemically induced (mass or con-
centration dependent) volumetric strain (e.g. osmotic
swelling) and again, mass or concentration change In both pc and M the chemical process is represented
dependent elasticity modulus. by an accumulated change of mass of a chemical
An example of such a law for swelling clays was pl
given by Loret et al. (2002) or Gajo et al. (2002) in the species, mkS . However, this variable is highly process
form as follows: specific, as will be discussed in the next paragraphs.
4 CHEMICALLY INDUCED SEDIMENT

COMPACTION
In this section we shall discuss modeling of chemically

induced compaction of sediment enhanced by inun-
dation with water and leading to mineral dissolution
Notably, the chemical potential evolution rule is and possibly decrease of permeability by precipitat-
needed to complement the description of the coupled ing mineral in the open pore space. Such a process is a
chemo-elastic process. component of diagenesis, in the from of pressure solu-
Chemo-plasticity formulation is needed to properly tion, as well as as that of reservoir compaction during
describe the effect of all dissipative processes and in extraction of oil or gas, when the emptied pores are
particular a number of chemical changes in the mate- filed with water, either in a natural process or during
rial on its mechanical processes. Physically, the most water injection.
important factor is that of chemical softening, as it may The natural process of normal consolidation of sed-
in many instances to be the most important factor. It iments is believed to be a superposition of several
describes the weakening of the material in terms of coupled processes regarding the sediment solid matrix
its strength and all its derivatives as a result of mass and pore fluids. The processes involved are: deforma-
removal and other concentration changes. From the tion of sediment grains, including plastic deformation,
27
Figure 2. Schematic of a rigid grain indentation into a
adjacent rigid-plastic grain.
solid grain mineral. Stress and strain are microstress

and microstrain at the scale of a fraction of the grain,
Figure 1. Scenario of multiple processes involved in chem-
ically induced compaction after Hueckel & Hu, 2009.
positive in compression.
The material of individual grain is taken as rigid
plastic. To describe mathematically a near contact fail-
especially near contact between grain asperities and ure one needs to identify a micro-stress (σij ) yield
corresponding locally smooth surface of another grain; locus, f (δij ) ≤ 0, within which no strain (εij ) occurs,
dilatant damage induced by the plastic yielding; infil- whereas at yielding, the strain rate is entirely irre-
tration of the dilatant zone by pore water; activation versible
of damage (microcracking) related internal interfaces
between pore liquid and solid; dissolution of minerals
at these interfaces; diffusion within the grain of the dis-
solved species away from the reaction sites; interpore
whereas
diffusion of the dissolved minerals in the pore water
and precipitation of those on the free surface of adja-
cent or remote grains with the consequent reduction
of pore space and permeability.
The above sequence of processes is common to The superimposed dot over a symbol denotes a
other phenomena such as aging (see Mitchell & time rate. Dilatancy damage at the sub-grain scale
Solymar, (1984), Hueckel et al. (2001), Hueckel is a critical variable for this concept as a vehicle of
et al. (2005)), or pressure solution in petroleum reser- the chemo-mechanical coupling. It is linked to micro-
voir modeling (Rutter (1976), Lehner (1995). Hu & cracking, which first, provides a connected network
Hueckel (2007) have developed a sequence of three- across the damaged part of the grain that becomes
scale models to simulate aging, and identified leading instantly permeated with water as a result of suc-
feedbacks and feedforwards of intersacale coupling to tion induced by dilatancy. Second, micro-crack walls
evaluate the evolution of meso-scale stiffness and perform new solid-fluid interface, which constitute a
meability (Hueckel & Hu, 2009). A summary of the source of dissolution of mineral species, that is solid
feedbacks and feedforwards involved in the process mass removal. Third, irreversible micro-slips occur
are redrawn in Figure 1. across micro-cracks. As a result of micro-slips and
As can be seen in the schematic the whole phe- mass removal, the yielding behavior of the material is
nomenon of compaction consists of a number of affected by two competing plastic - hardening mech-
coupled processes that are either parallel or sequen- anisms: deviatoric strain-hardening and mass removal
tial in time. They can be arranged to form alternative softening. Hence, pc which is the apparent preconsolia-
scenarios, which can be modeled, to select the most tion stress, i.e. isotropic size characteristics of the yield
reasonable one. locus, depends on two hardening parameters, which
pl
One such three-scale scenario for chemo-plastic- are mechanical (εq ) or chemical (ξ) in nature (Hueckel
diffusional processes is described in detail by Hu and 2002).
Hueckel 2007). In what follows only the main con-
cepts are presented. The models at the micro-scale
include stress-strain relationship, dissolution law, and
intra-grain diffusion. At the meso-scale the models
include intergranular diffusion with precipitation and Chemical softening parameter ξ ≤ 1 is an accumu-
the resulting increase of material stiffness. The macro- lated relative mass removal of a single mineral species
scale is not shown in figure 1. The micro-scale the that dominates the material strength, computed with
medium is a porous material of the solid grain miner- respect to the original total mass of that particular
als. Its porosity is the internal grain porosity, and the species. The relationship between change in the appar-
mechanical properties of the material are those of the ent preconsolidation stress pc and a reaction progress
28
variable is established empirically. A mineral in the f˙ (δij , εq ξ) = 0 (see e.g. Hueckel (2002) therefore it is
pl
example, dominating the strength of quartz sand is a function of rates of stress and reaction progress
silica. Dissolution reaction of silica in water may be
measured through a change in activity of its product,
which is silicic acid H4 SiO4 formed in the aqueous
solution. The rate of silica dissolution is determined
by Rimstidt and Barnes (1981) formula as propor-
tional to the specific surface area of the solid-fluid
interface A, normalized with respect to 1 m2 to yield a
non-dimensional quantity Ã
In particular, at constant stress, an irreversible strain

rate is generated proportional to the reaction rate. The
yield locus in the form of a set of linear functions for
single principal stress components has the advantage
where ai are activities, and γ i ,activity coefficients, of a very simple kinematics, yet capturing the essence
of i-th species, while k+ and k− are rate constants of straining,
of respectively forward and backward reactions. Ã is
a dimensionless specific interfacial surface area, as
above, per unit mass of pore fluid. For details see Hu
and Hueckel, 2007). ξ is a reaction progress variable
constrained by the inequalities: 0 ≤ ξ ≤ 1; when ξ = 1,
the reaction is completed, that is all silica is removed
from the material. where α and β are constants, σ3 remains undetermined.
In the current context the damage occurs within a As discussed earlier, ξ is the relative mass removal,
single grain and consists in opening of micro-cracks. as its rate is described by eq. (20). At this point the
A new scalar variable, ã, represents the amount of precipitation term is ignored, as it would lead to a con-
the added interface surface area per unit volume of stitutive non-linearity, making it impossible to solve
the grain medium. It is linked to the relative reaction the equation system semi-analytically.
area, Ã, The intra-grain diffusion of silicic acid is limited
to the damaged zone (in radial coordinates with axial
symmetry here), as shown in Fig. 3a. It is described
by a linear reactive diffusive transport law combined
with a mass balance law, with a reaction term (last
term of the RHS), which through its link couples it to
where ρ0 = 1 kg/m3 , ρw is the density of water, ng is deformation via eq. (21):
porosity of the grain solid. The new internal inter-
face surface area generated by the micro-cracking per
unit volume is proposed to be proportional to the
volumetric strain. Hence,
where XH4 SiO4 is molar fraction of aqueous silica in the

fluid phase within the grain. Assuming the pore fluid
to be a dilute solution, the molar fraction of any of its
where φ is a constant, whereas φc represents the spe- species k, has mass content mkF = MkF /V0 , linked to
cific surface area of pre-existing voids. For εv > 0; molar fraction, XkF through V0 is the reference volume
φ = 0 which aims at excluding the compressive strain of the entire grain medium. D is the solute diffusion
for which there are no micro-cracks and hence no coefficient.
change in dissolution surface area. Therefore The above system of constitutive equations is to be
supplemented by equilibrium equations and kinemat-
ics expression to yield displacements. Hu & Hueckel
(2007) provided a solution to the coupled chemo-
mechanical deformation-diffusion through what is
known as Johnson’ (1985) approximation for the
contact problem, by assuming it is axisymmetric.
Notably, precipitation at the dissolution site is Experiments with compressed pure silica grains in
neglected. The irreversible strain rate mode is contact with water provide evidence of formation of
determined by the associated flow rule, whereas the solid coating of silica polymer links, growing in
its magnitude by the plastic multiplier λ̇ results density and strength in time as short as 14–21 days
from the extended Prager’s consistency condition, (Guo & Hueckel, 2013, 2015).
29
Figure 3. (a) A cartoon showing intra-grain diffusion of the dissolved species; (b) inter-grain pore with the flux of the
dissolved silica; (c) pore fluid flow through a dissolution/precipitation altered meso-scale inter-grain pore system.
Figure 4. (a) Inter-grain porosity change (n), as originated from deformation of the grain (n1 ) and from precipitation induce
grain coating; (b) Flux of dissolved silica from a grain as a function of intra-grain mass transfer coefficient M = k+ a2 /Dx0 ,
where x0 is the initial concentration of silicic acid within the grain.
5 CHEMO-MECHANICALLY INDUCED its variable part of ua as a function of time depending

EVOLUTION OF PERMEABILITY on the dissolution rate constant k+ .
The rate of the formation of coating can be calcu-
In a similar way to the evolution of stiffness, the evo- lated as a change in the coating thickness, ḋc over an
lution of permeability is evaluated at the meso-scale, area, dS of the free surface, as certain number of moles
at which we consider as a REV a quadruplet of 1/4th of silica are precipitated in the water volume, dV,
grains, with equal radii, R, as shown in Fig. 3b.
There are two processes that affect the inter-grain
pore system as shown in Fig. 3b: the increase of the
contact size between two stressed grains due to chemo-
plastic deformation of the grain and the precipitation of
the solute within the pore. In terms porosity variation,
the effect of contact area increase is more significant, where p = Vvoid /Sfree = (2/π − 1/2)R, while υSiO2
as seen in Fig. 4a. However, it is an open question if the and υHO2 are the molar mass of SiO2 and H2 O, respec-
same is true for the permeability evolution. The most tively. XH4 SiO4 (θ, t) is the distribution of precipitating
important from the present perspective is the result aqueous silica determined from a steady state solu-
concerning the flux of dissolved mineral, Fig. 4b. As is tion of diffusion-precipitation transport eqution, where
seen it heavily depends on the reaction rate coefficient θ, −π/4 < θ < π/4 is an angular coordinate along
and the diffusion coefficient. the grain surface, with a grain mineral mass flux ±
The current radius of contact may be approximated fp (t) = Ja πa/4 p , as a boundary conditions at the con-
as a = a0 + ua where a0 is the initial radius of the contact points, while Ja (t) the mass flux across the inner
tact, while ua is the vertical indentation advancment. boundary of the grain, r = a (see Hu and Hueckel,
Fig. 5a shows the evolution of the radius a in terms of (2007) for details).
30
Figure 5. (a) Increment of the contact area (also grain penetration) as function of time and rate constant; (b) relative coating
rate of the walls of inter-grain pore (also of free surfaces of grains) for different values of constant inter-grain pressure.
Normalized with respect to grain size, R (calculated for k+ = 1 × 10−12 s−1 ).
Table 1. Characteristics dimensions of the flow conduit.
Name Symbol Value
grain size R 1 mm
estimated damage zone radius b 0.7 mm
size of the indent radius a0 = b/10 0.07 mm
asperity size = contact slit δc 0.05 mm
initial (nominal) pore opening δp0 0.86 mm
As a result of linearization of eq. (x), the coating

rate becomes uniformly distributed, i.e. independent
of θ.
The two variables shown in Fig. 5 control the overall

change in permeability of the pore system as visualized Figure 6. Schematic of an idealized flow conduit F1 across
in Fig. 4c: the mechanical displacement of the inner the inter-grain pore space.
radius of the pore (Fig. 5a) and the relative coating rate
ḋc /R/R plotted against time in Fig. 5b.
Notably their dependence with time is different. The
change in the contact area appears to accelerate with communicating with the adjacent inter-grain pore, as
time, whereas the coating rate increases nearly linearly, shown in Fig. 6. The intergrain slit has a very limited
after an initial period. opening, equivalent to a size of an asperity, or a piece
To evaluate the effect of the two factors, a simple of mineral debris locked in the contact space. As such
formula based on Poisseuille flow through two con- it will be assumed not to change during the process of
nected in series planar tubes of different thickness, is deformation, nor affecting the contact extent.
is employed (after Hueckel et al., 1997) in examining Clearly the vessel is a very simplistic representa-
the meso-scale structural changes. Given that gravity tion of the conduit. Especially the shape of the portion
plays an essential role in the stress pattern, in which representing the inner-grain pore, with its initial size
horizontal contacts are usually acted upon by twice as taken as an average between the pore entrance and its
big grain force than vertical contact, we shall limit our maximum, is a major simplification. However, as our
consideration to conduit denoted as F1 . The explicit goal is to capture main features of the evolution of the
data needed to calculate the permeability evolution are flow, following the earlier experience with such mod-
listed in Table 1. els (Hueckel et al. 1997) these simplifications seem
The flow is envisioned as occurring through a acceptable. The specific discharge of the two conduits
2D flat vessel composed of the intergranular slit connected in series, per unit area over which the vessel
31
is the only conduit, in this case 2Rx1, is obtained from
the rules of a series connection, which are
where p, pc and pp are total, contact and pore
fluid pressure difference between the exit and entrance
to the respective segment of the conduit, where the flow
velocities in the contact portion and pore portion, and
the one with respect to the entire cross section area,
are respectively equal to Figure 7. Evolution of permeability: K t (t)-intrinsic perme-
ability factor (left scale); ratio of the thickness of contact slit
to the pore size, δc /δp (right scale); integrated precipitation
pore coating dc /R.
where µ is dynamic viscosity of the pore fluid, and

K s is a resultant intrinsic permeability of the medium.
Combining (26) and (27) one arrives at the expression
for evolution of the specific discharge of the medium
expressed in terms of a variable resultant intrinsic
permeability,
Hence,
Figure 8. Stress–strain curve from Uniaxial Compression

Test on Gravina di Puglia calcarenite, dry (d1&2), wet with
water (w1&2), and with acid solution (w-d1&w-d2) (from
Ciantia et al., 2015).
the increase of stiffness is concerned (Hu and Hueckel,

where K t is a coefficient describing the evolution of 2007), while it is not felt in terms of permeability until
the intrinsic permeability due to chemo-mechanical it reaches a critical moment when porosity drastically
coupling in the medium, υi denote the molar mass of drops in a short time. A practical conclusion from the
species “i”. simulation concerns the critical time to clogging that
As may be easily anticipated from the form of the may be easily estimated from the mass dissolution rate
expression for K t , the evolution of the permeability needed to produce a solute to fill almost completely
will be not much sensitive to the effect of pore coating, the pore.
until well into an advanced stage when the size of the
pore approaches the size of the contact slit, as seen in
Fig. 7. 6 CHEMICALLY ENHANCED CRACK
The intrinsic permeability factor (time dependent) PROPAGATION
is seen to decline slightly (30%) in the first 1200 hours
mainly due to the increase of the contact area, whereas Subcritical crack propagation resulting from a spon-
it seems to be unaffected by the pore wall coating taneous or engineered change in the rock chemical
by precipitating mineral. This abruptly changes in a environment is of relevance in several energy tech-
short time, when the pore is practically clogged by the nologies, among which are unconventional oil and
precipitate, and hence its size becomes of the order, gas recovery, and enhanced geothermal systems. The
and then even smaller than the intergranular contact enhancement consists of a combination of fluid pres-
slit. So, within a few days the permeability drops over sure and injection of acids. Acid chemically softens the
one order of magnitude, and more. It is therefore con- material, which occurs relatively quickly, especially
cluded that precipitation is felt very quickly as far as in carbonate rocks. The main question is to correlate
32
Figure 9. Circumferential stress distribution near the crack tip, for a chemo-elastic behavior with coupled chemical shrinkage
coefficient, dependent on the shear strain invariant for a) a chemically swelling rock (silicate) and b) a chemically shrinking
rock (carbonate).
the chemical flux to the rate of crack propagation. be physically most justified and acceptable under the
Hu and Hueckel (2013, 2014) addressed the effect assumption of elasticity (if limitation to monotonic
of mineral mass removal on the material strength, process is imposed). The key step is introduction of
via coupled chemo-plasticity approach. The chemical the Airy stress function φ (Airy, 1863) defined as
part of the processes being explicitly rate-dependent,
requires plasticity to be treated incrementally and
iteratively. Simplified calculations with Extended
Johnson approximation (all fields are axially symmet-
ric around the crack tip point) makes it possible to
follow the stress evolution as minerals are dissolved.
The approach is analogous to that used in the problem
of indentation discussed in the previous paragraphs,
and will not be elaborated here.
To investigate the effect of coupling of chemicals on Substituting the stress-strain relationship into the usual
elasticity in the vicinity of a crack subject to acidizing strain compatibility equation, with the use of equilib-
requires quite different tools (Hu and Hueckel, 2016). rium equations, we can eventually obtain fourth order
The release of mineral mass in a reaction into liquid equation for the single variable of the Airy function φ.
phase affects solute diffusion, while the rate of mass
release is dependent on local acidity. As embodied
through equation (9) there are two main mechanical
responses to the mass removal. There is an addition- where chemical shrinkage coefficient is a linear func-
ally induced (stress independent) strain, often seen to tion of deviatoric strain invariant.
be proportional to the mass removal, with the pro-
portionality (chemical deformation) coefficient, likely
dependent on the material damage. A good example
here is osmotic swelling in shale, or swelling of silica The solution of the set of equation follows the one
in contact with water. In contrast in carbonates water used for thermo-elasticity (Saad, 2005) and can be
induces shrinkage of rock (Ciantia and Hueckel, 2013. found in Hu and Hueckel, 2016.
The second effect results from the change in elastic The most significant finding concerns a substan-
stiffness modulus (shear or/and isotropic). Figure 8 tial difference in the effect of acidized water on the
shows a decrease in elasticity modulus, between a subcritical crack propagation in sandstone and carbon-
material that is dry, wet and wetted with acid water ate, the former one elastically swelling and the latter
(Ciantia et al. (2015)). one shrinking in response to acidized water injection.
In what follows only the first aspect is discussed. Comparing the symptomatic distribution of circumfer-
A particular form of the chemical shrinkage coeffi- ential stress in front of the crack tip, one finds that the
cient is chosen, as linearly dependent on deviatoric stress is much higher for the shrinking rock than for
strain invariant to simulate the effect of microcracking chemically swelling rock, Figure 9a and b. Hence, sil-
(Hu & Hueckel, 2016). Such dependence was found to icate rocks require much higher base pressure or acid
33
concentration than carbonates to activate subcritical Hu M.M. and T. Hueckel, 2013, Environmentally Enhanced
propagation. Crack Propagation in a Chemically Degrading Isotropic
Shale, Geotechnique, SIP 2013, 63, 4, 313–321.
Hu M.M. and T. Hueckel, 2016, A chemo-elasticity cou-
pling in an acid enhancement modeling of pressurized
7 CONCLUSIONS crack propagation, Geomechanics for Energy and the
Environment, v. 7, 48–57.
Chemo-mechanical coupling is widely engineered, or Hueckel, T. 1997, Chemo-plasticity of Clays Subjected to
when naturally occurring it needs to be dealt with in Flow of a Single Contaminant and Stress, International
the energy industry to enhance the recovery and pro- Journal for Numerical and Analytical Methods in Geome-
ductivity. However, the engineering practice is widely chanics, 21, 1, 43–72.
based on costly experience and errors often leading Hueckel, T., M. Kaczmarek and P. Caramuscio, 1997, The-
to disasters. The attempts were presented to show oretical Assessment of Fabric and Permeability Changes
a potential for developing predictive tools, allowing in Clays Affected by Organic Contaminants, Canadian
Geotechnical Journal, 34, 4, 588–603.
for an optimalization of engineering solutions with Hueckel, T., G. Cassiani, Fan Tao, A. Pellegrino and V. Fiora-
environmental constraints in mind. It is clear that vante, 2001, Effect of aging on compressibility of oil/gas
such predictable power can only be achieved if an bearing sediments and their subsidence, J. of Geotechnical
appropriate, and appropriately designed experiments and Geoenv. Eng, ASCE, 127, 11, pp. 926–938.
are available. This work should help to design the Hueckel T., 2002, Reactive plasticity for clays during dehy-
experiment. dration and rehydration. Part I: Concepts and options, Int
J Plasticity; 18: 281–312.
Hueckel, T., Cassiani G., Prévost J.H. and Walters D.A.
REFERENCES 2005, Field Derived Compressibility Of Deep Sediments
of Northern Adriatic, Proceedings Seventh International
Airy, G. B. 1863, “On the Strains in the Interior of Beams”. Symposium on Land Subsidence (Sisols 2005), Shanghai,
Philosophical Transactions of the Royal Society 153: China, October 2005, Spec. Vol., F.B.J. Barends et al., eds.,
49–80. 35–50.
Bayly, B. 1992. Chemical Change in Deforming Materials Hueckel, T. and Hu, L.B. 2009, “Feedback mechanisms
Oxford UP, New York. in chemo-mechanical multi-scale modeling of soil and
Ciantia, M.O., Castellanza R., Crosta, G.B. and T. Hueckel, sediment compaction”, Computers and Geotechnics, 36,
2015, Effects of mineral suspension and dissolution on 934–943.
strength and compressibility of soft carbonate rocks, Johnson K.L., 1985, “Contact Mechanics”. Cambridge:
Engineering Geology, 184, 1–18. Cambridge University Press.
Gajo, A., B. Loret, and T. Hueckel, 2002, Electro- Lehner F.K. 1995, “A model for intergranular pressure
chemo-mechanical coupling in saturated porous media: solution in open systems” Tectonophys, 245:153–70.
elasto-plastic behaviour of heteroionic expansive clays, Loret, B., T. Hueckel, A. Gajo, 2002, Chemo-mechanical cou-
International Journal of Solids and Structures, 39, 4327– pling in saturated porous media: elasto-plastic behaviour
4362. of homoionic expansive clays, International Journal of
Guo R. and T. Hueckel, 2015, Silica polymer bonding of Solids and Structures, 39, 2773–2806.
stressed silica grains: an early growth of intergranu- Mitchell, J.K. and Solymar, Z.V. 1984, Time-dependent
lar tensile strength, Geomechanics for Energy and the strength gain in freshly deposited or densified sand.
Environment, v. 1, 1, 48–59. J. Geotech. Engrg., ASCE, 110(11), 1559–1576.
Guo R. and T. Hueckel, 2013, Growth of Silica Polymer Rutter E.H., 1976, The kinetics of rock deformation by
Micro-Structures within Stressed Intergranular Contacts pressure solution, Phil. Trans R Soc A, 283:203–19.
in Sands: A Chemo-Mechanical Coupling, Geotechnique Rimstidt J.D., Barnes D.L., 1980, The kinetics of silica water
SIP 2013, 63, 4, 322–330. reaction” Geochim Cosmoch. Acta; 44, 1683–99.
Heidug W.K. and Wong, S. W., 1996, Hydration swelling of
water absorbing rocks: a constitutive model, Int. Jnl. Num.
and Anal. Meth. in Geomech., 20, 402–430.
Hu L.B. and Hueckel T. 2007, Coupled chemo-mechanics
of intergranular contact: toward a three-scale model,
Comput. Geotech, 34(4):306–327.
34
Thermomechanical properties of a new small-scale reinforced

concrete thermo-active pile for centrifuge testing
A. Minto, A.K. Leung, D. Vitali & J.A. Knappett

University of Dundee, Dundee, Scotland, UK
ABSTRACT: The use of thermo-active geo-structures has been recognised to be a sustainable engineering
solution that can reduce carbon emissions from civil infrastructure. Physical modelling in a geotechnical cen-
trifuge has been increasingly used to study the behaviour of this kind of geo-structures and their interaction
with the surrounding soil under cyclic heating/cooling loads. Previous studies have been limited by the choice
of materials used to model thermo-active geo-structures (such as aluminium and conventional concrete), due
to inaccurate scaling of the thermal properties and the inability to capture the quasi-brittleness and strength
properties of reinforced concrete (RC) due to improper scaling of aggregate sizes. The paper aims to develop a
new thermally-enhanced plaster-based model concrete which can realistically reproduce both the thermal and
mechanical properties that are representative of concrete at prototype scale. The new model concrete, combined
with steel wire reinforcement (i.e. geometrically scaled reinforcing bars and stirrups), was then used to create
1:20 scaled RC thermo-active piles. Effects of temperature on their thermomechanical behaviour, including the
coefficient of thermal expansion, moment capacity and flexural stiffness, were investigated. The suitability of
using the newly-developed RC thermo-active piles for future centrifuge testing is discussed.
1 INTRODUCTION internal diameter, and hence the second moment of

area, of the model pile. In terms of thermal proper-
Thermo-active geo-structures such as pile foundations ties, aluminium has a coefficient of thermal expansion
and diaphragm walls have been used to exploit shal- (CTE) of 22.2 µε/◦ C (where µ is micron; and ε is
low geothermal energy to provide heating and cooling thermally-induced axial strain). In their aluminium
to buildings (Brandl, 2006). A number of field tests piles, an epoxy resin with a thickness of 1.5 mm had to
(Laloui et al., 2006, Bourne-Webb et al., 2009, Murphy be applied on the pile shaft to protect the strain gauges.
and McCartney, 2015) have been conducted to investi- As the CTE of epoxy is 85 µε/◦ C (Rotta Loria et al.,
gate the thermomechanical behaviour of these types of 2015), the CTE of the overall instrumented aluminium
structures and their interaction with the surrounding pile should be higher than 22.2 µε/◦ C, which is thus
soil. It was demonstrated that cyclic heating/cooling at least 30% higher than the CTE of unrestrained RC
processes introduced additional stresses and strains to (15–16 µε/◦ C; Goode III et al., 2015). This makes it
the thermo-active geo-structures. In order to improve difficult for this kind of aluminium pile to realistically
understanding of the soil-structure interaction in more capture the soil-structure interaction of thermo-active
detail, small-scale physical model tests in a geotech- piles that are made of RC in the prototype. Numerical
nical centrifuge have been conducted (Stewart and studies (Yavari et al., 2014) have shown that the coef-
McCartney, 2014, Ng et al., 2015). By elevating the ficient of thermal expansion of a thermo-active pile
acceleration of a physical model by N times Earth’s plays a significant role in affecting the mobilisation
gravity (i.e., N -g), the stress levels experienced by and distribution of pile axial loads under a combina-
the soil in a small-scale physical model are simi- tion of mechanical and thermal loadings. Moreover,
lar to that in a full scale prototype, enabling the this type of elastic aluminium pile is not able to mimic
stress dependency of the soil behaviour to be correctly the nonlinear quasi-brittleness feature of RC under ten-
captured. A major challenge of testing a thermo- sile and bending stresses such as when subject to lateral
active geo-structure in the centrifuge is the selection loading conditions (Karihaloo and Huang, 1991).
of appropriate model materials that model both the Stewart and McCartney (2014) modelled a thermo-
mechanical and thermal properties realistically and active pile using RC with scaled steel reinforce-
simultaneously. ment and scaled aggregates in their centrifuge tests.
In the centrifuge tests performed by Ng et al., Although the thermal properties of prototype RC
(2015), aluminium alloy was used to model a thermo- might be properly scaled using RC as a model pile,
active pile. Scaling of the mechanical properties such the use of gravel with a maximum diameter of 6 mm
as the axial or flexural rigidity of the aluminium (in model scale) to model coarse aggregates could pose
pile can be achieved through careful selection of the a scaling problem of the mechanical properties. At 24g
37
which Stewart and McCartney (2014) adopted in their
centrifuge tests, the prototype diameter of the gravel
was 144 mm, which was 7 times larger than the diam-
eter of coarse aggregates in prototype concrete. This
size effect would potentially lead to an over-strength
of RC (Litle and Paparoni, 1966, Belgin and Sener,
2008). This modelling method is thus limited to rela-
tively low scaling factors (or g-level) to minimise any
over-strength of RC.
In order to more realistically capture the nonlin-
ear quasi-brittleness feature and failure mechanisms
of concrete in centrifuge at higher scaling factors,
Knappett et al. (2011) developed model concretes
using plaster-based mortars. In the mortar mix, fine sil-
ica sand was used to geometrically scale the aggregate Figure 1. Particle-size distribution of silica sand and copper
found in concrete. Such model concrete was shown to powder.
have representative mechanical strengths, in terms of
unconfined compressive strength and modulus of rup- distribution of the silica sand and copper powder were
ture. The model concrete has been successfully used measured using a laser diffraction analyser. The results
for modelling various engineering structures such as are compared in Fig. 1. It can be seen that both the sand
piles (Al-Defae and Knappett, 2014) and bridge piers and copper powder were uniformly graded, and that the
(Loli et al., 2014) in the centrifuge where simultane- size of copper powder was finer than that of the sand.
ous modelling of stiffness and strength are crucial.
For modelling thermo-active RC geo-structures, this
type of model concrete requires further modification 2.2 Thermomechanical properties of the new
to ensure correct scaling of the thermomechanical model concrete
properties.
In order to investigate the effects of the copper pow-
This study aims to develop a new type of model
der content on the thermal conductivity of the model
concrete that can realistically scale the mechanical
concrete, a series of laboratory testing was carried out
and thermal properties of real concrete for future
using a hot-box apparatus developed by Jones et al.
centrifuge testing of concrete energy geo-structures
(2007). It is an apparatus that can create a temperature
for larger scaling factors. An application of the new
gradient across a slab-shaped specimen (45 mm width,
model concrete to produce RC thermo-active piles
150 mm long and 150 mm height) and also can mea-
is presented. Temperature effects on thermomechan-
sure the corresponding heat flux. At the steady state,
ical properties, including the coefficient of thermal
the thermal conductivity of the specimen can be deter-
expansion, moment capacity and flexural stiffness, of
mined by dividing the heat flux by the temperature
a model pile were tested. By comparing the model and
gradient, according to Fourier’s law.
prototype properties, the suitability of testing this type
Model concrete mixed with five different percent-
of new RC model pile in centrifuge is discussed.
ages of copper powder (by volume), 0%, 1.5%, 3%,
6% and 12% were tested. The test results depicted in
Fig. 2 show that the thermal conductivity of the model
2 NEW MODEL CONCRETE
concrete originally designed by Knappett et al. (2011)
was 0.4 W/(m·K), which was lower than the typical
2.1 Constituents of the model concrete
range of concrete (i.e., 0.9 to 1.1 W/(m·K); Kanbur
The new model concrete developed in this study et al. (2013)). When copper powder was added, there
is based on the design previously proposed by was almost a linear increase in the thermal conductiv-
Knappett et al. (2011). The original design consisted of ity with the amount of copper added. This shows that
a mixture of β-form surgical plaster (manufactured by the addition of copper powder was effective to enhance
Saint Gobain), water and fine silica sand (Congleton the thermal properties of the model concrete. In par-
HST95). The sand was used to geometrically scale and ticular, 6% and 12% copper powder contents appeared
approximate the size of aggregates found in concrete. to match the prototype range reasonably well.
Knappett et al. (2011) suggested that a water/plaster Although adding copper powder could significantly
(W/P) ratio of 0.9:1 and a sand/plaster (S/P) ratio of improve the thermal properties of the new model con-
1:1 would result in a model concrete that can realisti- crete, one concern is any detrimental effects of such
cally mimic the mechanical properties of concrete in addition on the mechanical properties. For this pur-
prototype. pose, a series of four-point bending (FPB) tests were
In order to properly scale and mimic the thermal conducted to measure the modulus of rupture (fr ) of
properties of prototype concrete, a new constituent, prismatic specimens (25 × 25 × 250 mm) when differ-
copper powder (manufactured by Phoenix Scientific), ent percentages of copper powder were added to the
was added to the design mix to enhance the thermal model concrete. The testing procedures outlined by
conductivity of the model concrete. The particle-size Knappett et al. (2011) were adopted.
38
due to the rapid evolution of the pore structure during
the hydration process (Song et al., 2009).
Fig. 4(c) shows the model concrete mix at × 300
where both the sand and copper particles are visible.
There was a gap between the irregular-shaped sand
particles and plaster, hence creating some weakened
interfaces. However, such gapping was not found for
the copper particles. It must be pointed out that no
chemical bonding was formed along the plaster, sand
and copper particles after the hydration process. These
constituents were bonded physically through weak
inter-particle van der Waal force. Such physical bond
is an important feature of the model concrete to real-
istically mimic the non-linear quasi-brittleness nature
Figure 2. Effects of copper powder content on thermal con- of concrete in prototype, which cannot be captured
ductivity of model concrete. Error bars represent standard by existing elastic model piles made of aluminium,
errors (n = 3). and which avoids the potential over-strength of using
cement as the binder.
3 NEW MODEL RC THERMO-ACTIVE PILE
3.1 Structural analysis and design

Before producing a RC thermo-active pile, a detailed
structural analysis and design was carried out. In
future centrifuge tests, model thermo-active piles with
a square cross-section will be used to stabilise a silty
slope in a 1:20 scale model tested at 20 g. Preliminary
analysis and a literature review of previous studies
using piles to reinforce soil slopes (Al-Defae and
Figure 3. Effects of copper powder content on fr of the new Knappett, 2014, Hayward et al., 2000) suggests that
model concrete. Error bars represent standard errors (n = 6). a moment capacity of 250 kNm would be sufficient
for the piles to maintain the slope stability for repre-
Fig. 3 shows the variations of copper powder con- sentative soil properties and slope angles. Based on the
tent with normalised fr (by the values obtained from targeted moment capacity, the pile reinforcement was
0% copper powder content, fr0 ). It is clear that the designed according to the design procedures outlined
addition of copper powder up to 12% did not cause in Eurocode 2: Design of Concrete Structures (BSI,
significant changes in the mechanical properties. The 2004). Hence, at model scale, the thermo-active pile
addition of copper powder could thus enhance the will have a square cross section with the side of 25 mm
thermal conductivity of the model concrete, while and a length of 250 mm. In the following discussion,
simultaneously maintaining its ability to mimic the all dimensions are expressed in model scale, unless
mechanical properties of concrete. stated otherwise.
The arrangement of the reinforcement is detailed in
Fig. 5. It was a doubly-reinforced concrete pile, since
2.3 Micro-structure of the new model concrete
to allow bending in either or both horizontal direc-
In order to have a better understanding of the inter- tions in future centrifuge testing. All the longitudinal
action between the plaster, sand grains and copper reinforcement bars chosen were 1.25 mm in diameter,
powder as well as the uniformity of these constituents while the diameter of shear reinforcement (aka stir-
within the mixture, the micro-structure of the new rups) was 0.6 mm. The thickness of concrete cover was
model concrete was imaged by a scanning electron 2 mm. The total area of steel reinforcements (As ) was
microscope (SEM). Fig. 4(a) shows an SEM image at 12.93 mm2 for a cross-sectional area of 625 mm2 of the
a magnification of × 55. The image depicts some sil- model pile. This means that the reinforcement ratio,
ica sand physically interlocked with the flake-shaped As /bd (where b and d is the breath and the depth of the
plastic matrix. The distribution of the sand appears model pile, respectively), is 2.1%, which falls within
uniform. a typical range found in high-strength doubly rein-
At a higher magnification of ×500 (Fig. 4(b)), forced concrete beams under flexural loads (Rashid
the fine copper particles can be viewed. They were and Mansur, 2005).
spherical in shape and physically embedded into the The model reinforcement used in this study was
fibre-like irregular structures of the plaster. This kind a stainless steel (Grade 316), which was manufac-
of micro-structure is a key feature of the β-form plaster tured by Ormiston Wire Ltd. It has a yield strength
39
Figure 5. Cross-section and reinforcement details of a 1:20
model RC thermo-active pile.
Figure 6. Overview of the formwork used to cast two model

piles.
reinforcement bars and stirrups were coated with

epoxy resin and then immersed in a pool of HST95
silica sand for 24 hours to ensure an intimate mechan-
ical frictional contact between the reinforcements and
the sand, preventing the reinforcements from slipping
inside the model concrete.
In order to allow water to be circulated inside
the new model concrete, two pairs of closed-loop
U-shaped silicon pipes were attached to the internal
surface of the stirrups (Fig. 5), which is a typi-
cal arrangement for a thermo-active pile prototype
(Loveridge and Powrie, 2013). Placing the pipes closer
to the pile perimeter aimed to minimise their thermal
interaction and the heat loss to the concrete material,
hence achieving a more uniform distribution of pile
temperature. Cecinato and Loveridge (2015) showed
that circulation pipes spaced too closely can reduce
the heat transfer efficiency to the surrounding soil.
Figure 4. SEM images of at (a) ×55; (b) ×500; and The diameter of the silicon pipe was 3 mm, which was
(c) ×300 magnification. the minimum internal size of the pipe required to min-
imise the thrust force caused by the change in water
of 460 MPa (Knappett et al. 2011) and a coefficient of flow direction at the U-turn of the pipe (see Fig. 6).
thermal expansion of 18.5 µε/◦ C (Lee et al., 2007). Three thermocouples were attached to both ends
and the centre of the reinforcement bar for monitoring
the temperature distribution along the pile.
3.2 Model preparation
When the reinforcements, silicon pipe and ther-
Formwork with a dimension identical to the model mocouples were fixed in position, the mix of model
pile (25 mm × 25 mm × 250 mm; Fig. 6) was con- concrete (including plaster, water, silica sand and cop-
structed to cast the model piles. Before casting, all per powder) was poured into the formwork. The mix
40
was allowed to cure for 28 days in a room maintained
at an ambient temperature of around 20◦ C.
In order to heat and cool the model pile, a heating
system (Julabo Ltd; Model F12) that can control and
maintain constant water temperature between 1 and
99◦ C was connected to the silicon pipes. Change in
pile temperature was monitored by the three thermo-
couples embedded in the model pile.
4 TESTING OF THE MODEL PILE
4.1 Test plan

Laboratory testing was undertaken to determine the Figure 7. Typical setup of a FPB test for model pile.
thermo-mechanical properties of the small-scale RC
thermo-active pile made by the new model concrete. The CTE tests were started by submerging each
The primary aim was to ensure the design of the model model pile in a water bath with a controlled temper-
pile was mechanically and thermally representative ature of 50◦ C for at least two hours. When a uniform
of the prototype RC thermo-active pile. An impor- distribution of pile temperature of about 50◦ C was
tant thermal property that needed to be characterised reached at the steady state (as indicated by the three
was the coefficient of thermal expansion (CTE). Upon thermocouples), each model pile was removed from
pile temperature changes, the pile expansion (upon the water bath and any change in pile length between
heating) and contraction (upon cooling) would have the two reference points was measured immediately
direct effect on the soil-pile contact, hence affecting using a calliper with an accuracy of 0.01 mm. The
the soil-pile interaction and the mobilisation of pile measurements were taken as quickly as possible to
axial load (Yavari et al., 2014). minimise heat loss from each model pile.
Another test series aimed to determine any tem- For the measurements of the pile moment capac-
perature dependency on the flexural properties, which ity, the four-point bending (FPB) testing method was
is a key thermo-mechanical property that governs the adopted (Balendran et al., 2002). Two pile tempera-
pile behaviour when the piles are subjected to lateral tures were considered: circulating water at ambient
loading in slope stabilization applications. temperature (20◦ C) and an elevated level of 50◦ C.
In both types of tests, effects of the percentage At each pile temperature, model piles cast with three
of copper powder addition were studied. This aimed different amount of copper powder (0%, 6% and 12%
to determine any optimum amount of copper powder by volume) were tested. 17 FPB tests were carried out
that could realistically scale both the mechanical and (7 tests at 0%, 4 tests at 6% and 6 tests at 12%).
thermal properties of the model pile simultaneously. A typical set up for a FBD test is shown in Fig. 7.
The model pile was roller-supported, having a span
4.2 Test procedures (L) of 210 mm. Vertical loading was applied symmet-
rically at two locations, both of which were 35 mm
The testing method of CTE followed a modification of away from the centre of the pile to ensure that the
the procedures outlined in AASHTO TP 60. CTE was loads and reactions were spaced at L/3 as per typical
calculated by the change in the length of a model pile FPB testing procedures (Huurman and Pronk, 2009).
when there is an increase in pile temperature. After By using the heating system, water was controlled at a
casting the model piles, they were coated with a sil- constant temperature of 20 or 50◦ C and was circulated
icon conformal coating spray to ensure a waterproof to the model pile. When all thermocouples showed
condition, as would be required in future centrifuge a temperature matching the targeted value set on the
tests. Then, two reference points were defined in each heating system, test began by loading the model pile at
model pile to determine their initial pile length. The a rate of 0.6 mm/min. The test was complete when the
two reference points were fixed along the axis of the pile section exhibited large deformation and formed
two different longitudinal sides, which are orthogo- prominent cracking.
nal with respect to each other. From Fig. 5, ‘side 1’
refers to the four longitudinal reinforcement bars laid
in parallel, while ‘side 2’ refers to the adjacent side,
5 RESULTS AND DISCUSSION
where no longitudinal reinforcement bars were added.
By comparing the test results obtained from sides 1
5.1 Coefficient of thermal expansion
and 2, any effects of heat-induced elongation of the
reinforcement bars on the overall length of the model Table 1 summarises the measured values of CTE for
pile may be quantified. In total, three tests were con- the Piles A to C. It is found that the CTE of the
ducted, one with no copper powder added (Pile A), and model pile without copper powder (i.e., Pile A) was
two replications with 6% copper powder content (Piles 15.7 µε/◦ C, which was close to the values (<3% dif-
B and C). ference) obtained from side 2 of Piles B and C when
41
Table 1. Summary of the CTE test.
Pile Side Copper powder (%) CTE (µε /◦ C)
A 1 0 15.7
B 1 6 16.5
2 15.3
C 1 6 16.0
2 15.5
6% of copper powder was added. This is because the

CTE of copper (i.e., 17.5 µε/◦ C; (Davis, 2001)) was
not very significantly different than that of the plaster
(13.9 µε/◦ C; (Hyer, 2009)) and quartzite (i.e. sand par-
ticles), which ranged from 7.0 to 13.5 µε/◦ C. (Johnson
and Parsons, 1944). This suggests that the addition of
copper powder did not introduce significant effects on
the CTE.
It can also be seen in Table 1 that the CTE obtained
from side 1 (where reinforcement bars were present)
was slightly higher than that on side 2, by not more than
8%. This was somewhat expected because the CTE
of plaster (i.e., 13.9 µε/◦ C) did not differ much from
that of the steel reinforcement used (i.e., 18.5 µε/◦ C).
By taking the average measured values obtained from
sides 1 and 2, the CTE of Pile B (15.9 µε/◦ C) was
close to that of Pile C (15.8 µε/◦ C). These average
Figure 8. Effect of copper powder on (a) moment capacity,
values match well with the typical range of CTE of (b) flexural stiffness of model piles at 20◦ C and 50◦ C. Error
unreinforced concrete (i.e., 9.4 to 14.4 µε/◦ C; depend- bars represent standard error.
ing on the aggregate mineralogy; Choi and Chen,
2005). Goode III et al., (2015) reported that the added (Fig. 3(b)). This implies that the presence of the
measured CTE of their RC pile (which was made silicon pipes and the water being circulated in model
of conventional concrete) under unrestrained con- pile played a significant role on the thermo-mechanical
dition (i.e., free to expand upon pile heating) was responses. The CTE of silicon is ∼2.3 µε/◦ C (Vasiliev,
15–16 µε/◦ C, which was also very close to the aver- 2008), while that of water ranges from 200 µε/◦ C at
age value of the model RC thermo-active pile created 20◦ C to 450 µε/◦ C at 50◦ C (Kell, 1975). This sug-
in this study. gests that the thermal expansion of the water has a
greater influence than the expansion of the silicon
pipes. When the copper powder content is higher, the
5.2 Temperature effects on moment capacity and
thermal expansion of the copper particles, silicon pipe
flexural stiffness
and the circulating water might have created more
In this study, pile failure was defined at the onset of weakened interfaces, hence reducing the Mult of the
the first observable flexural cracking on the surface model pile.
of the model pile. The rationale of defining this fail- Flexural stiffness (EI, where E and I are theYoung’s
ure criterion is that as flexural cracking develops, the modulus and the moment of inertia of the un-cracked
model pile would no longer be serviceable as an energy section of a model RC, respectively) can be obtained
geo-structure. This is because the embedded pipes, by determining the gradient of the linear portion of a
which have limited flexural resistance, would be split, moment-curvature plot. The linear portion is defined
causing leakage of the circulating water. as the range of Mult from zero to a value where the
Fig. 8(a) shows the effects of copper powder first crack was observed in the model piles. Effects of
percentage on normalised ultimate bending moment copper powder and pile temperature on normalised EI
(Mult ) at different pile temperatures. At both tempera- are shown in Fig. 8(b). At pile temperature of 20◦ C, the
tures, there were not very substantial changes in Mult copper powder percentage has negligible effects on the
as the amount of copper powder increases from 0 to EI. In contrast, at an elevated level of 50◦ C, an increase
6%, given the presence of sample variability. When the in copper powder percentage causes a significant drop
percentage of copper powder increased further to 12%, of EI.
significant drops of Mult were found. The drop under Based on the test results presented in this paper,
the elevated pile temperature at 50◦ C was much more it appears that 6% was an optimum percentage of
substantial (up to 86%). Interestingly, such a drop was copper powder that could provide a model RC thermo-
not observed in plain model concrete, where fr was active pile with a reasonably close CTE (Table 1)
almost unaffected by the percentage of copper powder and thermal conductivity (Fig. 2) when compared to
42
the properties of a prototype thermo-active pile. At Al-Defae, A. H. & Knappett, J. A. 2014. Centrifuge Mod-
6% copper powder content, the effects of temper- eling of the Seismic Performance of Pile-Reinforced
ature on the mechanical properties, including both Slopes. Journal of Geotechnical and Geoenvironmental
Mult and EI may be practically negligible (see Fig. Engineering, 140.
Balendran, R., Zhou, F., Nadeem, A. & Leung, A. 2002.
8). This is consistent with the BS EN 1992-1-2:2004 Influence of steel fibres on strength and ductility of nor-
Clauses 3.2.2 and 3.2.3, which respectively state that mal and lightweight high strength concrete. Building and
any temperature effects on the mechanical properties environment, 37, 1361–1367.
of concrete and steel are negligible when the tempera- Belgin, C. M. & Sener, S. 2008. Size effect on failure
ture is below 100◦ C.A study reported by Li and Purkiss of overreinforced concrete beams. Engineering Fracture
(2005) also shows that theYoung’s modulus of concrete Mechanics, 75, 2308–2319.
is reasonably constant for temperatures between 20 Bourne-Webb, P. J.,Amatya, B., Soga, K.,Amis, T., Davidson,
and 60◦ C. C. & Payne, P. 2009. Energy pile test at Lambeth College,
London: geotechnical and thermodynamic aspects of pile
response to heat cycles. Geotechnique, 59, 237–248.
6 CONCLUSION Brandling, H. 2006. Energy foundations and other thermo-
active ground structures. Geotechnique, 56, 81–122.
A new type of model concrete that is highly suitable British Standards Institution. 2004. Eurocode 2: Design of
Concrete Structures. London, BSI.
for realistically modelling thermo-active energy geo- BS EN 1992-1-2:2004: Eurocode 2: Design of concrete struc-
structures in centrifuge testing is developed. The new tures. Part 1.2: General rules – Structural fire design.
model concrete is a mixture of plaster, silica sand, British Standards Institution, London, 2004.
water and copper powder, which results in a material Cecinato, F. & Loveridge, F. A. 2015. Influences on the ther-
that has both mechanical and thermal properties scaled mal efficiency of energy piles. Energy, 82, 1021–1033.
simultaneously. The test results showed that adding 6% Choi, J. H., & Chen, R. H. L. 2005. Design of continu-
or 12% copper powder content to the new model con- ously reinforced concrete pavements using glass fiber
crete could increase the thermal conductivity to a value reinforced polymer rebars. Publication No. FHWA-HRT-
close to those found in prototype concrete. Such an 05-081. Washington, D.C.
Davis, J. R. 2001. Copper and copper alloys, ASM interna-
addition was shown not to cause significant changes tional.
in modulus of rupture. Goode III, J. & McCartney, J. 2015. Centrifuge modeling
This study also demonstrated an application of this of end-restraint effects in energy foundations. Journal of
type of new model concrete to create 1:20 model rein- Geotechnical and Geoenvironmental Engineering, ASCE,
forced concrete (RC) thermo-active piles. The effects DOI: 10.1061/(ASCE)GT.1943-5606.0001333.
of copper powder content (ranging from 0% to 12%) Hayward, T., Lees, A., Powrie, W., Richards, D. & Smethurst,
on the thermomechanical behaviour of the model RC J. 2000. Centrifuge modelling of a cutting slope stabilised
pile were quantified. It was found that copper powder by discrete piles, Transport Research Laboratory.
content of 6% is an optimum amount that would pro- Huurman, M. & Pronk, A. 2009. Theoretical analysis of the 4
point bending test.AdvancedTesting and Characterization
vide the model pile with a highly representative value of Bituminous Materials, A. Loizos, MN Partl, T. Scarpas,
of the coefficient of thermal expansion as compared and IL Al-Qadi, eds., CRC Press, Boca Raton, 749–759.
to prototype. At this percentage of addition, effects of Hyer, M. W. 2009. Stress analysis of fiber-reinforced
pile temperature between 20 and 50◦ C on pile bending composite materials, DEStech Publications, Inc.
moment capacity and flexural stiffness are practically Johnson, W. H. & Parsons, W. H. 1944. Thermal expansion
negligible. This correctly models the thermomechani- of concrete aggregate materials, US Government Printing
cal behaviour of prototype RC piles. Such a model RC Office.
thermo-active pile was also capable of mimicking the Jones, M. R., Zheng, L., McCarthy, A., Dhir, R. K., &
nonlinear quasi-brittle nature of real concrete, which is Yerramala A. 2007. Increasing the use of foamed concrete
incorporating recycled and secondary aggregates. WRAP
a key feature that is not achievable using elastic model Project Report:AGG79-001.
piles. Kanbur, B. B., Atayilmaz, S. O., Demir, H., Koca, A. &
Gemici, Z. 2013. Investigating the thermal conductivity
of different concrete and reinforced concrete models with
ACKNOWLEDGEMENTS
numerical and experimental methods. Recent Advances in
Mechanical Engineering Applications, Recent Advances
The authors would like to acknowledge the stu- in Mechanical Engineering Series, 95–101.
dentships and the research cost supported by the Karihaloo, B. L. & Huang, X. 1991. Tensile response of
Energy Technology Partnership (ETP), Scottish Road quasi-brittle materials. Pure andApplied Geophysics, 137,
Research Board (SRRB) from Transport Scotland and 461–487.
the EPSRC Doctoral Training Award. Kell, G. S. 1975. Density, thermal expansivity, and com-
pressibility of liquid water from 0. deg. to 150. deg..
Correlations and tables for atmospheric pressure and satu-
REFERENCES ration reviewed and expressed on 1968 temperature scale.
Journal of Chemical and Engineering Data, 20, 97–105.
American Association of State Highway and Transporta- Laloui, L., Nuth, M. & Vulliet, L. 2006. Experimental
tion Officials 2007. Provisional Test Method for the and numerical investigations of the behaviour of a heat
Coefficient of Thermal Expansion of Hydraulic Cement exchanger pile. International Journal for Numerical and
Concrete. Washington DC, AASHTO. Analytical Methods in Geomechanics, 30, 763–781.
43
Lee, H.-Y., Lee, S.-H., Kim, J.-B. & Lee, J.-H. 2007. Creep– pile performance in saturated sand. Canadian Geotechni-
fatigue damage for a structure with dissimilar metal welds cal Journal, 52, 1045–1057.
of modified 9Cr–1Mo steel and 316L stainless steel. Rashid, M. & Mansur, M. 2005. Reinforced high-strength
International Journal of Fatigue, 29, 1868–1879. concrete beams in flexure. ACI Structural Journal, 102,
Li, L.Y. & Purkiss, J. 2005. Stress–strain constitutive equa- 462–471.
tions of concrete material at elevated temperatures. Fire Rotta Loria, A. F., Gunawan, A., Shi, C., Laloui, L., &
Safety Journal, 40(7), 669–686. Ng, C. W. W. 2015. Numerical modelling of energy piles
Litle, A. W. & Paparoni, M. 1966. Size Effect in Small- in saturated sand subjected to thermo-mechanical loads.
Scale Models of Reinforced Concrete Beams. Journal Geomechanics for Energy and the Environment, 1, 1–15.
Proceedings, 63. Song, K.-M., Mitchell, J. & Gladden, L. F. 2009. Observ-
Loli, M., Knappett, J. A., Brown, M. J., Anastasopou- ing microstructural evolution during plaster hydration.
los, I. & Gazetas, G. 2014. Centrifuge modeling of Diffusion Fundamentals, 10, 22.1–22.3.
rocking-isolated inelastic RC bridge piers. Earthquake Stewart, M. A. & McCartney, J. S. 2014. Centrifuge Modeling
Engineering & Structural Dynamics, 43, 2341–2359. of Soil-Structure Interaction in Energy Foundations. Jour-
Loveridge, F.A. & Powrie, W., 2013, April. Pile heat exchang- nal of Geotechnical and Geoenvironmental Engineering,
ers: thermal behaviour and interactions. Proceedings of ASCE, 140.
the Institute of Civil Engineers: Geotechnical Engineering Vasiliev, L. 2008. Micro and miniature heat pipes–Electronic
(Vol. 166, No. 2, pp. 178–196). component coolers. Applied Thermal Engineering, 28,
Murphy, K. D. & McCartney, J. S. 2015. Seasonal Response of 266–273.
Energy Foundations During Building Operation. Geotech- Yavari, N., Tang, A. M., Pereira, J.-M. & Hassen, G. 2014.
nical and Geological Engineering, 33, 343–356. A simple method for numerical modelling of energy
Ng, C. W. W., Shi C., Gunawan, A., Laloui, L. & Liu, H. L. pile’s mechanical behaviour. Geotechnique Letters, 4,
2015. Centrifuge modelling of heating effects on energy 119–124.
44
Monitoring elastic parameter changes in the vicinity of salt caverns due to

cyclic loading by seismic waveform inversion
D. Köhn, D. De Nil, S.A. al Hagrey & W. Rabbel

Institute of Geosciences, Department of Geophysics, Christian-Albrechts-University Kiel, Germany
K. Khaledi, D. König & T. Schanz

Faculty of Civil and Environmental Engineering, Chair for Foundation Engineering, Rock and Soil Mechanics,
Ruhr-University Bochum, Germany
ABSTRACT: Increased emission of greenhouse gases into the atmosphere lead to a strong requirement of
renewable energy resources. However, they are intermittent and need buffer storage to bridge the time-gap
between production and public demands. The storage of compressed gas energy in sealed underground structures
like salt caverns is one approach to bridge this time gap. The cyclic loading of compressed air in the cavern,
with periods ranging from hours to days, leads to a response of the host rock in the vicinity of the cavern. To
predict the mechanical behavior of rock salt an elastic-viscoplastic-creep model with damage is implemented in
the finite-element modeling software CODE-BRIGHT. The resulting changes of the elastic material parameters
after 400 loading cycles under extreme loading conditions are converted into seismic P- and S-wave velocity
variations. Changes occur in a thin layer with a maximum thickness of five meters near the ceiling and floor of
the cavern. Within this area P-wave velocities are reduced by 50 m/s to 160 m/s and S-wave velocities by 20 m/s
to 80 m/s, respectively. The local character of the material parameter changes prohibits a resolution by classical
traveltime based seismic tomographic approaches. Therefore, we develop an elastic Full Waveform Inversion
(FWI) strategy for two different acquisition geometries. Acquisition setup 1 consists of a vertical source and
receiver line inside the cavern, while in acquisition geometry 2 the receivers are moved into boreholes outside of
the cavern, resembling a classical cross-hole tomography. To mitigate the non-linearity of the inversion problem
due to strong multiple reflections of elastic waves radiated by the cavern, a time-damping approach is applied
to the recorded seismic data. Both acquisition geometries allow an accurate and highly resolved reconstruction
of the elastic material parameter changes in the vicinity of the cavern due to the cyclic loading process.
1 INTRODUCTION strong cycle loading, changes of the elastic material

parameters in the cavern wall are very small and not
Mitigation of anthropogenic Greenhouse gases, detectable by ”classical” traveltime based tomographic
including CO2 emissions in the terrestrial atmo- approaches. To solve this problem, we investigate the
sphere demands developments of viable alternative of application of a high-resolution elastic time-domain
renewable energy resources including hydroelectric, Full Waveform Inversion (FWI). In a first step the
biomass, solar, wind, marine (wave/tides) and geother- mechanical behavior of the rock salt is predicted by
mal sources. Most of these sources produce energy an elastic-viscoplastic-creep model including damage
only when suitable weather conditions are prevail- implemented in the finite-element (FE) modeling soft-
ing and not when energy is directly demanded. These ware CODE-BRIGHT. The FE modelling covers the
sources are intermittent and need buffer storage to whole cavern construction process, first filling and
bridge the time-gap between off-peak production and strong cyclic loading phase. Modelled changes of elas-
demand peaks. One storage option already commonly tic material parameters in the vicinity of the idealized
used for natural gas are salt caverns. Most cavern numerical cavern after the cyclic loading phase are
monitoring techniques are focused on the accessible used for the subsequent FWI of the seismic wavefield.
cavern space and the injection borehole. The condition The resolution of two different seismic acquisition
and shape of the cavern are mapped by sonar meth- geometries are compared. While the FWI of acqui-
ods in combination with mine surveying techniques at sition geometry 1 is mainly based on the reflected and
the earth surface to determine any ground movement refracted wavefield, acquisition geometry 2 adds trans-
(Reitze et al. 2014), while the sealing of the injection mitted waves. Advantages, disadvantages, resolution
borehole is verified by pressure monitoring. However, and areal coverage of the different setups are discussed
all these methods are not able to monitor changes for the synthetic cavern model example, as well as a
in the host rock surrounding the cavern. Even under checkerboard resolution test.
45
Figure 2. The loading pattern applied to the inner boundary
of the cavern.
step is assumed to be 160 days. That means the whole

cavern is excavated after 4 years.
Debrining phase: the debrining phase has been
modeled in 9 steps from top to bottom of the cav-
ern. In each step, the brine pressure is replaced by the
gas pressure which is equal to the weight of brine col-
umn from ground surface to the current brine level.
Figure 1. Geometry and boundary conditions of the ideal- Thus, when the pressure of injected gas is equal to the
ized cavern model. brine pressure at the bottom of cavern, the whole brine
is ejected. In this numerical example, each debrining
2 NUMERICAL CAVERN MODEL UNDER step takes 20 days and the whole process is carried out
CYCLIC LOADING within 6 months.
First filling phase: in this phase, the pressure inside
A typical salt cavern with a simplified geometry has the cavern reduces to the minimum pressure of the
been modeled in GID software. GID is used as the cavern. It is assumed that the pressure reduction is
pre-processor and post-processor for the finite element carried out within 5 days.
solver CODE-BRIGHT. For a detailed description of Cyclic loading phase: to model the cyclic operating
the underlying mechanical model we refer to Khaledi condition, the internal pressure of the cavern fluctuates
et al. (2016). The axisymmetrical model with a height within a predefined range. Furthermore, it is assumed
of 800 m and a width of 500 m is shown in Fig. 1. The that the pressure variations during the charge and dis-
cavern itself has a diameter of 37.5 m and a height of charge of the cavern are linear. In this paper, an extreme
233 m and its top and bottom have a spherical shape. loading scenario is defined for the cyclic loading
On the upper model boundary a load of 10 MPa is operation. That means, the stresses around the cavern
applied which represents the overburden load at the top during the cyclic loading phase may locate inside the
of the model. The vertical displacement at the model dilatancy zone. Thus, the unfavorable consequences
bottom is restrained. The density of compact rock salt such as the damage development and microcracking
is assumed to be ρsalt = 2000 kg/m3 and the numerical can be experienced. In this scenario, the pressure varies
simulation is done at constant temperature T = 318 K. between 3 and 8 MPa. The duration of each cycle
The initial stress field is assumed to be isotropic (i.e. is assumed to be 1 day and 400 loading cycles are
σxx = σyy = σzz ). simulated. Fig. 2 describes schematically the loading
The following simulation phases have been consid- pattern which has been defined to simulate the cavern
ered to model the construction process as well as the excavation process as well as its cyclic loading oper-
cyclic loading operation. ation. Fig. 3 shows the final Lamé parameter distri-
Initial phase: At time t = 0 it is assumed that no bution with maximum reductions of 1.1 GPa for λ and
excavation has been performed. For this reason, a uni- 0.7 GPa for µ with respect to the undisturbed host rock.
formly increasing load equal to the geostatic pressure Changes occur at the curved ceiling and floor of the
is applied to the inner boundary of the cavern. Since cavern in a layer with a thickness of approximately 5 m.
the initial stress state is isotropic, the principal stresses
around the cavern are identical before the excavation.
Leaching phase: in order to model the leaching pro- 3 THEORY OF SEISMIC MODELLING AND
cess, the applied load inside the cavern which is equal FULL WAVEFORM INVERSION
to the geostatic pressure is gradually reduced to the
brine pressure (with ρbrine = 1100 kg/m3 ). This pro- Before applying the elastic FWI to the CODE-
cess has been performed in 9 excavation steps from BRIGHT cavern modelling results, we shortly
bottom to the top of the cavern. The duration of each summarize the fundamental physics of the seismic
46
Figure 3. True distribution of the Lamé parameters λ (a) and µ (b) in the cavern host rock after modelling the construction
process and strong cyclic loading operation.
forward problem and introduce the theory of full parallelized by domain decomposition using the Mes-
waveform inversion. sage Passing Interface (MPI). For a detailed descrip-
tion of the forward code, we refer to Bohlen (2002)
3.1 The seismic forward problem and Köhn et al. (2014).
The behavior of seismic waves in a 2D isotropic linear- 3.2 2D elastic full waveform inversion
elastic medium for the PSV case can be described by
the following equations of motion The classical FWI approach relies on the minimization
of the data residuals delta δv = vmod − vobs between
modelled seismic data vmod and field data vobs to
deduce high resolution models of elastic material
parameters in the subsurface. To solve this nonlin-
ear optimization problem an appropriate objective
function E has to be defined.
The objective function Eq. (2) can be minimized by

iteratively velocity Vs) at iteration step n, starting
with an initial background model m0 using the quasi-
Newton limited memory Broyden-Fletcher-Goldfarb-
Shanno (l-BFGS) method (Nocedal & Wright 2006)
where λ, µ denote the Lamé parameters, (vx , vz )

particle velocity vector, σxx , σzz , σxz stress tensor
components and (fx , fz ) directed body force vector,
respectively. For the numerical solution of eqs. (1) we
use a time-domain 2D finite-difference (FD) scheme where the product of the inverse Hessian H −1 with
with 2nd order operators in time and 4th order in space the gradient ∂E/∂m is iteratively approximated by
on staggered grids (Virieux 1986, Levander 1988). At finite-differences. The time-domain gradients can be
all boundaries convolutional PML (C-PML) absorb- effectively calculated using the adjoint state method
ing boundary conditions are implemented (Komatitsch (Tarantola 2005, Köhn et al. 2012). The step length
& Martin 2007) to avoid artificial boundary reflec- µn is estimated by an inexact parabolic line-search
tions.To reduce computation time, the resulting code is (Nocedal & Wright 2006).
47
4 MONITORING OF ELASTIC PARAMETER
CHANGES IN THE CAVERN MODEL
4.1 Model parametrization and synthetic seismic

data acquisition
Because the FWI approach is based on a model
parametrization in terms of P-wave and S-wave veloc-
ities, the modelled Lamé parameter distribution λ, µ
in the vicinity of the cavern after the modelled con-
struction process as well as the strong cyclic loading
operation from section 2 is converted to seismic
velocities via:
Figure 4. Seismic acquisition geometries used for the

FWI monitoring strategy: pure reflection seismic (a) and
cross-hole tomography geometry (b). White squares and
green dots denote the positions of source and receiver lines,
The changes of seismic velocities with respect to the respectively.
initial state of the host rock Vp0 = 4100 m/s, Vs0 =
2193 m/s
are shown in Fig. 6(a) and (b), respectively. Near

the curved ceiling and floor of the cavern, P-wave
velocities are reduced by 50 m/s to 160 m/s and S-
wave velocities by 20 m/s to 80 m/s, respectively. To
test the applicability of elastic FWI to resolve these
small material parameter changes, synthetic datasets
for two different acquisition geometries are calculated
by solving the forward problem (1). Acquisition setup
1 consists of a vertical source and receiver line inside
the cavern (Fig. 4a). 113 airgun sources radiate a pres-
sure wavefield, which covers a frequency range from
10 Hz to 100 Hz. The seismic waves are recorded by
226 hydrophones. This acquisition configuration has
the advantage, that no monitoring boreholes outside
of the cavern are required, sources and receivers can
be setup via the injection tube, similar to the equip-
ment used for sonar surveys. On the downside, this Figure 5. Seismic section for shot 1 (acquisition geome-
geometry can only record the reflected, refracted and try 1).
diffracted wavefield. To investigate to what extent this
limitation has an effect on the FWI resolution, we also 4.2 Application of elastic FWI
test a second acquisition geometry (Fig. 4b), where
560 multi-component receivers are placed into two To mitigate the non-linearity of the inverse problem,
boreholes outside of the cavern, while the shot posi- introduced by the complexity of the recorded wave-
tions inside the cavern are unchanged. This resembles field, we apply a sequential time-damping approach
a more classical cross-hole tomography approach with for the FWI (Brossier et al. 2009). Each trace of the
a much higher ray-coverage than acquisition geometry seismic section is damped from the first time sam-
1. As an example Fig. 5 shows the recorded pressure ple by the exponential function exp(−γt), with γ =
wavefield excited by shot 1 for acquisition geometry 1. 160, 80, 40, 20, 0. Starting model for the inversion is
Note the dominant direct wave and its reflection from the initial state of the host rock Vp0 , Vs0 including the
the cavern floor, the weak but still distinct refracted cavern. We assume that the geometry of the cavern was
wave and numerous reflections and their correspond- accurately estimated by sonar techniques and therefore
ing multiples. Due to the strong material parameter do not allow any model changes within the cavern. The
contrast at the interface between air-filled cavern and FWI result using acquisition geometry 1 are shown in
host rock, it is also likely that significant P to S and S Fig. 6 (c+d). Despite the limitation to reflected and
to P conversions occur. refracted seismic waves, the FWI could successfully
48
Figure 6. Cavern model: Comparison of the true changes of P-wave velocity δVp (a) and S-wave velocity δVs (b) with the
corresponding FWI results using acquisition geometry 1 (c+d) and acquisition geometry 2 (e+f).
reconstruct distribution and magnitude of the seismic ±80 m/s. Corresponding to the approach in section
velocity variations due to the cyclic loading of the 4.1, synthetic data is calculated for the true seismic
cavern. Introducing transmitted seismic waves with models and inverted by FWI as in section 4.2. The
acquisition geometry 2 does not significantly improve results for acquisition geometry 1 (Fig. 7 c+d) show
the FWI result (Fig. 6 e+f). The S-wave velocity vari- that the P-wave velocity checkerboard can be resolved
ations show a little bit less artifacts below the cavern in a layer with a thickness of roughly 2 m along the
floor. circumference of the cavern. In case of the S-wave
velocity model, this layer extents to a thickness of
about 7 m. Acquisition geometry 2 (Fig. 7 e+f) is not
4.3 Resolution analysis able to add a significant improvement in terms of reso-
For a quantitative estimation of the FWI resolution lution for the P-wave velocity model in the vicinity of
and areal coverage of the two different acquisition the cavern compared to acquisition geometry 1. This
geometries we apply a checkerboard test. Except for can be explained by the long wavelength of the P-wave
the air-filled cavern the homogeneous seismic back- within the host rock of about 40 m. Nevertheless, near
ground velocity models Vp0 , Vs0 are covered with the monitoring boreholes some tiles can be resolved
5 m × 5 m large tiles (Fig. 7a+b). For comparison the by FWI, maybe due to possible trade-offs between the
minimum wavelength of the P-wave in the cavern Vp and Vs model. Tiles of the Vs model are visible in
equals 3 m, in the host rock 40 m, while the minimum a layer with a thickness of 12 m around the cavern and
S-wavelength is 20 m. In each tile the P-wave veloc- also in a maximum radius of 5 m around the receiver
ity varies between ±150 m/s, the S-wave velocity by boreholes. Midway between boreholes and cavern
49
Figure 7. Checkerboard test: Comparison of the true changes of P-wave velocity δVp (a) and S-wave velocity δVs (b) with
the corresponding FWI results using acquisition geometry 1 (c+d) and acquisition geometry 2 (e+f).
surface, the checkerboard resolution is strongly resolution of a pure reflection based acquisition setup
reduced. In summary, both acquisition setups are able within the cavern is comparable with an acquisition
to monitor the thin layer in the cavern vicinity affected geometry based on transmitted waves, requiring addi-
by the strong cyclic loading operation, especially tional monitoring boreholes near the cavern to host the
S-wave velocity changes. receiver lines. The cylinder-symmetric cavern geom-
etry and material parameter distribution is a strong
simplification compared to a 3D asymmetric cavern.
Nevertheless, we assume that a similar seismic FWI
5 CONCLUSIONS monitoring concept could be applied to more complex
cavern geometries, if acquisition geometries based
We presented a feasibility study for a cavern mon- on seismic arrays and beam-forming techniques are
itoring concept based on seismic FWI, focusing on adapted to the problem.
elastic parameter changes within the host rock under
strong cyclic loading. For a realistic prediction of
the rock salt response, the cavern excavation and
strong cyclic loading operation is modelled using an ACKNOWLEDGMENTS
elastic-viscoplastic-creep model with damage, imple-
mented in the FE modeling software CODE-BRIGHT. This study has been carried out within the frame-
While the mechanical changes are very small and work of the ANGUS+ research project (Bauer et al.
localized in a thin layer around the cavern, the applied 2013) funded by the German Federal Ministry of Edu-
seismic FWI approach is able to resolve these vari- cation and Research (BMBF). The FWI inversions
ations. As demonstrated by a checkerboard test, the were performed on the NEC-HPC-Linux-Cluster at
50
Kiel University. The 2D FWI code DENISE Black- Köhn, D., D. De Nil, A. Kurzmann, A. Przebindowska, &
Edition is available under the terms of GNU GPL 2.0 T. Bohlen (2012). On the influence of model parametriza-
at https://github.com/daniel-koehn. tion in elastic full waveform tomography. Geophysical
Journal International 191(1), 325–345.
Köhn, D., A. Kurzmann, D. De Nil, & L. Groos (2014).
DENISE - User manual. available at http://www.geophysik.
REFERENCES uni-kiel.de/%7Edkoehn/software.htm.
Komatitsch, D. & R. Martin (2007). An unsplit convolutional
Bauer, S., C. Beyer, F. Dethlefsen, P. Dietrich, R. Duttmann, perfectly matched layer improved at grazing incidence for
M. Ebert, V. Feeser, U. Görke, R. Köber, O. Kolditz, the seismic wave equation. Geophysics 72(5), 155 – 167.
W. Rabbel, T. Schanz, D. Schäfer, H. Würdemann, & Levander, A. (1988). Fourth-order finite-difference P-SV
A. Dahmke (2013). Impacts of the use of the geological seismograms. Geophysics 53(11), 1425–1436.
subsurface for energy storage: an investigation concept. Nocedal, J. & S. Wright (2006). Numerical Optimization.
Environmental Earth Sciences 70(8), 3935–3943. Springer, New York.
Bohlen, T. (2002). Parallel 3-D viscoelastic finite-difference Reitze, A., F. Hasselkus, & P. Wurmbauer (2014). Today’s
seismic modelling. Computers & Geosciences 28(8), technology for monitoring caverns. BHM Berg- und
887–899. Hüttenmännische Monatshefte 159(4), 149–153.
Brossier, R., S. Operto, & J. Virieux (2009). Seismic imaging Tarantola, A. (2005). Inverse Problem Theory. SIAM.
of complex onshore structures by 2D elastic frequency- Virieux, J. (1986). P-SV wave propagation in heteroge-
domain full-waveform inversion. GEOPHYSICS 74(6), neous media: velocity-stress finite-difference method.
WCC105–WCC118. Geophysics 51(4), 889–901.
Khaledi, K., E. Mahmoudi, M. Datcheva, & T. Schanz (2016).
Stability and serviceability of underground energy storage
caverns in rock salt subjected to mechanical cyclic load-
ing. International Journal of Rock Mechanics and Mining
Sciences 86, 115 – 131.
51
Deep geothermal energy wells based on artificial large area fractures
J. Frankovská & M. Ondrášik

Faculty of Civil Engineering, Slovak University of Technology in Bratislava, Slovakia
Ch. Källberg
Dividend Industries, Sweden
ABSTRACT: Large diameter energy shafts is a method to utilize geothermal energy. The research presented
in this paper does not only cover the extraction of the energy from geological environments by heating of water,
but also storage of the energy in geological environments. Extraction of energy doesn’t have to be performed
through usual methods of deep hydrothermal wells, but energy can be stored and extracted from deep energy
shafts with a diameter of 6 m from which horizontal galleries is cut. The galleries is to be interconnected by
series of parallel narrow fractures cut by diamond wire at different depths. Water flowing through the spacing
of the fractures will be heated (extracting energy) or cooled (storing energy). Such a system will allow for a
controlled flow of water exchanging heat with the hosting rock through a large contact surface.
1 INTRODUCTION
Storage and extraction of energy in deep geologi-

cal environments is considered an important method
of energy management of the future. Many research
teams around the world are developing competing
technologies for storage and extraction of energy,
suitable for the characteristics of local geological envi-
ronments (Barbier 2002, Lund et al. 2005). Most con-
temporary technologies rely on uncontrolled natural
fractures (singenetic or postgenetic), artificial frac-
tures (made through blasting or hydraulic pressure), Figure 1. The schemes of basic types of deep energy wells.
or confined aquifer in deep geological environments a) Closed or open system based on natural or artificial set of
(fig. 1). The heat exchange is then provided by circula- fractures.
tion of groundwater in open or closed systems between b) System based on confined aquifer.
deep boreholes (injection and production wells) and c) System based on a large scale energy well with a set of
parallel mechanically cut fissures.
an exchange facility on the ground. Some systems can
only extract the heat from geothermal wells, provided
that the well has a sufficient discharge rate and low of Technology in Bratislava. The technologies needed
mineralization of the geothermal water (Vranovská for the development of large volume energy wells
et al. 2000). are tested on the first pilot testing field located in
The concept of geothermal energy storage and Stockholm archipelago (Sweden) on the island Stora
extraction in a deep geological environment is based Höggarn.
on narrow, open and controlled artificial fractures cut
from large diameter shafts and galleries (fig. 1). 2 MAIN FEATURES OF THE NEW CONCEPT
Its main features are:
As mentioned in the introduction, the concept is not
– shafts and galleries
based on geothermal boreholes (wells), but on deep
– large energy exchange contact surface
vertical shafts and horizontal galleries. The diameter
– controlled heat exchange fissure/slot
of the shaft (large volume energy well) is set to 6 m.
– nondestructive technology of shaft and gallery
The depth is adjustable to the local geological envi-
excavation
ronment and to the needs of the energy management.
– deposition and extraction of the energy
The vertical shaft is divided into sections due to con-
– commercially usable and valuable byproducts
struction and security reasons. Each section represents
The concept is recently being developed by Divi- one closed geothermal system independent from other
dend Industries in cooperation with Slovak University sections.
53
Figure 2. Distribution of shaft, galleries and the open fis-
sures of the heat exchanger (a, b). Detail of the diamond wire
cut fissure (c).
Horizontal galleries can be excavated at the end of

each vertical section. The galleries are used for cut-
ting open controlled fissures. Four galleries from two
sections are needed to cut one or more fissures (fig.
2a). The cutting is made with use of diamond wire
(fig. 2a). Several parallel fissures can be cut in each
section to provide sufficient surface for heat exchange
and rock volume for sufficient capacity of the heat
storage or extraction. The surface of one of the two Figure 3. The middle and lower platforms with the handling
walls of the fissure in a section can be up to 5 000 m2 , crane prior to transport to the place of testing.
and we can make up to 100 fissures in a section.
The resulting exchange surface is much greater than 3 TECHNOLOGY OF THE ENERGY WELL
the surface provided by common systems based on EXCAVATION
drilling.
The surface, volume, and dimensions of the fis- The proposed technology of the energy well excavation
sures cut by the nondestructive controlled method can by diamond wire cutting was selected for two reasons.
be regulated to meet the requirements of desired heat The first reason is the conscious effort not to dam-
exchange. These properties can be prepared to be suit- age the rock environment through blasting. Fractures
able for a specific geothermal storage capacity, heat from the blasting could create uncontrolled water seep-
exchange rate, fluid flow etc. age of the heated water from closed systems and also
To avoid destruction of the geological environ- complications when cutting fissures using the dia-
ment, as occurs when using blasting-techniques, the mond wire. The second reason is an effort to decrease
shaft and galleries are preferably also cut by diamond the construction cost. The commercial use of the
wire. Special machines have been developed for this removed rock blocks contributes to lowering the price
purpose, as described in the next chapter. of the construction of the whole geothermal storage
The concept of large volume energy wells can be facility.
expanded to both energy extraction from sources with Operations with heavy mining machines and exca-
dry geothermal energy, as well as for storage and sub- vated rock blocks are secured by a middle and lower
sequent extraction of energy where there is low or platform (fig. 3a). The platforms were developed and
nonexistent geothermal energy potential. designed specifically for the needs of the deep energy
The removed material from the shaft and galleries wells research project during 2012–2015. The middle
are rock blocks with dimensions 1.5 m × 1.5 m × 3 m platform, called “Runa” (fig. 3b), is located on rails
and a weight per piece of about 18 000 kg. Depending above the entrance to the shaft. It serves for the con-
on the geology of the site and related rock quality, trolling and lifting of heavy loads above and within
the removed rock blocks from the shaft and gal- the shaft. The lower platform, called “Kovlad” (fig.
leries can be presented on the commercial market 3c), fits into the higher part of the middle platform
among stone material for construction or decorative and can independently move up and down the verti-
purposes. According to preliminary calculations, in cal shaft like a spider. It serves as a working platform
ideal geological conditions only 2% of the removed for operating of heavy machinery and rock blocks and
material volume will be turned to waste. The other for drilling and cutting at the bottom of the shaft. All
98% will result in rock blocks usable for commercial the processes on the lower platform are intended to be
sale. automated and remote controlled.
54
the shaft and galleries from one section are prepared,
the galleries are interconnected with series of parallel
boreholes, through which diamond wire is inserted to
cut the fissures for the heat exchanger.
4 FIELD TESTING
Most of the described technologies are being tested on

the pilot site on the island Stora Höggarn. The island
is located at the tectonically stable geologic unit of
the Baltic Shield consisting of massive crystalline pre-
Cambrian rock (intrusive igneous and highly metamor-
phous rock). The geology of the island and engineering
geological Conditions of the underground facility
where the shaft entrance is located is described in the
article by Ondrášik et al. (2015).
Thanks to ideal geological conditions, the recent
activities have been focused on the engineering and
mining aspects of the concept. The middle and upper
platform are installed on the site and their functionality
is being tested. Cutting with diamond wire on pulley
inserted into the wide diameter boreholes is tested as
well, as are the most suitable drilling machines and
injection mixtures and their application.All supporting
installations and technologies are installed and built
on the island. After solving the engineering and min-
ing problems the concept will also be extended to less
ideal geological conditions (fractures, tectonic discon-
tinuities, rock layers) to find the geological limits of
the concept for storage and extraction of geothermal
energy.
5 CONCLUSIONS
Figure 4. Sequences of excavation of large scale energy well The presented concept for storage and extraction of
cut by diamond wire fixed on a system of pulleys: geothermal energy from large energy wells is not
a) drilling 300 mm holes for the pulleys, imagined as being used in singleton individual use
b) and c) cutting the rock with the diamond wire, scenarios, but our vision cover large urban munici-
d) lifting the extracted rock blocks with hydraulic jack and
splitting the blocks in half,
palities with seasonal excess of, and need for, energy,
e) lifting the rock block out from the shaft. which can be deposited and/or extracted. Solving
the challenges of establishing the deep energy well
in a geological environment is to be seen as a step
The operations on the lower platform include towards an integrated prospective green energy man-
drilling, cutting, and block manipulation. The agement ecosystem, which includes excessive energy
sequence of these processes is illustrated in figure 4. production, distribution, accumulation, and storage.
It starts with core drilling for geological probing, and Depending on the season, the direction of the energy
injection to seal any fractures and discontinuities. This flow is reversed from, or to, the storage and customers.
is followed by drilling 2 m deep boreholes with diam-
eter 300 mm (fig. 4a) along the perimeter of the shaft,
into which pulleys with the diamond wire are inserted REFERENCES
(fig. 4b, 4c) for cutting the rock blocks between the
boreholes. After cutting, the rock blocks are lifted Barbier, E., 2002. Geothermal energy technology and current
status: an overview. Renewable and sustainable energy
with hydraulic jacks (fig. 4d). For better control of reviews 6(1–2): 3–65.
the blocks in the narrow space of the shaft the blocks Lunda, J.W., Freestonb, D.H., & Boyd, T.L., 2005
are cut in half (fig. 4e). Direct application of geothermal energy: 2005 Worldwide
Finally, the blocks are lifted out from the shaft review. Geothermics 34(6): 691–727.
(fig. 4f). Similarly, but without the support of the lower Ondrášik, M., Kopecký, M., Frankovská, J., & Brček, M.,
platform, rock blocks are cut from the galleries. After 1915. Swedish granite – host rock environment of the deep
55
energy well. SGEM 2015. 15th International Multidisci- Vranovská, A., Beňovský, V., Drozd, V., Halas, O., & Vana,
plinary Scientific GeoConference. Science and Technolo- O., 2000. Investigation for geothermal energy utilisa-
gies in Geology, Exploration and Mining; Proc. intern. tion in the town Košice, Slovak Republic. Proceedings
conf., Albena, Bulgaria, 18–24. 6. 2015. Sofia: STEF 92 World geothermal congress 2000, Kyushu–Tohoku, Japan,
Technology, Volume 1: pp. 799–806. May 28–June 10, 2000. pp. 2283–2288.
56
Numerical study of bearing capacity and consolidation settlement

of energy piles in fine-grained soils
M. Khodaparast, M.A. Kiani & H. Bayesteh

Civil Engineering Department, University of Qom, Qom, Iran
ABSTRACT: Energy geostructures represent a renewable and clean source of energy that can be used for
heating and cooling of buildings. This innovation uses deep foundations as energy supply and support of super-
structure. However, this use introduces new engineering challenges because of temperature variation in the
foundation pile and surrounding soils. There will be additional deformations and forces in the bearing founda-
tion system, coupled thermohydro-mechanical phenomena in soil and some effects on the strength and settlement
properties of soil. In situ tests are powerful experimental tools as they give an insight into real conditions under
real scales. However, they are expensive and rather time-consuming. For this purpose, in this paper the finite
element method is used to study the bearing capacity and consolidation settlement of energy piles in fine-grained
soils. This method in comparison with experimental tests is inexpensive and needs less time. In this study the
thermal induced stresses and the variation of displacement in the pile and adjacent soils are studied.
Keywords: energy piles, finite element, thermal response of soils, bearing capacity, consolidation settlement,
fine-grained soils
1 INTRODUCTION experimentally and numerically, and found vertical

effective stress and strain of pile influenced by ther-
Due to the growth of energy consumption around the momechanical loading applied to pile. They reported
world, managing energy resources is one of the most that the intensity of vertical stress and strain induced
important problems of the current century. One of in pile depended on properties of surrounding soils.
the main energy resources is fossil energy which is In the view of comparison, thermomechanical load-
limited in terms of amount and production. More- ing induces more vertical stress along the pile than
over, its transfer is not only costly but also dan- mechanical loading. Webb et al. (2009) and Amatya
gerous. Furthermore, their environmental effects are et al. (2011) present a mechanism that can describe
drastic, increasing greenhouse effect, the spread of the pile response under thermomechanical loading
cancer along with other diseases in human society. imposed on it. They found that the intensity of induced
Consequently, researchers are seeking new ways to vertical stress in the pile depended on end restrained
discover further resources and substitute fossil fuels condition. Freitas et al. (2013) studied numerically the
with renewable energy ones which are inexpensive in effect of thermal expansion coefficient, mechanical
terms of costs and friendly with the environment. One properties of surrounding soils and thermal bound-
of such resources is geothermal energy. One of the ary on the thermomechanical behavior of energy pile.
newest methods of using geothermal energy is using After that Yavari et al. (2014) studied thermome-
energy foundations which couples the structural role of chanical behavior of energy pile in dry sand with
geostructures with energy supplies. It causes changes physical tests and observed that the total pressure
in the temperature of piles and surrounding soils. below the pile change intensively due to mechanical
Because of the fact that the primary role of these piles and Thermomechanical loading on the pile. Also, they
is the stability of the superstructure and the settlement found skin friction change under thermomechanical
of structure sensitivity, these effects must be studied loading. With numerical analysis Jeong et al. (2014)
to overcome the obscurity of temperature effects on studied mechanical response of pile group under sev-
the pile and soil response. A suitable and helpful tool eral conditions such as different pile arrangements and
for this study is implementing numerical methods, distance between piles and different soil type. Due to
especially the finite element method being efficiency the primary effect of temperature on the behavior of
approved by many researchers (Laloui (2006), Fre- fine grain soils such as high plasticity clays, Akrouch
itas (2013), Donna (2015) and etc.). Laloui et al. (2014) implemented experimental test in this condi-
(2003, 2006) studied the temperature effects on piles tion to study the response of pile to thermomechanical
57
loading and found the pile response and creep rate is
dependent on time and suggested the distribution of
energy piles should be as much symmetrical as possi-
ble to avoid differential settlement below the structure.
Moreover, temperature effect on the pile, it has impor-
tant effects on surrounding soils such as an increase in
initial elastic module and shear strength (Laloui et al.,
2004), volume change behavior and effective stress
(Mitchell, 2005). It is known that heating a soil might
induce excessive pore water pressure because the ther-
mal expansion coefficient of water is higher than the
one of solid skeleton. Another change is observed
when the horizontal effective stress increases during
heating and decreases during cooling because of the
radial thermal expansion of the pile (Donna, 2014).
So it is important to investigate the various effects of
Thermomechanical loading in a pile and soil response.
Since temperature changes stresses in soils, this paper
concentrates on the temperature effects on soils and
geotechnical aspects of this phenomenon.
2 FINITE ELEMENT METHOD
Finite element is one of the most common tools which

is used for the study of the Thermomechanical behav-
ior of energy pile by a plethora of researchers namely Figure 1. Finite element mesh, Model geometry and
Laloui et al. (2006), Jeong (2014). Moreover, it has mechanical boundary.
been approved that it should have suitable results con-
cerning experimental and physical tests. The present 3 MATERIAL PROPERTIES
model is based on a numerical simulation which was
implemented by Laloui et al. (2006). The model con- In this paper, surrounding soils are in line with
sists of a pile with 1m diameter and 26 m length in Drucker-Prager thermo-elastoplastic model. Pile and
the corner of a four-story building at the Swiss Fed- layer D behave as thermo-elastic material. Material
eral institute of technology in Lausanne. The software parameters were extracted from Laloui et al. (2006)
used in this paper is ABAQUS V6.13 encompassing and except for the expansion of the coefficient of layer
three analytical procedures: Heat Transfer, Geostatic B and C, there is no difference for inputted parameters.
and Soil consolidation. Results from Heat Transfer Batini (2015) reported lower a value for this parameter,
analysis was imported as predefined in soil consol- differing from Laloui et al. (2006). For the conversion
idation analysis to apply temperature effects in the of bulk modulus (K) and shear modulus (G) to the elas-
analysis. Figure 1 represents geometry, finite element tic modulus (E) and Poisson’s ratio (υ), Equation 1 and
mesh and mechanical boundary condition. The dimen- 2 were implemented. Material properties inputted are
sions of elements in soils are 0.5 m × 0.5 m and in illustrated in Table 1.
pile 0.25 m width and 0.5 m length. Element type
used in the Heat Transfer analysis is a 4-node lin-
ear axisymmetric heat transfer quadrilateral (DCAX4)
and in Geostatic and Soil Consolidation whose layers
surround the pile in an 8-node axisymmetric quadrilat-
eral, biquadratic displacement, bilinear pore pressure,
reduced integration (CAX8RP). The pile’s element
is an 8-node biquadratic axisymmetric quadrilateral
(CAX8) and for the layer under the pile an 8-node 4 INITIAL AND BOUNDRY CONDITION
biquadratic axisymmetric quadrilateral, reduced inte-
gration (CAX8R). Reduced integration property pre- Two sides of the model were restrained in horizontal
vents shear locking in soil elements (ABAQUS V6.13 direction, and the bottom of the model was restrained
user’s guide). Interaction between pile and the sur- in both vertical and horizontal directions. Soils were
rounding soils is assumed rough, and contact between permitted drainage from top and the right hand side of
the master surface and slave surface are tied together the model. Layer D and the pile were assumed imper-
to restraint relative movement between them; these vious and hence no drainage occurs from the soil-pile
assumptions are in good agreement with Laloui et al’s. interface. The thermal boundary consists of allowing
58
Table 1. Mechanical, thermal and hydrological properties of pile and soils.
Unit Permeability Young Friction Cohesion Thermal Heat Solid dilation

Soil weight ρ Porosity K modulus Poisson’s angle C conductivity capacity ρc coefficient βs
layer (kg/m3 ) η (m/s) E (MPa) ratio ν (◦ ) (kPa) (W/m/◦ C) (J/m3 ·◦ C) (◦ C−1 )
A1 2000 0.1 2×106 259 0.14 30 5 1.8 2.4×106 10−5

A2 1950 0.1 7×10−7 259 0.14 27 3 1.8 2.4×106 10−5
B 2000 0.35 1×10−6 451 0 23 6 1.8 2.4×106 2×10−5
C 2200 0.3 1×10−6 634 0 27 20 1.8 2.4×106 2×10−5
D 2550 – – 1273 0.16 – – 1.1 2.0 × 106 10−6
Pile 2500 – – 33700 0.176 – – 2.1 2.0 × 106 10−5
Note: Fluid dilation coefficient βf = 10−5
Figure 2. Temperature variation over tests.
heat to dissipate from the right hand side and the bot-
tom of the model. Also, the temperature in the top Figure 3. Validation of pile head displacement – test 1.
surface of the model is assumed constant.
pile. Finally, for validating the ability of current sim-

5 MECHANICAL AND ulation to predict the variation of vertical stress under
THERMOMECHANICAL LOADING thermomechanical loading, a comparison between ver-
tical stress in pile between results reported by Laloui
Laloui et al. (2006) assumed temperature as varying (2003, 2006) and current model was accomplished the
over time in test 1 and test 7 similar to Figure 2. In test 1, results of which are demonstrated in Figure 5. Alto-
the pile was free to expand and no restrained is imposed gether from figure 3, 4 and 5 it can be concluded that
on the head of the pile, but in test 7 a mechanical results show good agreement with the reported values
loading with a 1.3 MPa magnitude was imposed on by Laloui (2003, 2006).
the pile head which induced compression axial force
in the pile.
7 NUMERICAL RESULTS AND DISCUSSION
6 VALIDATION OF PROPOSED MODEL
7.1 Ultimate bearing capacity
The results of the Finite element analysis were com- In this section, the effects of temperature on soils were
pared with numerical and field load test data from studied. With respect to Figure 6, it can be compre-
Laloui et al. (2003, 2006). Figure 3 demonstrates ver- hended that temperature increasing result to increasing
tical head displacement of pile during test 1 based of vertical effective stress in soils surrounding the pile.
on numerical and experimental results from Laloui Donna (2014) observed horizontal effective stress in
(2003, 2006) and result from current model. It can be soil was influenced by the increasing of temperature
seen that there is good agreement between results pre- that cause radial thermal expansion of pile and the
sented by Laloui (2003, 2006) and current simulation increasing of horizontal effective stress in soils.
to predict head displacement under thermomechanical With respect to this observation and since tempera-
condition. Figure 4 displays comparison between verti- ture is increased in soils and soils are restrained from
cal strains along with the pile from Laloui (2003, 2006) thermal expansion by overpressure from the upper soil
and current model during test 1. It can be deduced body, it is expected that vertical effective stress con-
that current model have this ability to predict verti- tinues to increase. Based on Literature review, it was
cal strain resulted from thermomechanical loading on expected that increasing temperature in soils cause the
59
Figure 6. Mechanical and thermomechanical vertical stress
in soils — test 7.
mechanical and Thermomechanical loading based on

Figure 4. Validation of vertical strain in pile – test 1. CFEM (1992).
With Qt being the ultimate bearing capacity, Qb the tip

bearing capacity, Qs the skin friction bearing capacity,
qb the unit tip bearing, σt the vertical stress at tip level,
Nt the end bearing factor, Ap the pile area, qsi the skin
friction bearing capacity for each layer, fsi the unit skin
friction resistance for each layer, βi the skin resistance
factor, σv the averaged vertical stress in middle of each
layer, pp the pile perimeter and Li the length of each
layer.
Vertical effective stress increases with depth.
Hence, skin friction and tip bearing capacity should
increase with depth indefinitely. But in reality, Exper-
imental data indicate that skin friction and tip bearing
capacity will not increase with depth indefinitely
(Rajapakse, 2008).
Therefore, in this study for the purpose of simpli-
Figure 5. Validation of thermomechanical vertical stress in
fication assumed it keeps increasing up to a specific
pile – test 7. depth. With consideration of soil properties, critical
depth for both, tip and skin friction capacity were
assumed equal to 10D in which D is pile diameter.
increasing of pore water pressure. This occurs because Hence, the vertical effective stress was modified the
of the difference in thermal expansion coefficient of same as Figure 7. Table 2 and 3 show the variation
pore water and soil’s grains, but since thermal expan- of tip and skin friction bearing capacity of piles under
sion coefficient of pore water is less or equal to the mechanical and thermomechanical conditions. For the
thermal expansion coefficient of soils, an increase sake of safety, the unit skin friction resistance of layer
in pore water pressure is not expected to occurred. D has not been considered here. For unit skin friction
Whereby ultimate bearing capacity of pile depends on in layer A1, A2, B, C and D Laloui et al. (2006) report
vertical effective stress, its cause relies on the enhance- 0, 0, 30, 165 and 300 kPa respectively and considering
ment of bearing capacity of piles. Equations 3, 4 and tip bearing capacity, they reported 11 MPa. The dif-
5 show total, tip and skin bearing capacity of the ference between the values reported by Laloui et al.
pile, respectively. These equations are directed for the (2006) and results presented in this paper is because
computation of the bearing capacity of piles under of the exclusion of critical depth for vertical stress
60
by Laloui et al. (2006). From Figure 8 with respect changed due to temperature effects in the pile per se.
to variation of unit skin friction resistance, it can be For axial deformation of pile, the formula is as follows:
deduced that thermomechanical loading can result in
an increase or decrease of skin friction bearing capac-
ity of piles and depending on soil type, properties of
materials, and restraint of the pile in the soil and mag-
nitude of thermomechanical loading which is imposed
on the pile. with Qp is transmitted load to the soil at the tip, Qs
7.2 Pile and soil settlements is transmitted load to pile skin, Ep the pile elastic
modulus, A the pile area, L the pile length, a being a
Since the change in temperature results in the variation
coeffiecent equal to 0.5 for clay and 0.67 for sands. For
of vertical effective stress in pile and surrounding soils,
the inclusion of layered soil effects in the calculation,
it is expected that the settlement of pile be affected too.
aQs is to be calculated from Equation 7.
7.2.1 Pile settlements
Due to temperature effects on pile axial vertical stress,
a change in pile settlement is expected. So by means
of semi-empirical approach, the settlement of the pile
was calculated. Equation 6, 8 and 10 represent set-
with qs the skin friction mobilized in each condition,
tlements due to axial deformation of pile, skin friction
pp the pile perimeter and L the pile length. Since
and pile point respectively. With respect to Figure 9, the
A1 and A2 are fine-grained soil and B and C are
mobilized friction of soil-pile interface under mechan-
coarse grain soil, aQs was calculated separately for
ical and thermomechanical loading changed so, it can
each class of soil. The axial deformation in mechani-
be deduced that axial deformation of the pile will be
cal and thermomechanical condition is summarized in
Table 4.
Figure 7. Applied vertical stress in mechanical and thermo- Figure 8. Unit skin friction resistance in mechanical and
mechanical condition. thermomechanical conditions along pile length.
Table 2. Comparison between tip bearing capacity in mechanical and thermomechanical condition.

Depth Ap σt=0 σt=13.4 qbt=0 qbt=13.4 Qbt=0 Qbt=13.4
◦
Layer m m2 kPa kPa Nt kPa kPa kN kN
D 25.5–52 40 0.785 101.55 124.13 72 7311.26 8937.58 5739.34 7016.00
Table 3. Comparison skin friction bearing capacity in mechanical and thermomechanical condition.

Depth Length σvt=0(avg) σvt=13.4(avg) pp fst=0 fst=13.4 qst=0 qst=13.4
◦
Layer m m kPa kPa β m kPa kPa kN kN
A1 0–5.5 5.5 30 40.57 32.12 0.32 3.14 12.98 10.28 224.22 177.50
A2 5.5–12 6.5 27 81.13 101.57 0.28 3.14 22.72 28.44 463.62 580.44
B 12–22 10 23 101.55 124.13 0.30 3.14 30.46 37.24 956.56 1169.33
C 22–25.5 3.5 27 101.55 124.13 0.35 3.14 35.54 43.45 390.59 477.48
61
Table 4. Transmitted load and axial deformation of pile in is mobilized in the upper part of the pile parallel to
mechanical and thermomechanical conditions. mechanical loading direction. In the lower part, it is
reversed the mechanical loading direction. Hence the
Qp aQs Sa settlement due to skin friction will be varied. The value
of settlement due to skin friction is presented inTable 6.
kN kN mm
For pile point settlement:
Mechanical 248.45 369.22 0.607
Thermomechanical 564.8 125.2 0.678
Table 5. Emprical coefficent value for different soil types.

qb is the ultimate tip bearing capacity and D the pile
diameter. The magnitude of pile point settlement is
Soil Type Cp for Bored Pile summarized in Table 7.
Based on total settlement in Table 8, it can be seen
Dense Sand 0.09 the total settlement of the pile was increased in Ther-
Loose Sand 0.18 momechanical loading condition. Also, it is worth
Soft Clay 0.06 noticing that Figure 9 has good agreement with Ther-
momechanical mechanism proposed by Amatya et al.
(2011) for thermomechanical behavior in heating.
7.2.2 Soil settlements

Due to an increase in vertical effective stress in soils
similar to Figure 6, it can be deduced that the set-
tlement of the surrounding soil will be changed in
thermomechanical conditions. Hence, soils surround-
ing the pile experiment have a different settlement
nature from those under mechanical conditions. Since
there exist fine-grained soils in the field in layer A1
and A2, one of the primary soil settlement is consol-
idation settlement. As Laloui et al. (2004) and Naga
(2007) reported, compression index of clay will not
be affected by temperature and will be approximately
constant. Based on consolidation theory and consider-
ing Equation 12, volume compression coefficient mv
depends on the initial void ratio, compression index
Cc and variation of vertical effective stress. Therefore,
it can be concluded that mv will be changed and so it
causes a change in consolidation settlement of ther-
Figure 9. Mobilized skin friction in mechanical and thermomechanical condition compared with mechanical
momechanical condition. condition.
Consolidation settlement can be obtained from
For skin friction settlement: Equation 11 and mv was calculated from Equation 12.
Cs is an empirical coefficient which can be calculated

from Equation 9.
In this numerical simulation, vertical effective stress in
soils varied over time, depth and until precise distance
from pile skin. The affected distance by temperature is
clearly shown in Figure 10. It can be observed that the
In this equation Cp is an empirical coeffiecent the temperature effect up to a precise distance from pile is
values of which for different soil types are listed in approximately about 1.5 m from the pile. For simplify-
Table 5, with L being pile length and D the pile ing purposes, variation of vertical effective stress over
diameter. As there are two classes of soil, skin fric- time, depth and affected distance from the pile (1.5 m)
tion settlement was calculated for each one separately. was averaged linearly. Value of averaged vertical effec-
In mechanical loading, skin friction acts reversely to tive stress in initial, mechanical and Thermomechan-
the direction of mechanical load, but with respect to ical conditions are listed in Table 9. Considering the
Figure 9, it was observed under Thermomechanical fact that there is no available data to specify the com-
condition that it has more complicated behavior and pression index, it is assumed that it is equal to 0.05.
62
Table 6. Pile settlement due to skin friction.
Qs1 Qs2 Ssf

Cp1 Cp2 Cs1 Cs2 kN kN mm
Mechanical 0.06 0.18 0.3054 0.9162 139.24 244.23 1.78

Thermomechanical 0.06 0.18 0.3054 0.9162 −240.1 365.3 1.43
Table 7. Transmitted load and pile point settlement in

mechanical and thermomechanical condition.
Ssf St Sa S
mm mm mm mm
Mechanical 1.78 3.9 0.6 6.28

Thermomechanical 1.43 7.2 0.67 9.3
Table 8. Total settlement.
Ssf St Sa S
mm mm mm mm
Mechanical 1.78 3.9 0.6 6.28

Thermomechanical 1.43 7.2 0.67 9.3
According to Solanki and Desai (2008), research on

alluvial clay represents a correlation between primary
properties of soils to compression index. But the ini-
tial void ratio in this case is different from what they
already mentioned. It can be concluded that tempera-
ture causes the increasing of consolidation settlement.
It is worth noticing that vertical effective stress in adja-
cent soils will be increased, whereas in distance, after
this length, it decreased. Similar to this decrease was Figure 10. Vertical Stress variation in soils along time,
reported by Yavari (2014) who observed that the total depth, distance from pile shaft.
pressure in soils far from the pile will be decreased
over heating the pile. Table 9. Variation of consolidation settlement in mechan-
ical and Thermomechanical condition.
8 CONCLUSION σ0 σ mv H
kPa kPa e0 Cc m2 /kN mm
A numerical simulation was implemented to study
Mechanical 57.83 3.97 0.1 0.05 1.31 × 10−3 62.4
the variation of piles behavior under Thermomechan- Thermo- 57.83 8.81 0.1 0.05 2.8 × 10−3 296
ical loading. It was observed that the behavior of pile mechanical
and soils will be affected intensively. Temperature has
important effects on the behavior of piles and soils. It
depends on soil type, properties of soils and magnitude
(2) Thermomechanical loading can cause decrease or
of imposing mechanical and also the thermal load on
increase of bearing capacity of piles. Its variation
the pile.
depends on thermal properties of pile and the
In this case it was observed that:
surrounding soils, magnitude of mechanical and
(1) Temperature causes variation of soil stress, which thermal load imposed on pile and restraint of pile
can be the result of thermal expansion of the pile in soil.
and probably the effect of overpressure on soil (3) Settlements of pile and surrounding soils can be
blocking thermal expansion of soils. This conclu- influenced by thermomechanical loads compar-
sion is based on variation of vertical stress in soils ing with mechanical loads. This is because of the
over depth. By heating, vertical stress in shallow thermal loading that changes the deformation of
depth decreases whereas it increases in deep depth. the pile and soil.
63
Altogether, each case needs to be examined to Di Donna, A. and L. Laloui (2015). “Numerical analysis
ensure that the design of the pile for applied mechan- of the geotechnical behaviour of energy piles.” Interna-
ical and thermomechanical loads is correct. For this tional journal for numerical and analytical methods in
purpose, numerical analysis is an appropriate tool. geomechanics 39(8): 861–888.
Jeong, S., et al. (2014). “Thermally induced mechanical
response of energy piles in axially loaded pile groups.”
Applied Thermal Engineering 71(1): 608–615.
REFERENCES Laloui, L., et al. (2003). “Comportement d’un pieu bi-
fonction, fondation et échangeur de chaleur.” Canadian
Abuel-Naga, H. M., et al. (2007). “Thermally induced volume
Geotechnical Journal 40(2): 388–402.
change and excess pore water pressure of soft Bangkok
Laloui, L., et al. (2006). “Experimental and numerical inves-
clay.” Engineering Geology 89(1): 144–154.
tigations of the behaviour of a heat exchanger pile.”
Akrouch, G. A., et al. (2014). “Thermo-mechanical behavior
International journal for numerical and analytical methods
of energy piles in high plasticity clays.” Acta Geotechnica
in geomechanics 30(8): 763–781.
9(3): 399–412.
Mitchell, J. and K. Soga (2005). “Fundamentals of soil
Amatya, B., et al. (2012). “Thermo-mechanical behaviour of
behavior.” Fundamentals of soil behavior (Ed. 3), Wiley
energy piles.” Géotechnique 62(6): 503–519.
New York.
Batini, N., et al. (2015). “Energy and geotechnical behaviour
Rajapakse, R. A. (2008). Pile design and construction rules
of energy piles for different design solutions.” Applied
of thumb, Butterworth-Heinemann.
Thermal Engineering 86: 199–213.
Saggu, R. and T. Chakraborty (2015). “Cyclic thermo-
Bodas Freitas, T., et al. (2013). The response of energy foun-
mechanical analysis of energy piles in sand.” Geotechnical
dations under thermo-mechanical loading. Proceedings of
and Geological Engineering 33(2): 321–342.
the 18th International Conference on Soil Mechanics and
Solanki, C. and M. Desai (2008). Preconsolidation Pres-
Geotechnical Engineering.
sure from Soil Index and Plasticity Properties. The 12th
Bourne-Webb, P., et al. (2009). “Energy pile test at
International Conference of International Association for
Lambeth College, London: geotechnical and thermody-
Computer Methods and Advances in Geomechanics. Goa,
namic aspects of pile response to heat cycles.” Géotech-
India. 2008, Citeseer.
nique 59(3): 237–248.
Yavari, N., et al. (2014). “Experimental study on the mechan-
Cekerevac, C. and L. Laloui (2004). “Experimental study
ical behaviour of a heat exchanger pile using physical
of thermal effects on the mechanical behaviour of a clay.”
modelling.” Acta Geotechnica 9(3): 385–398.
International journal for numerical and analytical methods
in geomechanics 28(3): 209–228.
64
Differences and challenges in salt cavern design

for hydrogen, air and natural gas storage
M. Schlichtenmayer & M. Klafki

ESK GmbH, Freiberg, Germany
ABSTRACT: In the gas storage industry a growing demand for the storage of alternative gases like hydrogen
and compressed air is expected for the coming decades. Since natural gas storage in salt caverns gas is a mature
technology, substantial knowledge can be transferred to the storage of hydrogen and compressed air. However,
several open questions remain including the impact of the different thermodynamic gas properties on cavern
design or material requirements for the subsurface equipment. Furthermore, the conversion of existing caverns
from natural gas to hydrogen or air storage can be a promising option for cost reduction in this context, but
poses additional technical challenges. In this paper an overview of the main differences between the storage of
hydrogen, air and natural gas is given focusing on typical storage design requirements, cavern thermodynamics
and requirements on the subsurface equipment. The gases are compared regarding their compressibility, density,
specific heat capacity and viscosity and the impact of the different properties on storage capacity and performance
is discussed. Exemplary, the main steps of a technical feasibility evaluation for the conversion of an existing
natural gas cavern for alternative gas storage are presented. It can be concluded that, in most cases, the conversion
of an existing cavern is technically possible and provides a cost-efficient alternative to the leaching of a new
cavern. The findings facilitate the further development of technologies for underground energy storage and help
identifying new business perspectives for underground gas storage operators.
1 INTRODUCTION new business fields like hydrogen or air storage. The

conversion of existing gas storage caverns for hydro-
In the last years a growing interest could be observed gen or air storage is such an option, since it can allow
within the gas storage industry for alternative gases the usage of a cavern as an existing asset in a more prof-
like hydrogen or air. This is connected with the sub- itable way. The actual cost saving potential depends on
stantial changes of the energy system that are currently the cavern properties and on the planned hydrogen or
taking place in several countries foremost in cen- air storage profiles.
tral Europe. Even though natural gas as an energy
carrier is generally regarded as a partner of this ongo-
ing transition, storage operators seem interested in a
diversification of their business concepts. 2 FUNDAMENTALS
In this context hydrogen and compressed air both
are energy carriers, which can be stored using the same The different thermodynamic behavior of hydrogen,
technologies as for natural gas storage. However, in air and natural gas is caused by their different molecu-
contrast to natural gas both can be generated from elec- lar masses, molecule dimensions and intermolecular
tricity allowing them to serve as buffers for surplus forces. For technical applications like gas storage
renewable electricity. these microscopic properties are converted into macro-
The general storage design is the same for hydrogen, scopic quantities like density or heat capacity. Many
air and natural gas. Also, the techniques for design, real gas properties can then be conveniently described
erection and operation of the storage can be transferred by the deviations from an ideal gas. An example for
from natural gas to hydrogen and air. However, some this is the real gas law in the formulation employing
differences between the gases have to be observed in the compressibility factor Z:
order to get an optimized cavern for the storage of
hydrogen or compressed air. Especially, with respect
to the thermodynamic behavior of the gases and to the
material requirements on the equipment hydrogen and where p = pressure; V = Volume; n = amount of gas;
air can differ substantially from natural gas. T = Temperature and R = 8.314 J/mol K.
The storage operators are currently working under Gas properties like the compressibility factor or the
high economic pressure making it necessary to identify specific heat capacity determine the behavior of gas
cost saving options to facilitate the development of storage caverns. During the design phase of a cavern
65
Figure 1. Compressibility factor of hydrogen, air and Figure 3. Specific heat capacity of hydrogen, air and natural
natural gas at 30◦ C. gas at 30◦ C.
Figure 2. Density of hydrogen, air and natural gas at 30◦ C. Figure 4. Dynamic viscosity of hydrogen, air and natural
gas at 30◦ C.
usually numerical simulations based on the gas proper- Table 1. Composition of air and natural gas (L-gas).
ties are carried out to make predictions on the pressure
and temperature response of the cavern. These allow Natural gas Air
an optimization of cavern parameters like depth, vol-
ume or well diameter to fit to the planned operation Component Vol % Vol %
cycles without oversizing the cavern or the well.
CH4 81.554 –
C2 H6 2.916 –
C3 H 8 0.420 –
3 COMPARISON OF HYDROGEN, AIR AND n-C4 H10 0.143 –
NATURAL GAS C5 H12 + 0.082 –
N2 13.723 78.084
The thermodynamic behavior of the three gases hydro- CO2 1.162 20.942
O2 – 0.038
gen, air and natural gas differs significantly from each
Ar – 0.936
other. Especially natural gas consists of rather large
molecules, which can interact with each other, causing
a strong deviation from ideal gas behavior. On the other Natural gas has the lowest compressibility factor
hand, hydrogen, consisting of lightweight two-atomic of the three gases meaning that for storing the same
molecules with only weak intermolecular interactions, amount of gas a lower pressure is required for natu-
behaves almost like an ideal gas. These differences ral gas than for air or hydrogen. In other words, the
are visible in various gas properties, which are impor- working gas volume of a cavern operated within fixed
tant for the storage of gas in salt caverns, especially in pressure limits is highest for natural gas (and lowest
the compressibility (Figure 1), density (Figure 2), spe- for hydrogen).
cific heat capacity (Figure 3) and viscosity (Figure 4). Furthermore, the maximum rates for injection and
The gas compositions used in this paper are stated in withdrawal of gas from a cavern are often limited by
Table 1. The gas properties shown in this paper are the rate of pressure change at the last cemented cas-
in good agreement with the values stated in literature ing shoe due to geomechanical integrity constraints.
(Voigt 2014a, b, c, Kunz et al. 2007, Vargaftik et al. A low compressibility factor also allows injecting or
1996). withdrawing a larger amount of gas upon a fixed rate
66
of pressure change. Therefore, the performance of a
given cavern is higher for storing natural gas than for
air or hydrogen.
Thus, the thermodynamic properties have a direct
impact on the profitability of a cavern for storing dif-
ferent types of gases, since income is usually based on
the working gas volume and the maximum injection
and withdrawal rates.
The density of the gas has an indirect impact on
the cavern operation, since it determines the weight
of the gas column in the well and, therefore, the static
pressure difference between well head and cavern. For
air, the density is slightly higher than for natural gas
leading to higher static pressure differences. The den-
sity of hydrogen is approximately a factor of 10 lower
than the other gases and, accordingly, hydrogen shows
almost no static pressure difference between well head
and cavern. Since usually the operating pressure range
at the last cemented casing shoe is fixed, this can cause
higher operating pressures in the surface facilities for Figure 5. Energy density and cycle efficiency for natural
a hydrogen storage facility than for air or natural gas gas (or synthetic methane), hydrogen and air.
storage.
The differences between the gases regarding the to an adiabatic compressed air energy storage system
specific heat capacity can lead to rather complex with a cycle efficiency of 70 % using the full pressure
differences in the temperature development of the cav- range as stated above (note that other systems may have
erns. Natural gas has the highest specific heat capacity lower maximum pressures (Zunft & Freund 2014)).
and hydrogen the lowest. Thus, natural gas can trans- This comparison shows that the amount of energy that
port heat more effectively and can lead to a stronger can be stored in a cavern with natural gas is much
coupling of the cavern temperature and the tempera- higher than for the same cavern with hydrogen or air.
ture at the well head during injection and withdrawal Additionally, the power-to-power cycle efficiency
of gas. For hydrogen this coupling is much weaker, for energy storage systems based on methane (rep-
so, less heat is transferred with the gas flow during resenting natural gas), hydrogen and air (adiabatic
injection and withdrawal. The temperature distribu- compressed air energy storage (Zunft & Freund 2014))
tion in this case is driven stronger by horizontal heat are illustrated in Figure 5. The values for methane and
conduction in the surrounding rock. hydrogen are taken from literature and refer to full
In this context it should also be noted that under systems including electrolysis, storage operation and
typical storage conditions hydrogen has a very small a fuel cell or an additional methanization step and a
negative Joule-Thomson coefficient and does not cool combined cycle gas turbine (Hartmann et al. 2012,
down upon a pressure drop as it is known for air Sterner & Stadler 2014). This comparison shows that
and natural gas. Instead, it shows a slight temperature from a technical point of view air is a good choice as
increase, which is negligible for underground storage an energy carrier for storing energy with a high cycle
applications. efficiency. Hydrogen (power to gas (H2 ) to power) is
Hydrogen is also known for its very low viscos- suited better if a higher storage density is required
ity, which is approximately 50 % lower than for air and a lower efficiency can be tolerated. For even
or natural gas. Therefore, hydrogen storage caverns higher storage densities an additional methanization
show very low dynamic pressure losses during injec- step can be included in the power to gas system to pro-
tion and withdrawal of gas. Regardless of the current duce synthetic methane at the cost of an even lower
operation mode, in hydrogen storage caverns the well efficiency.
head pressure is always almost the same as the cavern Additional arguments for the choice of an energy
pressure. storage medium from an economic point of view are
A more detailed discussion of the gas properties the availability and cost of the technical equipment.
including thermodynamic cavern simulations can be Especially, the electrolysis and the methanization still
found in Schlichtenmayer et al. (2015). act as bottlenecks, since the capacities available today
It is also useful to compare the amount of energy are in the order of a few kW to MW.
that can be stored in a given cavern using different
storage gases.This is illustrated in Figure 5 for a typical
cavern in the depth of 1000 m (last cemented casing 4 CONVERSION OF EXISTING CAVERNS FOR
shoe) operating in the pressure range between 4 MPa HYDROGEN OR AIR STORAGE
and 18 MPa. For natural gas and hydrogen the higher
heating value is used to convert working gas volumes In a changing energy system it can be profitable to con-
into energy contents. For air the energy content refers vert a cavern from natural gas storage to hydrogen or
67
air storage. Historically, caverns have been converted Table 2. Summary of main cases for cavern conversion.
for instance in the late 20th century in the eastern The costs are compared to the erection of a new cavern in
part of Germany, when former town gas storages were a developed location with leaching infrastructure.
switched to natural gas. However, in these cases the
Required works Costs Comments
requirements on the converted caverns were different
than they are for the storage of hydrogen or air.
Leaching and almost as economically
Generally, the conversion of a cavern for alternative well upgrading new cavern not feasible
gases starts with an in-depth assessment of the existing
Leaching only mainly cavity unlikely for
cavern and well. Depending on the results several main
costs H2 or air storage
conversion steps will follow as outlined below:
Major well similar to new technically most
1. Assessment of existing cavern and well upgrading well & completion likely
2. Flooding of the cavern New completion mainly equipment if casing integrity
3. Additional cavern leaching (optional) only & installation is guaranteed
4. Removal of existing well completion Flooding only almost negligible unlikely for
5. Well and casing operations H2 or air storage
a. Upgrading of last cemented casing or
b. Drilling of new well and abandonment of old In summary, the following questions need to be
well answered for assessing the feasibility of a cavern
6. Installation of new well completion conversion:
7. First fill with new storage gas 1. Do the cavern size and depth and the completion
This is a general outline of the conversion process diameter fit to the planned gas storage cycles? Is
and some steps may be skipped if the situation allows. it necessary to enlarge the cavern or drill a larger
For instance, the flooding of the cavern could be omit- well to meet these requirements?
ted, if the new storage gas does not react chemically 2. Are the existing casing and well completion suited
with natural gas and the purity requirements for the for the planned type of gas (especially regarding
storage gas are low and it is not required for a safe material compatibility)?
conduction of any work at the cavern well. However, 3. Is it necessary to protect the existing last cemented
these three conditions will only coincide in rare cases. casing from direct contact with the new storage
The additional leaching of the cavern is necessary if gas? Which additional installations are required to
the original cavern volume is too small or if an infiltra- achieve this?
tion of natural gas into the cavern wall is expected and 4. What are the preconditions to conduct the work on
a later re-migration of this gas into the cavern should these installations?
be avoided. In this case, the necessary extent of the 5. To what extend can a mixture of original and new
additional leaching is a result of the previous cavern storage gas be tolerated?
assessment. 6. Is a flooding or additional leaching of the cavern
One of the main questions for the conversion of a necessary?
cavern is, whether the existing last cemented casing is 7. What are the costs and the time consumption for
suited for the new gas or can be upgraded with rea- the complete cavern conversion process?
sonable effort. In many cases a standard API-steel is In most cases the conversion is technically possi-
used for the last cemented casing. These types of steel ble, but an economically feasible option usually only
are not optimized for highly oxidizing environments exists if either the initial cavern volume fits to the
like moist high-pressure air. For hydrogen these types requirements or if the well can be reused without major
of steel may be suited better, but in a real conversion alterations. In the best cases only the well completion
situation an individual evaluation based on the exact needs to be exchanged or even all components can
alloy composition including the material stress condi- be reused. A summary of the major cases is given in
tions in operation is required to ensure the long-term Table 2.
material integrity.
If the integrity of the casing material cannot be guar-
anteed, it must be protected from direct contact to the 5 CONCLUSION
gas. An option to achieve this can be to install an addi-
tional cemented casing, which has to reach down into In this paper it is shown that the different thermody-
the cavern neck. Requirements for the installation are namics of hydrogen, air and natural gas has an impact
that the cavern neck is sufficiently long and that a on the design and operation of a gas storage cavern.
temporary plug can be set in this section. This option The highest energy density is achieved with natural gas
reduces the inner diameter of the well and, thus, the followed by hydrogen and air. On the other hand, syn-
well and cavern performance. Therefore, this option thetic methane as a substituent for natural gas can only
may work better for a hydrogen storage cavern than be produced from electricity with low cycle efficiency.
for an air storage cavern, since the well performance Using compressed air or hydrogen as energy carriers
is more critical for the latter. allows higher cycle efficiencies.
68
The main steps for the conversion of an existing Kunz O. et al. 2007.The GERG-2004Wide-Range Equation of
natural gas storage cavern for hydrogen or air stor- State for Natural Gases and Other Mixtures. Düsseldorf:
age and the main aspects of a feasibility assessment VDI Verlag GmbH.
have been outlined. Technically, a conversion is possi- Schlichtenmayer M. et al. 2015. Renewable Energy Storage
in Salt Caverns – A Comparison of Thermodynamics and
ble in many cases. The economic feasibility, however, Permeability between Natural Gas, Air and Hydrogen.
depends strongly on the current state of cavern and Clarks Summit: SMRI.
well and on the requirements of the planned storage Sterner M. & Stadler I. 2014. Energiespeicher – Bedarf,
cycles. Technologien, Integration. Berlin Heidelberg: Springer.
There are still some open topics before a widespread Vargaftik N. B. et al. 1996. Handbook of Physical Proper-
usage of salt caverns for air and hydrogen storage, ties of Liquids and Gases: Pure Substances and Mixtures.
mostly concerning the optimization of completion New York: Begell House.
materials and certification of equipment. Current Voigt H. D. 2014a. Thermodynamische und Transport-
and planned research projects like ADELE or Hypos Eigenschaften von Wasserstoff – I. Realgasfaktor und
Dichte. Erdöl Erdgas Kohle 130(5): 192–194.
are expected to bring valuable results for closing Voigt H. D. 2014b. Thermodynamische und Transport-
remaining gaps of knowledge and developing meth- Eigenschaften von Wasserstoff – II. Joule-Thomson-
ods required to demonstrate the technical possibility Effekt und spezifische Wärmekapazität. Erdöl Erdgas
of cavern conversion and storage operation. Kohle 130(7/8): 297–280.
Voigt H. D. 2014c. Thermodynamische und Transport-
Eigenschaften von Wasserstoff – IV. Dynamische
Viskosität, Schallgeschwindigkeit und Wärmeleitfähigkeit.
REFERENCES Erdöl Erdgas Kohle 130(11): 444–446.
Zunft S. & Freund S. 2014. Large-scale Electricity Storage
Hartmann N. et al. 2012. Stromspeicherpotenziale für with Adiabatic CAES – the ADELE-ING project. Energy
Deutschland. Universität Stuttgart. Storage Global Conference, Paris, 19–21. Nov. 2014.
69
Effects of transient phenomena on the behaviour of thermo-active piles
K.A. Gawecka, D.M. Potts, D.M.G. Taborda, W. Cui & L. Zdravković

Imperial College London, London, UK
ABSTRACT: Thermo-active geotechnical structures are becoming increasingly popular due to their dual
function – providing stability as foundations and low-carbon heating and cooling as heat exchangers. Although
they have been used for decades, their geotechnical performance is yet to be fully understood. Current practice
generally involves the application of large factors of safety, which may lead to an over-conservative design. In
this paper, the thermo-mechanical behaviour of a single pile is investigated through a series of numerical analyses
performed using the Imperial College Finite Element Program (ICFEP), which is capable of simulating the fully
coupled thermo-hydro-mechanical (THM) behaviour of porous materials. Particular emphasis is given to the
transient thermo-mechanical response of the surrounding soil, which influences greatly the overall behaviour of
the pile. This is accomplished by performing parametric studies focussing on the adopted thermo-mechanical
properties of the soil. Lastly, implications for the design of these thermo-active structures are investigated.
1 INTRODUCTION The limited field and experimental data have led to

investigations of the behaviour of thermo-active piles
Shallow geothermal energy – i.e. the heat energy stored through numerical analysis (e.g. Knellwolf et al., 2011,
in the ground up to depths of 300 m (Banks, 2012), is Di Donna et al., 2016). Such studies vary in the com-
a renewable energy source capable of providing a reli- plexity of the numerical formulation and focus mostly
able and effective means of space heating and cooling. on reproduction of field tests and/or the effects of ther-
Buildings’ foundations have a great potential for mal loads on the pile-soil interaction. Bourne-Webb
utilising this thermal energy if installed with heat et al. (2016) examined the effect of pile length, ground
exchanger pipes and coupled with a ground source surface temperature and the soil’s thermal expansion
heat pump. The major advantage of these structures coefficient, which is a source of a great uncertainty,
over other types of closed-loop systems is their dual however, only a thermal-steady state condition was
function: providing stability as foundations and energy considered.
as ground source heat exchangers. The heat trans- This paper presents a series of coupled Thermo-
fer is also significantly higher than in borehole heat Hydro-Mechanical (THM) finite element simulations
exchangers owing to the higher thermal conductivity of a single pile performed using the Imperial College
and heat storage capacity of concrete. Finite Element Program – ICFEP (Potts & Zdravković,
Piles, which are the focus of this paper, are the most 1999). The parametric study aims to explain the
common foundation structures used as heat exchang- effects of the soil’s properties – coefficient of thermal
ers. Although they have been in use for over 30 years expansion and thermal conductivity – on the transient
(Brandl, 2006), their behaviour and performance have response of the soil and the mechanical behaviour of a
only been studied during the last 10 years. Despite thermo-active pile. It should be noted that the issue of
this, few issues regarding serviceability have been thermal performance is not considered in this paper.
reported (Loveridge & Powrie, 2013). The reason for
this may be highly conservative design, since cur-
rent practice involves greater factors of safety than
otherwise employed.
2 FINITE ELEMENT MODELLING
Field and laboratory tests investigating the thermo-
mechanical behaviour of piles (e.g. Bourne-Webb
2.1 Governing formulation
et al., 2009, Stewart & McCartney, 2013) as well as
their geothermal performance (e.g. Pahud & Hubbuch, ICFEP has recently been upgraded to model fully cou-
2007) have provided an invaluable insight into this pled THM behaviour of porous materials. The program
relatively new technology. However, further studies, accounts for heat transfer through conduction as well
especially on the accurate simulation of their thermo- as convection.
mechanical behaviour and the long term effects on Another feature of ICFEP that is of particular rel-
both the structure and the surrounding soil, are evance to this study is the ability to simulate the
required. generation of pore water pressures due to changes
71
in temperature, caused by the difference in the ther-
mal expansion coefficients of the soil particles and the
pore water, which can be described by the following
expression:
where u is the change in pore water pressure, T is

the change in temperature, n is the porosity, αw and αs
are the linear thermal expansion coefficients of pore
water and soil particles and Kf and Ks are the bulk
moduli of pore fluid and soil particles, respectively.
The details of the finite element formulation,
together with the program’s validation, can be found Figure 1. (a) finite element mesh and (b) zoomed in mesh
in Cui (2015). detailing the pile.
where RG,min and RK,min are the minimum normalised

2.2 The Imperial College Generalised Small-Strain values of Gtan and Ktan , respectively, a, b, r and s are
Stiffness (IC.G3S) model degradation parameters, and Ed and εvol are the devia-
Small-strain stiffness models have been widely used in toric and the volumetric strain invariants, respectively,
numerical analysis in conjunction with elasto-plastic defined as:
models in order to improve the prediction of soil
behaviour in the small to medium strain range. In
this paper, the Imperial College Generalised Small-
Strain (IC.G3S) model, which accounts for the effects
of a number of variables on the soil stiffness, was
used in all simulations. This section summarises the
model’s capabilities, with the complete formulation 3 NUMERICAL ANALYSIS
being described in Measham et al. (2014) and Taborda
et al. (2016). 3.1 Problem description
The maximum shear and bulk moduli are calculated
using: The parametric study presented in this paper involves a
single 30 m long pile with a diameter of 1.0 m installed
in London Clay.
The initial ground conditions are characterised by
a hydrostatic pore water pressure profile with ground
water level at the ground surface, a K0 value of 1.0 and
a uniform temperature of 15◦ C. The bulk unit weight
of the soil is 20 kN/m3 .
where G0 and K0 are the maximum shear and bulk The pile was wished in place, subjected to a mechan-
stiffness at a reference mean effective stress, pref , and ical load of 3300 kN and subsequently cooled by 15◦ C,
mG and mK control the stiffness dependency on the which is representative of a thermal load applied in
mean effective stress, p . practice (Brandl, 2006).
The model is also capable of taking into account the
influence of void ratio on the stiffness of the material, 3.2 Numerical modelling
though for simplicity such an effect is not considered
in this study. In order to investigate the thermo-mechanical
The degradation of the tangent shear and bulk behaviour of the considered pile, axisymmetric cou-
moduli is calculated from: pled THM analyses were performed using ICFEP.
Figure 1 shows the finite element mesh consisting
of 1375 quadrilateral elements, each with displace-
ment and temperature degrees of freedom at eight
nodes, and pore water pressure degrees of freedom
at the four corner nodes.
The concrete pile was modelled as linear elastic,
whereas the soil was modelled as non-linear elastic-
plastic with a Mohr-Coulomb failure surface and the
IC.G3S small-strain stiffness model. The Jardine et al.
(1986) small-strain stiffness model parameters for
72
Table 1. Material properties. Table 2. List of analyses.
Concrete Analysis Loading Thermal state Soil properties
Young’s modulus, Poisson’s A Mechanical – –

E (kPa) 20 × 106 ratio, µ (–) 0.15 B – –
Coefficient of 1.2 × 10−5 C λ, αs
thermal expansion, D1 Mechanical 0.1λ, αs
α (m/m/K) D2 & thermal Transient 10λ, αs
E1 λ, 0.1αs
Soil E2 λ, 10αs
Specific weight, γs 20 G0 (kPa) 16653.8

(kN/m3 )
Coefficient of thermal 1.2 × 10 −5
K0 (kPa) 21400.7 thermal pipes or the heat conduction between the ther-
expansion of soil, mal pipes and the concrete. It is important to note that
αs (m/m/K) cooling the pile over 10 days, while more realistic,
Coefficient of thermal 6.9 × 10−5 pref (kPa) 100 should not be regarded as a fully undrained process.
expansion of pore A time-step of 1.0 day was chosen for theTHM anal-
water, αw (m/m/K) ysis, based on the expressions and recommendations
Volumetric heat 2970 mG (–) 1.0 of Cui et al. (2015), in order to avoid the numerical
capacity, ρCp shock problem.
(kJ/K/m3 ) All analyses presented in this paper are summarised
Thermal conductivity, 2.0 mK (–) 1.0
in Table 2. In order to facilitate comparison, Analysis
λ (W/m/K)
Permeability, kw (m/s) 2 × 10−10 a (–) 7.22 × 10−4
A was isothermal, i.e. the pile was loaded and the soil
c (kPa) 5.0 b (–) 0.9757 was allowed to consolidate for six months. In Anal-
ϕ (◦ ) 22.0 RG,min (–) 0.1133 ysis B, loading and cooling of the pile was followed
ψ (◦ ) 0.0 Gmin (kPa) 2667 by six months of consolidation, with no heat trans-
r (–) 1.23 × 10−4 fer in the soil being simulated (i.e. no change in soil
s (–) 2.3106 temperature was allowed). Analysis C was a coupled
RK,min (–) 0.1353 THM analysis where the soil was allowed to transfer
Kmin (kPa) 5000 thermal energy. Analyses A, B and C will be referred
to as the base cases for the parametric study (sets D
London Clay adopted by Jurečič et al. (2013) were and E). In analyses D1 and D2 the effect of the soil’s
converted to the IC.G3S model parameters following thermal conductivity is studied by using a value that
the proposal by Measham et al. (2014). The strength is, respectively, ten times smaller or larger than the
properties were also adopted from Jurečič et al. (2013). value provided in Table 1. Similarly, analyses E1 and
The thermal parameters of both soil and concrete are E2 investigate the effect of the soil’s thermal expansion
within the range of the values quoted in the literature coefficient.
(e.g. Banks et al., 2013, Bodas Freitas et al., 2013). The
thermal expansion coefficient of soil was assumed to 4 RESULTS
be the same as that of concrete. The full set of adopted
material properties are listed in Table 1. The sign convention adopted in this paper is such
The domain was restrained in the x-direction along that compressive normal stresses, compressive strains
the axis of symmetry (left mesh boundary) and in the and downward shear stresses applied to the pile are
y-direction along the bottom mesh boundary. Pore positive.
water pressure on the right-side mesh boundary was
assumed not to change from the initial hydrostatic
4.1 Effect of transient thermal behaviour
state. The thermal boundary condition included zero
heat flux along the left mesh boundary as well as Analyses A, B and C were performed in order to illus-
the top of the mesh, which simulated a conservative trate the time-dependent behaviour of a thermo-active
scenario of an insulated boundary. pile and to provide a base case for the parametric study
The analyses were performed in two stages. In the described in the following sections.
first stage, the pile was loaded in an undrained man- Figure 2 shows the evolution of the vertical dis-
ner, reaching the final load of 3300 kN. The second placement of the pile head with time, whereas Figure 3
stage involved a coupled THM analysis simulating six presents the vertical displacement of the pile along its
months of extraction of thermal energy. The pile was length at different stages of the analyses. In all three
gradually cooled over the first ten steps of the analy- analyses, the pile initially moves downwards due to
sis by prescribing a temperature change of −1.5◦ C loading, which, in analyses B and C is followed by
per step to all elements of the pile. This is a sim- contraction of the pile due to cooling. Hence, the pile
plified modelling approach as it simulates a uniform head displaces further down, whereas the toe moves
temperature in the pile and does not take into account upwards. During the six months of consolidation, the
the transient effects of heat convection along the pile settles in all three cases, with the displacement
73
Figure 2. Evolution of pile head vertical displacement with Figure 5. Pile axial stress along its length.
time.
restriction to pile displacement and results in larger
shear stresses due to cooling. The total shear stress
at the end of cooling in the upper part of the pile is
larger than that mobilised during the mechanical load-
ing stage. From Figure 4, it is clear that the soil close
to the ground surface reaches its yield limit with the
depth of failure being greater in Analysis B than in
C, as expected. During the six months of consolida-
tion, the shear stresses caused by cooling reduce more
in Analysis C than B, as in the former the soil cools
down, alleviating the restriction imposed to the pile
movement.
Figure 5 shows the total axial stress in the pile at var-
ious stages. Cooling induces a tensile axial stress and
hence reduces the total compressive axial stress in the
Figure 3. Pile vertical displacement along its length.
pile. In the presented case, this change is favourable
as it increases the factor of safety. However, if the
applied axial load is smaller and/or the cooling load
is larger, the total axial stress in the pile may become
tensile which must be accounted for in the design. The
axial stresses induced by temperature are explained
further in Figure 6. At the end of cooling, the maxi-
mum change in axial stress caused by temperature is
approximately −650 kPa in Analysis B and −610 kPa
in Analysis C. As previously mentioned, this difference
can be explained by the gradual cooling of the sur-
rounding ground which is simulated in the latter case
but is absent from the former. Indeed, during the six
months of consolidation the thermally induced axial
stress reduces to −605 kPa (7% reduction) in Analysis
Figure 4. Pile shear stress along its length. B, and to −345 kPa in Analysis C (43% reduction) as
the mechanical restriction imposed by the soil is allevi-
ated through cooling. Figure 7 depicts the variations of
being the largest in Analysis C, since the soil is allowed the maximum thermally-induced axial stress changes
to cool down, and therefore contract, enabling higher with time for all analyses presented in this paper.
pile settlement. In a thermo-active pile, the effect of temperature on
The shear stresses acting on the pile are pre- the radial stresses must also be investigated. As the pile
sented in Figure 4. The application of the mechanical cools down and contracts, tensile thermally-induced
load causes an upward shear stress, whose magni- radial stresses are generated, as shown in Figure 8.
tude increases with depth due to the mean effective Additionally, the more the soil cools down, the more
stress dependency of the soil stiffness simulated by the it contracts and the thermally-induced radial stress
IC.G3S model. Cooling (i.e. contraction) of the pile increases, which explains the differences between
leads to an additional upward shear stress in the upper analyses B and C. However, it should be noted that
part of the pile and downward shear stress in the lower these tensile stresses are relatively small compared to
part. In Analysis B, the soil is not thermally active, the compressive horizontal stress on the pile at the start
and therefore does not contract, which increases the of cooling.
74
Figure 6. Pile thermally-induced axial stress along its
length.
Figure 9. Thermally-induced pore water pressures (a) at the

end of cooling, (b) after 2 months and (c) after 6 months.
although the volume of soil affected by changes in pore

water pressure increases as the soil continues to cool
down. It should be noted that Analysis C is a coupled
THM analysis and therefore dissipation of pore water
pressures takes place during the initial cooling stage
as well. For comparison, an undrained cooling gener-
ates a maximum pore water pressure of approximately
−125 kPa.
Figure 7. Evolution of maximum thermally-induced axial

stress with time.
4.2 Effect of soil’s thermal conductivity
In the first part of the parametric study, the soil’s ther-
mal conductivity was varied by respectively reducing
(Analysis D1) and increasing (Analysis D2) the value
used inAnalysis C by one order of magnitude, such that
λD1 = 0.1λ and λD2 = 10λ (where λ = 2.0 W/m/K), as
listed in Table 2.
A higher thermal conductivity implies that the heat
transfer through the material is faster. Therefore, for
the same time instant, the average temperature in the
soil is lowest in Analysis D2 and highest in Analy-
sis D1. A larger change in temperature causes larger
thermal strains, which are contractive in the case of
cooling. This is clearly illustrated in Figure 10, which
shows the ground surface settlement increasing with
increasing thermal conductivity. Additionally, Figure
Figure 8. Pile radial stress along its length. 11, which depicts the pile’s vertical displacement, fur-
ther confirms that the rate of settlement is higher
Temperature changes in the soil also affect the pore for greater thermal conductivity due to the increased
water pressure, provided that the thermal expansion reduction in temperature.
coefficients of soil particles and pore water are differ- Figure 12 shows that the change in shear stress due
ent. Equation (1) implies that cooling (i.e. a negative to the initial cooling is the greatest for Analysis D1
temperature change) generates tensile changes in pore and smallest for Analysis D2. As before, this can be
water pressure when αw is larger than αs . Figure 9 explained by the fact that, as the soil cools down more,
shows contours of thermally-induced pore water pres- it contracts more, and therefore the mechanical restric-
sures in Analysis C. At the end of the cooling stage, tion imposed to the pile by the soil reduces, causing a
the largest tensile pore pressures, of approximately smaller change in shear stress. It should be noted that at
−58 kPa, are generated around the toe of the pile. Dur- the end of cooling the total shear stress in the upper part
ing the six months of consolidation these pore water of the pile exceeds that mobilised during the mechan-
pressures around the toe of the pile dissipate slightly, ical loading. Therefore, Analysis D1 is more critical
75
Figure 10. Ground surface vertical displacement. Figure 12. Pile shear stress along its length.
Figure 11. Pile vertical displacement along its length. Figure 13. Ground surface vertical displacement.
in the short term as the failure zone reaches greater

depths.
Figure 7 shows that the maximum thermally-
induced axial stresses at the end of cooling are largest
in Analysis D1, as is also suggested by the shear stress
profile. During the six months of consolidation, the
reduction in the thermally-induced axial stress is great-
est in Analysis D2, where the thermally induced axial
stress becomes positive, increasing the total compres-
sive axial stress in the pile. This result can be explained
by the fact that a larger contraction of the soil in the
case of higher thermal conductivity causes a larger
pile settlement. As, in Analysis D2, this settlement is
greater than the one due to the mechanical load (as Figure 14. Pile vertical displacement along its length.
shown in Figure 11), the total compressive axial stress
in the pile increases. Analysis E2 experiences the largest ground surface, as
well as pile, settlements.
Figure 15 shows that, at the end of cooling, the
4.3 Effect of soil’s thermal expansion coefficient
largest change in shear stress due to the effect of
The second part of the parametric study involves vari- temperature is observed in Analysis E1, where the vol-
ation of the soil’s thermal expansion coefficient such umetric contraction of the pile is greater than that in the
that αE1 = 0.1α and αE2 = 10α (where α = 1.2 × soil, resulting in a larger mechanical restriction being
10−5 m/m/K), while that of the concrete remains applied by the soil. As the soil cools down more with
constant and equal to α. time, the thermally-induced shear stresses reduce in all
The effect of thermal expansion coefficient is three analyses. However, in Analysis E2, the contrac-
clearly demonstrated in Figures 13 and 14, which tion of the soil becomes considerably larger than that
depict the vertical displacement of the ground surface of the pile, resulting in a greater upward shear stress
and the pile, respectively. As the thermal expansion in the lower part of the pile than that mobilised by the
coefficient increases, the soil contracts more, hence mechanical load.
76
Figure 15. Pile shear stress along its length.
These changes in shear stresses cause changes in

axial stresses and, therefore, the greatest thermally-
induced axial stresses at the end of cooling are
observed in Analysis E1, as shown in Figure 7. Dur-
ing the six months of consolidation the thermally-
induced axial stresses decrease. The largest reduction
is recorded in Analysis E2, where the effect of temper-
ature is so significant that the total axial stress in the
pile is more compressive than at the end of mechanical
loading.
Figure 16 shows the thermally-induced pore water
pressures for analyses E1 and E2. As expected from
Equation (1), a larger thermal expansion coeffi-
cient leads to the generation of higher tensile pore
water pressures. The maximum thermally-induced
pore water pressures are −56 kPa in Analysis E1 and Figure 16. Thermally-induced pore water pressures (a) at
−65 kPa in Analysis E2. the end of cooling, (b) after 2 months and (c) after 6 months.
5 CONCLUSIONS greater in the case where the soil was thermally-active

and was allowed to contract as a result of cooling.
This paper describes the thermo-mechanical behaviour The series of parametric studies demonstrate the
of a single pile subjected to a cooling load through a effects of the soil’s thermal conductivity and thermal
series of finite element analyses. expansion coefficient. It was found that a lower ther-
It was shown that pile cooling leads to down- mal conductivity or thermal expansion coefficient may
ward shear stresses in the lower part of the pile, lead to the yielding of the soil around the pile head dur-
whereas in the upper part, it increases the upward ing initial cooling. Moreover, it was shown that, despite
shear stress, which could cause failure in the soil. the initial reduction in the axial stresses due to pile
Additionally, cooling induces tensile axial stresses cooling, in the long term, higher thermal conductivity
in the pile. However, for the magnitudes of thermal or thermal expansion coefficient may actually cause
and mechanical loads considered in this paper, this larger pile settlement than that due to the mechanical
thermally-induced axial stress is not sufficiently large load, increasing the total compressive axial stress in
to overcome the compressive stress due to mechan- the pile, and therefore reducing the factor of safety.
ical loading. Clearly, when designing thermo-active The typical ranges of the soil’s thermal parameters
piles, it is important to assess the significance of quoted in the literature are 1–4 W/m/K (e.g. Banks
the thermally-induced stresses by comparing to the et al., 2013) for the thermal conductivity, and 1–
magnitude of the mechanically-induced stresses. 2 × 10−5 m/m/K (e.g. Bodas Freitas et al., 2013) for the
The effect of the transient phenomena was inves- linear thermal expansion coefficient. The values used
tigated by comparing the results of simulations with in the parametric study were exaggerated in order to
and without heat transfer in the soil. The changes in illustrate extreme pile behaviour. Cleary, the research
stresses due to the initial cooling of the pile are greater presented in this paper highlights the importance of
in the latter case as the soil imposes a larger restric- accurate estimation of soil’s thermal parameters in
tion to the pile displacement. It was also shown that order to allow for an efficient and safe design of
the long term pile settlement and stress changes are thermo-active piles.
77
These studies improve our understanding of the finite element analysis, PhD thesis, Imperial College
behaviour of thermo-active piles. However, further London.
research is still required. The main limitations of the Cui, W., Gawecka, K. A., Taborda, D. M. G., Potts, D. M. &
analysis presented in this paper include the assumption Zdravković, L. (2015) Time-step constraints in transient
coupled finite element analysis. International Journal for
that the pile is uniformly cooled instead of modelling Numerical Methods in Engineering.
the heat transfer between the heat exchanger pipes Di Donna, A., Rotta Loria, A. F. & Laloui, L. (2016) Numer-
and the concrete, as well as the use of mechanical ical study of the response of a group of energy piles
properties that are independent of temperature. More under different combinations of thermo-mechanical loads.
insight is also needed into the effects of cyclic heating Computers and Geotechnics, 72, 126–142.
and cooling. Jardine, R. J., Potts, D. M., Fourie, A. B. & Burland,
J. B. (1986) Studies of the influence of non-linear
stress–strain characteristics in soil–structure interaction.
ACKNOWLEDGEMENTS Géotechnique, 36, 377–396.
Jurečič, N., Zdravković, L. & Jovičić, V. (2013) Predicting
ground movements in London Clay. Proceedings of the
This work is funded by Engineering and Physical Institution of Civil Engineers – Geotechnical Engineering,
Sciences Research Council (EPSRC). 166, 466–482.
Knellwolf, C., Peron, H. & Laloui, L. (2011) Geotechnical
Analysis of Heat Exchanger Piles. Journal of Geotechni-
REFERENCES cal and Geoenvironmental Engineering, 137, 890–902.
Loveridge, F. & Powrie, W. (2013) Pile heat exchangers:
Banks, D. (2012) An Introduction to Thermogeology: Thermal behaviour and interactions. Proceedings of the
Ground Source Heating and Cooling, Chichester, Wiley- Institution of Civil Engineers: Geotechnical Engineering,
Blackwell. 166, 178–196.
Banks, D., Withers, J. G., Cashmore, G. & Dimelow, Measham, P. G., Taborda, D. M. G., Zdravković, L. & Potts,
C. (2013) An overview of the results of 61 in situ D. M. (2014) Numerical simulation of a deep excavation
thermal response tests in the UK. Quarterly Jour- in London Clay. In: Hicks, M. A., Brinkgreve, R. B. J.
nal of Engineering Geology and Hydrogeology, 46, & Rohe, A. (eds.) Numerical Methods in Geotechnical
281–291. Engineering, Delft, The Netherlands, Taylor & Francis,
Bodas Freitas, T. M., Cruz Silva, F. & Bourne-Webb, P. J. 771–776.
(2013) The response of energy foundations under thermo- Pahud, D. & Hubbuch, M. (2007) Measured thermal perfor-
mechanical loading. In: Proceedings of the 18th inter- mances of the energy pile system of the Dock Midfield
national conference on soil mechanics and geotechnical at Zürich Airport. European Geothermal Congress 2007.
engineering, Paris, France, 3347–3350. Unterhaching, Germany.
Bourne-Webb, P. J.,Amatya, B., Soga, K.,Amis, T., Davidson, Potts, D. M. & Zdravković, L. (1999) Finite Element Analysis
C. & Payne, P. (2009) Energy pile test at Lambeth College, in Geotechnical Engineering: Theory, London, Thomas
London: geotechnical and thermodynamic aspects of pile Telford.
response to heat cycles. Geotechnique, 59, 237–248. Stewart, M. A. & Mccartney, J. S. (2013) Centrifuge Mod-
Bourne-Webb, P. J., Freitas, T. M. B. & Assunção, R. eling of Soil-Structure Interaction in Energy Founda-
M. F. (2016) Soil–pile thermal interactions in energy tions. Journal of Geotechnical and Geoenvironmental
foundations. Géotechnique, 66, 167–171. Engineering, 140.
Brandl, H. (2006) Energy foundations and other thermo- Taborda, D. M. G., Potts, D. M. & Zdravković, L. (2016) On
active ground structures. Geotechnique, 56, 81–122. the assessment of energy dissipated through hysteresis in
Cui, W. (2015) Development, implementation and finite element analysis. Computers and Geotechnics, 71,
application of thermo-hydro-mechanical couplings in 180–194.
78
Determining the maximum operating pressure of gas storage caverns

as function of the horizontal in situ geostatic stress
E. Passaris & M. Clemente

Atkins Ltd., UK
ABSTRACT: The analysis of the geomechanical stability of gas storage salt caverns is greatly dependent upon
the magnitudes of the geostatic stresses that exist in the underground formations before the caverns are leached.
In the salt formations an isotropic geostatic stress field is typically assumed and this assumption is customarily
accepted as basis of design in determining the maximum operating pressure in gas storage caverns, ranging
customarily between 80% and 85% of the vertical lithostatic stress at the depth of the last cemented casing
shoe. However, a series of hydraulic/hydrofrac tests carried out in the UK’s Triassic salt fields identified that the
concept of isotropic geostatic stress field was not a valid assumption. The horizontal geostatic stresses, which
essentially provide the lateral confinement to the last cemented casing shoe of the cavern’s well, were found to be
larger than the respective vertical stresses. Geomechanical numerical modelling of an underground gas storage
cavern in Cheshire, was used to investigate whether the maximum operating pressure could be larger than the
applied limit of 80% of the vertical lithostatic stress. The results have shown that for the assessed maximum
pressure all principal stress components around the cavern were compressive, i.e. no development of tensile
stresses. Moreover, it was shown that the investigated cavern was able to endure effectively the developed shear
stresses and was also fully surrounded by a zone of finite thickness where no gas infiltration could occur.
1 THE CONCEPT OF THE ISOTROPIC that characterises the Triassic salt field in Cheshire,
GEOSTATIC STRESS FIELD UK, provided evidence of an anisotropic geostatic
stress field (Passaris & McCusker, 2012). Assessment
In domal Halite formations an isotropic geostatic stress of the derived gradients of the σH and σh compo-
field is typically assumed, i.e. the vertical and the two nents confirmed that the concept of isotropic geostatic
horizontal geostatic stress components are taken to be stress field was not a valid assumption for the inves-
equal implying that σH = σh and ko = 1 where ko is the tigated Triassic salt field, since both horizontal stress
ratio of the horizontal geostatic stress σh to the vertical components were clearly higher than the respective
geostatic stress σv . In the absence of any in situ stress vertical stresses. However, the maximum horizontal
measurements, this assumption may also be extended stress calculated from the breakdown pressure data
to layered salt formations. of the hydraulic fracture tests is not reliable due to
However, in reviewing the implementation of inconsistencies between the requirement of elasticity
hydraulic fracture tests, Horvath & Wille (2009) iden- for the involved stress analysis and the potential creep
tified that fracture tests have been observed to provide behaviour of the Halite (Haimson & Cornet, 2003).
formation pressure values about 5% higher than the As a result, the gradient of the maximum horizontal
vertical stress inferred from density logging. geostatic stress was disregarded and only the mini-
Further evidence of the anisotropic geostatic stress mum horizontal stress results that were consistent with
field in salt formations is provided by Schreiner et al. ko = 1.4 were considered.
(2004) whereby pneumatic fracture tests in a salt dome
of northeast Germany at various depths between 900 m
and 1,400 m indicated that the minimal stress values 2 LIMITING CONDITIONS THAT MUST BE
were significantly higher, i.e. 1.0 MPa–1.5 MPa, than SATISFIED WHEN CONSIDERING THE
the estimated vertical lithostatic pressure. USABILITY OF A GAS STORAGE CAVERN
An accurate assessment of the geostatic stress field
is of great importance and when in situ stress mea- When a cavern is subjected to the maximum permissi-
surements are available, the influence of the identified ble internal pressure, its integrity must be guaranteed
stress field must be taken into consideration in the by ensuring that no macroscopic fracture can be initi-
determination of the maximum permissible cavern ated in the surrounding rock mass. Moreover, consid-
pressure. ering the potential issues of serviceability or usability
The results of a series of hydraulic fracture tests that of the cavern, the surrounding rock mass has to be
were carried out to determine the geostatic stress field tight not only in a structural sense but also in terms of
79
infiltration of small gas volumes into the rock mass.
In this respect, Durup (1994) was first to highlight the
significance of micro-permeation of the stored mate-
rial in a storage salt cavern. Therefore, in evaluating the
maximum admissible pressure of a cavern, it is evident
that the starting phenomenon for the beginning of loss
of its usability may be related to gas infiltration. This
is why, Rokahr et al. (1997) suggested that the deter-
mination of the permissible maximum internal cavern
pressure should be based on a criterion that limits the
infiltration of the storage medium into the rock mass.
They further suggested that a, so called, ‘safety zone’
of sufficient extent (i.e. thickness) must be allowed to
develop around the cavern to provide a barrier against
loss of tightness. In this zone, the effective tangential
∗
stress σθθ should remain below a limiting value σθθ
ensuring that:
Figure 1. Perspective view of the sonar survey of the
investigated gas storage salt cavern.
where σθθ is the is the local tangential stress, corre-

sponding to the smaller stress component tangential cavern, as well as to comply with the development of
to the flow direction and po is the local pore pressure the ‘safety zone’ specified in Section 2.
(negative value). The geomechanical numerical analysis of the gas
To avoid possible local paths of infiltration due to storage salt cavern (shown in Figure 1), which was
the non-uniform shape of a cavern, Riekenberg et al. developed at a depth of approximately 300 m in the
(2004) revised the criterion specified by Equation 1 Triassic Halite formation in Cheshire, UK, was car-
ried out employing an axisymmetric analysis using the
and instead of using the effective tangential stress σθθ ,
they incorporated the effective minor principal stress finite difference package FLAC.
σ3 as: In the numerical modelling the cavern sump, which
has the shape of an inverted cone, was represented as
being available for the storage of gas. In real terms,
the sump will collect insoluble material during the
where σ3∗ is the limiting value of the effective minor solution mining of the cavern and although the insol-
principal stress at the boundary of the safety zone ubles will provide some support to the cavern walls, the
around a cavern that prevents infiltration of the stored sump was modelled as being devoid of such material.
medium. This approximation lends conservatism to the model,
Brouard et al. (2007) focused their approach although the overall structural response of the cavern
on the cavern wall and specified that the effec- will be little influenced by the approach used to repre-
tive minor principal stress must be smaller than the sent the sump. The axisymmetric model of the cavern,
rock tensile strength. They suggested that the related including the assumed shape of the sump is shown in
infiltration/micro-fracturing criterion generally may Figure 2.
be written as:
3.2 Numerical analysis configuration
During the implementation of the finite difference
analysis the modelled structure of the virgin ground
where pmax is the maximum internal cavern pressure
was pre-stressed, in conformance with the in situ
(which is a negative quantity since tension is assumed
geostatic stress field, before an alteration can be intro-
positive) and σt is the rock tensile strength. When this
duced in the model by simulating the leaching of the
criterion is met, an effective tensile zone develops (cor-
cavern. To investigate the influence of the nature of
responding to the infiltration zone), micro-fracturing
the geostatic stresses on the geomechanical stability
occurs, permeability drastically increases and salt
of the cavern, the numerical analysis was carried out
softens.
by pre-stressing the model employing a ko = 1.4.
The leaching of the cavern was modelled by reduc-
ing, over a period of approximately 2.5 years, the
3 GEOMECHANICAL NUMERICAL ANALYSIS
internal pressure of the cavern from the mean geostatic
stress to the halmostatic brine pressure, increased by
3.1 Introduction
an additional 0.5 MPa. Following the modelling of the
Cavern geomechanical stability is defined, in this leaching operations, the de-brining of the cavern was
paper, as the fitness of the cavern to endure any poten- simulated by increasing the internal cavern pressure
tial shear failure which may develop in its walls of the from the halmostatic level to pmax that corresponds
80
Figure 2. Axisymmetric model of the investigated gas
storage salt cavern, incorporating the idealized shape of the
sump.
Figure 3. Distribution of the minimum principal stress (σ3 )
at the end of the 11 years of cyclic loading using pmax equal
to 80% of the σv at the depth of the last cemented to 80% of σv .
casing shoe. After the de-brining process, the cav-
ern was subjected to a cyclic loading by varying the
pressure between pmax and the minimum gas pressure
(pmin ) that corresponds to 30% of the σv at the last
cemented casing shoe. This cyclic pressurization was
modelled over a period of eleven years and then it was
continued for an additional five years by raising the
pmax to a level corresponding to 95% of the σv at the
depth of the last cemented casing shoe.
3.3 Results of the geomechanical numerical

modelling
Examination of the distribution of the principal stress
components that developed throughout the cyclic load-
ing of the cavern indicated the complete absence of
any tensile stresses since all principal stresses, dur-
ing the first eleven years and also during the later five
years (when the pmax was increased) were negative (i.e.
compressive). Moreover, the lowest value of the cal-
culated shear strength factor for the Halite formation,
that surrounds the cavern, was 5.3. Thus, confirming
the ability of the cavern to sustain the shear stresses Figure 4. Distribution of the minimum principal stress (σ3 )
that develop when the cavern is subjected to a cyclic at the end of the 5 years of cyclic loading using pmax equal to
95% of σv .
pressurization which employs the increased pmax .
It remains now to investigate the limiting condi-
tion which specifies that the infiltration zone must to 80% of σv at the depth of the last cemented casing
have an acceptable finite thickness and must be fully shoe.
surrounded by a zone where no infiltration occurs. Furthermore, when we compare the limiting value
The limiting σ3 value that satisfy Equation 3 for of σ3 = −4.55 MPa with the distribution of the minor
pmax equal to 80% of σv is −3.55 MPa, while the principal stress shown in Figure 4, it is possible to see
corresponding σ3 for a pmax equal to 95% of σv is that the thickness of the infiltration zone that develops
4.55 MPa. above the cavern roof is 81 m when the pmax is equal
Comparison of the limiting value of the minor prin- to 95% of σv at the depth of the last cemented casing
cipal stress σ3 = −3.55 MPa with the distribution of shoe.
the minor principal stress shown in Figure 3, indicates Although the infiltration zone above the roof cavern
that the thickness of the infiltration zone that develops is significantly larger when the pmax is increased, there
above the cavern roof is 12 m when the pmax is equal is still a considerable barrier of Halite with a thickness
81
of at least 17 m between the boundary of the infiltra- In conclusion, if a gas storage salt cavern is sur-
tion zone and the rock-head, i.e. the Halite-Mudstone rounded by a formation in which the horizontal geo-
interface (see Figure 4). static stresses are larger than the respective vertical
The identified relatively large infiltration zone that stresses, it may possible to employ a maximum per-
develops above the cavern roof, in contrast with the missible pressure larger than the customarily limit of
infiltration zone that develops along the lateral direc- 80 to 85% of the vertical lithostatic stress.
tion, is essentially a consequence of the development
of the tensile zone that extends above the cavern
roof. REFERENCES
In declaring acceptable an infiltration zone that is
buffered by a 17 m thickness zone of Halite, consid- Brouard B., Bérest, P. & Karimi-Jafari, M. 2007. Onset of
tensile effective stresses in gas storage caverns, Solution
eration was given to the clarifications provided by
Mining Research Institute; Proc. tech. conf., Halifax, 8–10
Brouard et al. (2007) who identified that, although October 2007: 119–136
Equation 3 is valid at the cavern wall, is deemed to Durup, J. G. 1994. Long-term tests for tightness evaluations
be too pessimistic when a point inside the rock mass with brine and gas in salt (Field test no. 2 with gas), SMRI
is considered. In other words, the criterion expressed research and development project report no. 94-0002-
by Equation 3 overestimates the extent of the effective S, Fall meeting. Hannover: Solution Mining Research
tensile zone i.e. the thickness of the infiltration zone. Institute.
Haimson, B.C. & Cornet, F.H. 2003. ISRM suggested meth-
ods for rock stress estimation-Part 3: hydraulic fracturing
4 CONCLUSIONS (HF) and/or hydraulic testing of pre-existing fractures
(HTPF), Int. J. of Rock Mech. & Min. Sciences, 40:
Examination of the distribution of the minor principal 1011–1020.
stresses σ3 , determined in the Halite formation that sur- Horvath, P.L. & Wille, S.E. 2009. Determination of formation
pressures in rock salt with regard to cavern storage. Solu-
rounds the investigated gas storage cavern, indicates tion Mining Research Institute; Proc. tech. conf., Krakow
that σ3 is always compressive even when the pmax is 27–28 April 2009: 83–90.
raised to a level corresponding to 95% of σv , at the Passaris, E. & McCusker, P. 2012. The significance of
depth of the last cemented casing shoe. Consequently, anisotropic geostatic stress field concerning the Hill-
the possibility of developing stress concentrations that top Farm gas storage caverns in Cheshire, UK, Solution
may lead to tensile failure in the roof of the cavern is Mining Research Institute; Proc. tech. conf., Bremen 30
practically unattainable. September–3 October 2012: 210–221.
Calculation of the shear strength factor for the Riekenberg, R., Hartmann, U., Staudtmeister, K. & Zander-
Halite formation has shown that all strength factor val- Schiebenhöfer, D. 2004. Recommendation of maximum
cavern pressures for the gas storage caverns at Huntorf on
ues were higher than 5. The identified high strength the basis of three-dimensional numerical models, Solution
factor values verify the ability of the investigated cav- Mining Research Institute; Proc. tech. conf., Berlin 3–6
ern to accept effectively the resulting shear stresses October 2004.
when subjected to a cyclic pressurisation, even when Rokahr, R.B., Staudtmeister, K. & Zander-Schiebenhöfer, D.
the pmax is raised to a level corresponding to 95% of σv . 1997. Development of a new criterion for the determi-
The limiting condition concerning the infiltration nation of the maximum permissible internal pressure for
zone (that may develop if the criterion specified by gas storage in caverns in rock salt, SMRI research and
Equation 3 is met), is adequately satisfied when the development project report no. 97-0001, Spring meeting.
Houston: Solution Mining Research Institute.
maximum permissible cavern pressure is raised to a
Schreiner, W., Jäpel, G. & Popp, T. 2004. Pneumatic frac-
level corresponding to 95% of σv , at the depth of the ture tests and numerical modeling for evaluation of the
last cemented casing shoe. At this pmax , the infiltration maximum gas pressure capacity and the effective stress
zone is characterized by an acceptable finite thickness conditions in the leaching horizon of storage caverns in
and is fully surrounded by a zone where no infiltration salt diapirs, Solution Mining Research Institute; Proc.
occurs. tech. conf., Berlin 3–6 October 2004: 129–142.
82
Thermo-mechanical behaviour of energy pile in underground

railway construction site
M. Adinolfi, A. Mauro, R.M.S. Maiorano, N. Massarotti & S. Aversa

Università Degli Studi di Napoli “Parthenope”, Napoli, Italy
ABSTRACT: The use of geothermal pile foundations is an environmental friendly way to extract/storage energy
from/in the ground. The operating principle of these innovative structures is based on energy transfer between
the ground and the fluid flowing inside the probes inserted in the pile, in order to feed heat pumps dedicated
to buildings heating and cooling. Nevertheless, very limited installations are recorded in Italy, and only few
information are available regarding the impact of thermal processes on the structural performance of energy
piles. In the present work, the geo-energy research group has analysed the thermo-hydro-mechanical behaviour
of an energy pile installed in pyroclastic soils and rocks, as an element of a sheet pile wall. The developed model,
solved numerically by using finite elements, is then used to forecast the behaviour of an actual pile subjected to
thermo-mechanical loads and to design a proper experimental campaign in the underground railway construction
site of Piazza Municipio in Napoli, Italy.
The multi-pile experimental set-up is equipped with fibre optic sensors system that allows the evalua-
tion of strains and temperature distributions along the piles during cooling and heating processes. Once the
measurements will be available, the developed model will be verified and validated against on field experiments.
1 INTRODUCTION element THM model in poroelastic media allowing the

simulation of complex soil-structure interaction phe-
In recent times, the diffusion of geothermal pile foun- nomena for thermo-active piles, in order to evaluate the
dations to extract/store up energy from/in the ground performance of the geothermal pile in the investigated
is rapidly growing (Brandl 2006, Amis et al. 2014), area.
also for its economic advantages due to the reduction
of installation costs with respect to traditional geother- 2 EXPERIMENTAL SET UP
mal probes. The operating principle of these innovative
structures is based on energy transfer between the 2.1 The site
ground and the fluid flowing inside probes inserted
in the pile, in order to feed heat pumps dedicated to The experimental set-up has been realized in the con-
buildings heating and cooling (Brandl 2006). Experi- struction site of the underground station of Piazza
mental tests have been conducted (Laloui et al. 2006, Municipio in Napoli, Italy. The train platform of
Bourne-Webb et al. 2009) on energy piles, showing the original project is located at 13.5 m below the
that the thermally induced axial stress inside the pile ground level. A sheet pile wall is realized by means
strictly depends on the end-restraint provided by the of forty-five bored piles (800 mm nominal diameter),
overlying structure, thermal loads and the properties at a inter-axis distance of 1.10 m, and six of them
of the foundation material (Amatya et al. 2012). Many have been used for the present set-up, by incorpo-
numerical approaches are available in the literature. rating, in the concrete casting, the geothermal probes
Related to heat and fluid flow through porous media made of High Density PolyEthylene (HDPE). The total
is presented a novel method in Carotenuto et al. 2012 length of piles to support the excavation of 6 m is
and a comprehensive review of the developed models equal to 11.40 m. The average normal load on each
for thermo-fluid dynamic phenomena in low enthalpy pile is smaller than 1200 kN. The tested elements are
geothermal energy systems is proposed in Carotenuto located within the sheet-pile retaining wall. The HDPE
et al. 2016. Concerning the coupled phenomena ther- probes (32 mm nominal diameter) were fixed to the
mal, hydraulic and mechanical (THM) several models pile reinforcing cage (Fig. 1). Different probes con-
are available about effects on single energy pile and figurations have been provided in order to check their
the surrounding soil (Laloui et al. 2006, Salciarini performances: spiral, double and triple U-shape.
et al. 2012, Suryatriyastuti et al. 2012) and piles group
2.2 The instrumentation
(Dupray et al. 2014, Jeong et al. 2014, Salciarini
et al. 2015). The aims of this work are to describe The reinforcement cage (details available in Figure 1)
the experimental set-up and to propose a proper finite has been used also to support Optic Fiber Sensors
83
thermal conductivity and permeability tests on soil
samples, integrity tests on the concrete, performance
evaluation on heat exchangers and ground response
tests.
3 THERMO-HYDRO-MECHANICAL MODEL
3.1 Mathematical formulation

A coupled thermo-hydro-mechanical model has been
developed to simulate the behaviour of one of the
piles of the experimental set-up during the prelimi-
nary tests, to be carried out before the soil beneath
the sheet pile wall will be excavated. The governing
equations reproducing the mechanical behaviour of the
solid skeleton, the heat conduction phenomena and the
hydraulic motion in the soils are solved. The model is
based on the following assumptions: soils are modelled
as isotropic and linear elastic materials (the applied
mechanical loads are significantly smaller than the
collapse values); displacements and deformations of
the solid skeleton are small (linear kinematics); perfect
Figure 1. Experimental set up details: typical section of contact between soil and pile is considered; hydro-
equipped pile; pictures of set-up distribution of OFS cables static pressures are neglected. Taking into account the
and spiral probes before and after concrete casting; pile previous assumptions, the applied equations are:
section with geometric and ground features.
(OFS), for strain and temperature measurements. In

particular, about 180 m of OFS cables were installed
in three of the six probes-equipped piles. Two methods
of installation, one with the cable anchored continu-
ously along the length of the reinforcement cage at
regular steps and one with the cable strung between
the top and bottom of the cage were used. As concerns
strain measurements, three fibres, installed at 120◦
along the pile circumference, allow to derive the vec-
torial displacements of each pile section. The fourth where σ = total stress tensor (positive if tensile);
optical cable was strung along the pile for distributed b = volume external force vector; p = pore water pres-
temperature measurements and a traditional tempera- sure increase (positive if compressive); αB = Biot
ture sensor was used to obtain a reference temperature Willis coefficient of porous media; De = elastic stiff-
value. The monitoring system is made up by an opto- ness tensor; ε = strain tensor; εT = thermal strain
electronic reading unit, including all the electronic and tensor; T = temperature; 1 = second order unit ten-
optical components and cables. sor; αs = volumetric thermal expansion coefficient
of the soil; αw = water thermal expansion coeffi-
2.3 Soil conditions cient; v = Darcy velocity vector; ρ = soil density;
k = permeability tensor; evol = volumetric strain; βsw
The sheet pile wall is located inside a larger excava-
is thermal volumetric expansion coefficient of the bi-
tion with its top at a depth of 8.00 m below ground
phase mass; and S = storage coefficient taking into
water level (gwl). Ground conditions, are typical
account for porosity, permeability, Biot Willis coeffi-
for Neapolitan area: a superficial layer of granular
cient of soils. Heat transfer in porous media, includ-
pyroclastic products of volcanic activity called “poz-
ing conduction and convection phenomena, uses the
zolanas” (3.55 m thick) overlies a lithified stratum of
following equation:
yellow tuff, at depth of 11.55 m below gwl, which
extends for many meters under the piles toe. Since
the water level is well above the top of the sheet pile
wall, all the soils are considered as saturated (Fig. 1).
The energy piles installation was undertaken within
an active construction site. Therefore, the experimen-
tal campaign will be designed compatibly with all where n = porosity of soils; ρw = water density;
the other activities of the underground station. More- λeff = thermal conductivity tensor of the soil;
over several additional in situ and laboratory tests will Cps = solid skeleton thermal capacity and Cpw = water
be realized to characterize the materials, including thermal capacity.
84
Table 1. Thermal, hydraulic and mechanical properties of
soils.
Tuff
Density (ρ) 17 [kN/m3 ]

Young modulus (E) 1500 [MPa]
Poisson modulus (ν) 0.3 [–]
Thermal expansion (αs ) 4 × 10−5 [K−1 ]
Thermal conductivity (λeff ) 1.48 [W m−1 K−1 ]
Thermal capacity (Cp) 850 [J kg−1 K−1 ]
Permeability (k) 4 × 10−5 [m/s]
Porosity (n) 0.550 [–]
Pozzolanas

Young modulus (E) 520 [MPa]
Thermal expansion (αs ) 4 × 10−5 [K−1 ]
Thermal conductivity (λeff ) 2.30 [W m−1 K−1 ]
Permeability (k) 1 × 10−6 [m/s]
Porosity (n) 0.525 [–]
Concrete

Young modulus (E) 40 [GPa]
Thermal expansion (αs ) 8.5 × 10−6 [K−1 ]
Thermal conductivity (λ) 2.10 [W m−1 K−1 ]
Mottana & Campolunghi 2010; Bourne-Webb et al.

2009; Laloui et al. 2006) are reported in Table 1.
Figure 2. Axial-symmetric model region. 3.3 Loads time histories

The loading time history lasts 4 months operation and
it includes two phases: the application of the mechan-
3.2 FEM analysis
ical load at the pile head and the application of a
The time-dependent numerical model has been imple- seasonally cyclic thermal load under constant mechan-
mented and solved by using the finite elements code ical load. The applied mechanical load is 1200 kN. The
Comsol Multiphysics version 4.3. The 2D model thermal load is due to the temperature of the injected
region, in the axial symmetric system, has a width of fluid, supposed constant along the probes, accord-
4.00 m and a length of 20.00 m (Fig. 2). ing to the law presented in Figure 3a. The authors
The mechanical Boundary Conditions (BCs) are: are aware that this is an approximation, but for the
fixed constraints at the bottom and roller condition on present scope is retained acceptable. During the first
the side. The thermal BCs are: constant temperature on month only mechanical load is applied and tempera-
the heat exchanger, adiabatic conditions on all external ture is equal to 16.9◦ C (ground reference temperature
boundaries.The hydraulic BCs are: no flow on the right in Naples). During the cooling phase (winter mode),
and bottom boundaries and zero pressure on the top. the inflow temperature in probes is set to the value
Initial stress conditions are derived from the results of 7◦ C (T = 9.9◦ C) maintained for 30 days, then
obtained from steady-state simulations accounting for the temperature in the probes is kept constant and
only gravity effects. equal to the initial ground reference value during the
The model has been validated against experimen- third month. The increase from reference temperature
tal data available in the literature (Bourne-Webb et al. to 35◦ C (T = 18.1◦ C) occurs during the last month
2009). The geometric features are shown in Fig- (summer mode). The effects of only mechanical load
ures 1–2, and the material properties, taken from (M) are appreciated at time t = 720 h whereas both
the literature (Viggiani & Rippa 1998; Aversa et al. effects of mechanical-cooling (C+M) thermal loads
2013; Aversa & Evangelista 1993; Colombo 2010; are present at time t = 1440 h and mechanical-heating
85
Figure 3. Time histories of applied thermal and mechanical
loads (a); pile head vertical displacements (b). Figure 4. Pile and soil vertical displacements after mechan-
ical and thermal loads at pile head.
(H+M) thermal loads at time t = 2880 h (dotted lines
in Figure 3a).
4 RESULTS
4.1 Pile head and soil displacements

Pile head displacements (Fig. 3b), produced by this
thermo-mechanical loading cycle, are smaller than
1.4 mm and the maximum settlement is reached when
both mechanical and cooling loads are applied. The
analysis shows that the vertical displacement of the
pile head follows the temperature changes.
Pile and surrounding soil displacements at the
ground surface are plotted versus radial distance in the
left side of Figure 4. The pile is subjected to the nor- Figure 5. Axial load profiles and shear stress distribution
mal stresses of the soil at the toe, and to normal and along soil-pile interface after mechanical and thermal loads.
tangential stresses on the shaft. The pile head under
only mechanical load (t = 720 h) exhibits a maximum
settlement of 0.48 mm and the effects of the axial load mechanical load is applied. It results less compressed
influence the surrounding soil with a compressive pro- especially between 7.50 m and the pile’s tip at the end
file. When the thermal loads are added, the pile and the of the subsequent cooling phase. An increase of the
surrounding soil are more contracted during cooling compressive load can be appreciated during heating
phase according to their different thermal expansion phase with a consequent maximum load of 1450 kN
coefficients and different temperatures, involving a due to the combined thermo-mechanical effects.
settlement of 1.4 mm for head pile and 0.8 mm for sur- The distribution of shear stress along the interface
rounding soil. During the heating phase, the expansion is shown on the right side of Figure 5. The path follows
of the pile leads to an uplift of 1.00 mm of its head, a trend opposite to that of axial load in which at 3.55 m
whereas the surrounding soil shows a decreasing path depth the change from pozzolanas to tuff soils in under-
of upwards displacements versus radial distance. lined by a cusp, which is directly correlated to soil
The vertical displacements in the pile, along with strength and stiffness of layers. During the mechanical
depth, are plotted on the right side of Figure 4. Mechan- load the shear stresses act upward all over the length
ical loads imply downward displacements (t = 720 h) of the pile. In winter mode (C+M), due to different
maximum at pile head and minimum at the toe. At contractions between pile and soil, frictions mobilized
time t = 1440 h, the top of the pile tends to contract along the upper part of pile increase whereas along
downward upon temperature decrease more than the the lower part of pile decrease. Inversely in summer
toe. Contrariwise at time t = 2880 h, the pile tends to mode (H+M), in the upper part of the pile the shear
expand: the pile head moves upwards upon temper- stresses act downward, while the lower part of pile
ature increase, while the lower segment of the pile has additional frictions mobilized. Moreover, it can be
expands less reaching a null displacement at the toe. verified that when thermal loads are applied a neu-
tral point appears, located at 6.70 m during heating
process. It represents the point where the opposing
4.2 Axial loads
soil shear stresses along the pile are balanced.
Comparing the different plots of left side of Figure 5, it Consequently, the compressional stress due to pile
is possible to see, in terms of axial load, that the pile is elongation and the associated soil resistance is max-
obviously fully contracted at time t = 720 h when only imized at the neutral point where themo-mechanical
86
Figure 7. Pore water pressure field (in kPa) steps: thermal
Figure 6. Temperature field (in ◦ C) after the first month loading (cooling phase) application at t = 720 h (a), the end of
of analysis at t = 720 h (a), after cooling phase at t = 1440 h cooling phase at t = 1440 h (b) and the end of heating phase
(b) and at the end of heating phase at t = 2880 h (c). at t = 2880 h (c).
shear stresses acting on the pile reverse from down-

ward to upward along the pile length. The thermo- to flow away from the area, if the soil is sufficiently per-
mechanical stress distribution would be symmetrical meable that water flow is possible in a short time scale.
if the pile was free at both ends for an elastic soil with At time t = 2880 h (Fig. 7c) excess of pore water pres-
uniform stiffness throughout (Ozudogru et al. 2015). sure is not yet completely dissipated whereas residual
pressure at this time is about 60 kPa.
4.3 Thermal field
The temperature field produced by the heat exchanger 5 CONCLUSIONS
probes is shown in Figure 6 together with deformed
axial-symmetric domain.The figure presents three sig- Energy piles are foundations which support the loads
nificant temperature steps; the first one shows ground- of structure and provide thermal energy to the build-
pile system constant temperature at 16.9◦ C (Fig. 6a). ing as heat exchangers. They use the ground as energy
During the cooling phase (Fig. 6b), the temperature source sink, successfully considered as an innova-
ranges from 7◦ C inside the heat exchanger to 16.9◦ C tive environmentally friendly building’s technology.
in the ground. At the end of heating phase (Fig. 6c), In general, the amount of thermal energy exchanged
the temperature ranges from 35◦ C inside probes to depends on the type of soil, saturation degree, pres-
reaching ground initial temperature at boundaries of ence of groundwater flow, temperature of the ground
domain. and on the building energy needs. It is well known
that additional thermal deformation and thermal stress
4.4 Pore water pressure field that change the behaviour of material appear during
cooling/heating processes.
Temperature changes around the pile induce changes This study is finalized to evaluate geotechnical
through their effect on water and soil skeleton. In Fig- effects and energy-related problems that can result
ure 7a is plotted pressure field distribution at time from the use of energy pile under conventional tem-
t = 720 h. peratures in the tested site of the underground station
Due to the cooling loading stage a reduction about of Piazza Municipio in Napoli, Italy. A preliminary
−80 kPa of pore water pressure concentration (and numerical simulation is carried out with finite ele-
increases the effective stress of the soil) is computed ment method under mechanical and thermal loads for
in the lower part of the pile whereas compressive pres- cooling and heating operation on a single pile in the
sures of 20 kPa is computed for the length pile in tuff, construction site. The numerical results were in qual-
coherently with the differences in the soil hydraulic itative agreement with the experimental results of in
properties. At the end of cooling phase (t = 1440 h) situ tests at Lausanne and Lambeth College reported
the suction is dissipated and the residual induced pore in literature. For the Piazza Municipio energy pile, the
water pressure variation is lower than 40 kPa at lower results show that:
boundary of the domain (Fig. 7b). A reverse effect is
produced when an increase in temperature expires an • the pile head settlements are lower than 1.40 mm
excess pore water pressure concentration at pile toe and its maximum elongation is of 1.00 mm. Due to
and tensile pressure of in the upper part of tuff layer. low magnitude of these movements, there is slight
This increase in pore pressure can tend to cause water influence in mechanical behaviour of the structure;
87
• when thermal loads are applied, the neutral point Amis, T. & Loveridge, F. 2014. Energy piles and other ther-
where the opposing soil shear stresses along the pile mal foundations for GSHP, The REHVA European HVAC
are balanced appears: its position strictly depends Journal 51(1).
from pile constrains; Aversa, S. & Evangelista, A. 1993. Thermal Expansion of
Neapolitan Yellow Tuff. Rock Mech. Rock Engng. 26(4):
• cooling energy pile causes the thermal contrac-
281–306.
tion of the pile, resulting in decreasing axial Aversa, S., Evangelista, A., Scotto di Santolo, A. 2013.
load along the pile. Inversely, heating energy pile Influence of the subsoil on the urban development of
causes an increase in axial load under combined Napoli. In Carlo Viggiani (ed.), Geotechnical engineer-
thermo-mechanical loading reaching the maximum ing for the preservation of monuments and historical sites;
compressive value of 1450 kN. Typical problems for Proc. intern. symp., Napoli, 30–31 May 2013. Rotterdam:
energy piles are: stresses may exceed the acceptable Balkema.
design stress, while large-amplitude strain cycles Bourne-Webb, P.J., Amatya, B., Soga, K., Amis, T., Davidson,
alter the magnitude and distribution of shaft fric- C., Payane, P. 2009. Energy pile test at Lambeth College,
London: Geotechnical and thermodynamic aspects of pile
tion mobilised between the pile and the soil (Laloui response to heat cycles. Géotechnique, 59: 237–248.
et al. 2006, Bourne-Webb et al. 2009). Obtained Brandl, H. 2006. Energy foundations and other thermo-active
results, using a four-month time scale and conven- ground structures. Géotechnique 56: 81–122.
tional temperatures (7◦ C and 35◦ C), are far from Carotenuto, A., Massarotti, N., Mauro A. 2012. A new
collapse values; methodology for numerical simulation of geothermal
• temperature distribution extends throughout sur- down-hole heat exchangers. Applied Thermal Engineer-
rounding soil for a radial distance of 3d (2.40 m) ing, 48: 225–236.
and for 4.50 m above pile’s toe; Carotenuto, A., Ciccolella, M., Massarotti, N., Mauro
• the effect of the thermal behaviour of water in low A. 2016. Models for thermo-fluid dynamic phenom-
ena in low enthalpy geothermal energy systems: A
permeability soil creates changes in the effective review. Renewable and Sustainable Energy Reviews, 60:
stresses of the soil. The proposed analysis shows 330–355.
that heating phase does not reduce significantly the Colombo, G. 2010. Il congelamento artificiale del ter-
effective stresses at pile toe but the knowledge of reno negli scavi della metropolitana di Napoli: valu-
the soil permeability is necessary for the design of tazioni teoriche e risultati sperimentali. Rivista Italiana
an energy pile foundation. Indeed huge reduction in di Geotecnica 4.
effective stress could be detrimental to the stability Comsol. 2012. Comsol Multiphysics user manual version 4.3
of the building, due to a reduction in mobilized shaft Palo Alto CA.
friction along the piles. Dupray, F., Laloui, L., Kazangba, A. 2014. Numerical analy-
sis of seasonal heat storage in an energy pile foundation.
Future developments will focus on the evaluation of Computers and Geotechnics 55: 67–77.
Jeong, S., Lim, H., Lee, K., Kim, J. 2014. Thermally induced
the effects of variation of boundary conditions due to
mechanical response of energy piles in axially loaded pile
open excavations, required to accommodate the under- groups. Applied Thermal Engineering 71: 608–615.
ground stations, and on model the whole retaining wall Laloui, L., Nuth, M., Vulliet, L. 2006. Experimental and
activating the group of equipped six energy piles. Once numerical investigations of the behaviour of a heat
the measurements from the in situ campaign will be exchanger pile. Int. J. Numer. Anal. Meth. Geomech. 30:
available, the model will be verified against on field 763–781.
experiments. Ozudogru, T.Y., Olgun, C.G., Arson, C.F. 2015. Analy-
sis of friction induced thermo-mechanical stresses on a
heat exchanger pile in isothermal soil. Geotechnical and
Geological Engineering 33: 357–371.
Mottana, A., Campolunghi, M.P. 2010. Strutturazione di una
ACKNOWLEDGMENTS banca dati in ambiente G.I.S. per lo sviluppo di impianti
innovativi finalizzati alla gestione delle georisorse. Report
The authors gratefully acknowledge the financial sup- di ricerca si sistema elettrico. Accordo di programma
Ministero dello Sviluppo Economico – ENEA.
port of GeoGrid project PON03PE_00171_1. Further- Salciarini, D., Tamagnini, C., Cinfrignini, E. 2012. Model-
more the authors thank: Eng. F. Cavuoto as of Head lazione dei processi termo-idro-meccanici indotti in
Director of underground railway construction site of prossimità di pali geotermici. IARG,Padova, 2–4 Luglio
Piazza Municipio, Metropolitana di Napoli s.p.a. and 2012.
SudMetro s.c.a.r.l.; Eng. G. Normino and Eng. P. Salciarini, D., Ronchi, F., Cattoni, E., Tamagnini, C. 2015.
Marotta from CRAVEB (Consorzio di Ricerca per Thermomechanical effects induced by energy piles oper-
l’Ambiente i Veicoli l’Energia e i Biocombustibili) ation in a small piled raft. Int. J. Geomech. 15(2).
for the technical support in the design and realization Suryatriyastuti, M.E., Mroueh, H., Burlon, S. 2012. Under-
stages of the experimental set-up. standing the temperature-induced mechanical behaviour
of energy pile foundations. Renewable and sustainable
energy reviews, 16: 3344–3354.
REFERENCES Viggiani, C., Rippa, F., 1998. Linea 1 della metropolitana
di Napoli tratta Dante-Garibaldi. Relazione Geologia e
Amatya, B., Soga, K., Bourne-Webb, P.J., Amis, T., Laloui, Geotecnica.
L. 2012. Thermo-mechanical behaviour of energy piles.
Géotechnique, 62(6): 503–519.
88
Innovative deep-mixing methods for oil & gas applications
G. Spagnoli
Department of Maritime Technologies, BAUER Maschinen GmbH, Schrobenhausen, Germany
D. Bellato
Department of International Projects and Services, BAUER Spezialtiefbau GmbH, Schrobenhausen, Germany
P. Doherty & G. Murphy

GDG Geosolutions, Dublin, Ireland
ABSTRACT: Notwithstanding the increasing investments in the use of renewable sources of energy and natural
gas observed in recent years, oil still represents an important part of the global energy market. In particular,
oil overall consumption and production are expected to continue growing, especially in emerging economies,
though more slowly than in the past due to international policies aimed at slowing down climate changes. Plants
and infrastructures related to the oil industry are often inherently related to challenging projects, which, in most
of the cases, have to fulfill strict technical, environmental, and economical specifications. For instance, refinery
sites located in sub-urban contexts have to face stringent environmental requirements both during production
and after their dismantling to ensure that no pollution spreads across the surrounding areas. On the other hand,
the construction of offshore platforms in difficult geotechnical conditions presents issues, which can be solved
only by means of innovative technical solutions. This paper presents two case histories in which deep-mixing
techniques were used to produce improved-soil elements for environmental and structural purposes related to the
oil industry, i.e. a cut-off wall in an ex-refinery site and foundation piles for the support of offshore platforms in
carbonate sands. Deep-mixing (DM) of soil is a well-established methodology in geotechnics introduced more
than 50 years ago in Japan. DM methods have been used so far in numerous applications both on-shore and
offshore projects. The use of DM is expected to become more and more popular in the future due to its economic
and environmental advantages compared to other traditional construction techniques.
1 INTRODUCTION to 6 m (for offshore renewable energy) and lengths up

to 300 m (Gerwick 2007). However, there are situa-
Fixed platforms have been traditionally used to exploit tions where the pile installation by means of hammers
offshore hydrocarbon reserves in water depths of less is not possible, for instance in rock conditions, where
than 200 m, although there are several examples of boulders and/or cobbles are encountered or in calcare-
fixed jackets installed in deeper water, e.g. Magnus ous deposits. Therefore, drilled piles are normally used
platform installed in 186 m water depth in the North in these conditions. George and Wood (1976) identify
Sea (Clarke 1993), the Bullwinkle oil platform in the three basic types of offshore drilled piles:
Gulf of Mexico installed in 412 m water depth (Digre
• Single-stage drilled piles, which can be formed by
et al. 1989). Offshore fixed production platforms are
drilling an oversized hole to the required penetra-
globally installed. In the period 2008–2012, the major-
tion, inserting a steel tube pipe and grouting the
ity of the fixed platforms where installed in Asia and
annulus between pile and soil;
Middle East. As a matter of fact, Asia’s demand for
• Primary driven pile, which forms the casing for the
fixed platforms is increasing especially for develop-
upper section of the insert pile. An oversized hole
ments in Malaysia and China, with large proportion of
is then drilled to the required penetration below the
platforms situated between 25 and 50 m water depth.
primary pile tip, and the annulus between the insert
Platforms installed in depths over 100 m are common
pile and soil and the primary pile is grouted;
in South America and Africa. Regarding West Africa,
• Belled pile.
Angola and Nigeria are the key countries driving the
market growth (Infield 2013). It is possible to observe According to Poulos (1988), Gerwick (2007) and
from Fig. 1 that the majority of fixed production plat- Doherty et al. (2016) offshore drilled piles are installed
forms are pile-based. In fact, the most common type of with top drilling units with facilities for both direct and
offshore foundation is the open-end driven steel tube reverse circulation, with the latter normally used for
pile (Poulos 1988) with diameter ranging from 1m up relatively large-diameter holes. Offshore piles resist
91
shaft friction than for the driven counterparts. This has
resulted in the prominence of D&G piles in calcareous
sand deposits (e.g. King et al. 1980; Gerwick 2007).
However, D&G piles is a costly foundation solution
and therefore, despite the geotechnical properties of
carbonate sands, driven piles with closed-ended piles
have been installed in the past for saving costs (e.g.
De Mello et al. 1989) or grouted driven piles have
been suggested, where cement grout after driving is
injected (Barthelemy et al. 1987). Barthelemy et al.
(1987) tested a 762 mm diameter pile equipped with
grout pipes and instrumented with 36 strain gauges
was driven into calcareous sands down to 24 m. The
following paper presents and discuss the latest data
about a novel mixed-in-place pile (MIDOS) already
described in Igoe et al. (2014); Spagnoli et al. (2014);
Doherty et al. (2016). As the MIDOS is based on the
mixed-in-place technology, an onshore test with the
Cutter Soil Mixing (CSM) in an ex-refinery site is pre-
sented. Laboratory tests of MIDOS pile in carbonate
Figure 1. Fixed platforms installed in the period 2008–2012 sands will also be briefly discussed.
(modified after Infield 2013).
compression and tension load, by transferring it along 2 THE DEEP MIXING METHOD
its shaft and by end bearing on its tip (for compression
loads). However, calcareous soil deposits posed sev- 2.1 Introduction
eral problems regarding the pile capacity (e.g. Murff
The Deep Mixing Method (DMM) is an in-situ soil
1987; Le Tirant and Nauroy, 1994; Doherty et al.
treatment technology whereby binding materials are
2015). Calcareous (or carbonate) sands are found
added and blended with soils in order to improve
manly in the warm seas between latitudes 30◦ N and
their hydraulic and mechanical properties. Deep Mix-
30◦ S in coastal areas of Australia, India, Saudi Arabia.
ing techniques, originally developed in Sweden and
However they are also frequent up to latitudes 50◦ N
Japan during the 1960s and 1970s, are today well-
and 50◦ S (Le Tirant and Nauroy 1994). It is very well-
established procedures in the geotechnical engineering
known that they show comparable friction angles to
practice of an increasing number of countries. The rea-
silica sands at low confining stress, but at higher pres-
sons for this success can mainly be due to the several
sures, the material contracts, due to crushing of the
engineering purposes they serve as alternative, more
calcareous grains.
economic (e.g. Topolnicki 2004), and environmental
The tendency to contraction during shearing has
friendly solutions with respect to the traditional meth-
therefore serious implication for pile shaft resistance,
ods involved in ground improvement works and to
which depends on the development of lateral stress
the constant technological development carried out
along the pile wall (e.g. Murff 1987; Colliat et al. 1999;
on mixing rigs. The performance of deep mixing
Gerwick 2007; Spagnoli et al. 2015). The skin friction
structures depends significantly on the mixing pro-
of driven piles in carbonate sands (where the estima-
cess implemented at the site, which is much more
tion of K and tan δ is uncertain) is given by (Le Tirant
effective when a homogeneous distribution and uni-
and Nauroy 1994):
form blending of the binding material with the soil is
achieved (Mitchell 1981). First attempts to classify in
a rigorous way deep mixing techniques were made by
Bruce (2000). This classification depends on several
where β varies from 0.05 to 0.2 (Dutt and Cheng 1984;
fundamental operative features as follows:
Abbs et al. 1988). According to Randolph (1988) for
long piles, the very low β values suggest that it is not • The method with which the binding material is intro-
correct to relate the value of skin friction to the over- duced into the subsoil, namely in a Wet (pumped as
burden pressure. It seems more appropriate to think in a slurry) or a Dry (blown in pneumatically) form.
terms of absolute values of unit skin friction, which lie • The approach adopted to penetrate the ground
in the range of 5 to 15 kPa. Drilled-and-grouted (D&G) and/or to blend and homogenize the chemical agents
piles overcome the difficulties associated with parti- with the soil: purely by Mechanical methods adding
cle crushing and compression at the pile interface and the binder at relatively low pressures, or by a rotary
are frequently used for offshore foundations in these method aided by Jet systems which injects fluid
deposits (Lee and Poulos 1991). As a result, the radial slurry at high pressure. The classification does not
effective stress will remain close to the in situ hor- include jet-grouting, as it does not involve any
izontal stress, yielding significantly higher values of mechanical mixing to improve the stabilized mass.
92
Figure 3. CSM technique: (A) cement slurry flowing out the
nozzle located between the cutting wheels; (B) CSM mixing
Figure 2. General classification of deep mixing methods wheels and shear plates.
(modified after Topolnicki 2004).
The CSM unit can be mounted on a guided Kelly

• The vertical position of the mixing tools along the bar or a wire rope suspended cutter frame equipped
drill shaft: the incorporation of the binder and its with special steering devices. Depths up to 60m have
homogenization with the soil can be carried out at been successfully reached using wire rope-suspended
the End (or within a distance comparable with the rigs provided with four sets of mixing wheels. Pan-
characteristic size/diameter of the tool) or along a els having a width ranging from 2.4 m to 2.8 m and a
significant portion of the Shaft. thickness between 0.55 and 1.5 m can be created.
In practice, not all the combinations provided by this
flowchart are available for ground improvement appli- 2.3 The MIDOS technology
cations since wet slurry, jetted shaft mixing (WJS) and The MIDOS pile is a hollow steel pile, which is lowered
dry binder, rotary, shaft mixing (DRS) do not exist, and into the soil while at the same time a mixing tool, which
no jetting introducing dry binder (DJS or DJE) have is advanced inside the hollow pipe, mixes the in situ
been already developed. Furthermore, recent tech- soil with cement slurry (Fig. 4). Hence, a mixed soil
nologies based on driven mechanisms different from cement body is created.
those involving rotary systems cannot be correctly A test pile has been installed in the Bauer’s field test
identified using this scheme. facility in Bavaria (Germany) in a silica sand deposit
In order to overcome such limitations, Topolnicki with the following dimensions: the steel pile had an
(2004) elaborated a more general classification which external diameter of 1.5 m with a wall thickness of
allows distinguishing between different in-situ soil 15 mm. The steel has at its lower end a mixing chamber
mixing treatments (Fig. 2). The basic indicators used with an outer diameter of 1.9 m. Holes in the steel pile
to operate this distinction are the same considered and in the mixing chamber allow the soil-cement-mix
by Bruce (2000), even if the mixing principles are to communicate between inside and outside of the pile.
here extended to include recent advancements in deep The steel pile with the mixing chamber remains in the
mixing machinery. soil, surrounded by a soil-cement-mixture. The pile
length was 17 m and it was installed in less than 3
hours (see Igoe et al. 2014).
2.2 Cutter Soil Mixing (CSM) Because the MIDOS reached a static uplift capacity
of about 9 MN (Igoe et al. 2014) and a static bear-
The Cutter Soil Mixing (CSM) technique is a wet ing capacity mobilized by the dynamic pressure of
mechanical mixing technique developed in 2003 that the load test of 15.4 MN, it was decided to extend the
differs significantly from conventional DMMs, since it investigation to calcareous sands, which as previously
makes use of two sets of cutting wheels rotating around described, present foundation issues.
horizontal axis (Fig. 3A) to produce rectangular pan-
els of improved soil rather than one or more vertical
rotating shafts creating circular stabilized columns.
3 CASE HISTORY FOR THE ONSHORE
Therefore, CSM can be considered a WME technique.
MIXED-IN-PLACE TECHNOLOGY: LEUNA
The counter-rotating mixing wheels and the cutting
EX-REFINERY
teeth push the soil agglomerates through narrow gaps
between vertical steel plates, named “shear plates”
3.1 Introduction
(Fig. 3B). This produces a sort of forced mixing action
which enhance incorporation and homogenization of Between August and September 2005, a cut-off wall
the binder into the soil. composed of CSM panels was built in an ex-refinery
93
It included several boreholes and dynamic penetration
tests. Furthermore, laboratory tests for the determina-
tion of grain size distribution, permeability and triaxial
unconsolidated and undrained strength were carried
out on soil samples retrieved from the boreholes. Some
of these samples were also classified by means of
Atterberg limits and chemical analyses. Inside each
borehole a piezometer was installed to monitor the
groundwater level and the quality of the water before
and after treatment.
The resulting grain size distribution curves are
depicted in Fig. 5. Each curve in the graph is individu-
ated by an alphanumeric string reporting the borehole
number, the sample number, and the depth at which
the sample was taken. The curves show the great het-
erogeneity characterizing the subsoil at the site of
Leuna.
Triaxial undrained tests were performed on undis-
turbed cohesive samples collected from the site. The
undrained shear strength varied from 120 kPa to
247 kPa, denoting the high overconsolidation of the
clay deposits.
The classification of these soils based on the
Casagrande plasticity chart is shown in Fig. 6. The
finest fraction could be classified as inorganic silty
clay of medium plasticity (CL). Moreover, hydraulic
conductivities of the order of magnitude of 10−10 to
10−11 m/s were measured on soil specimens represen-
tative of the fine deposits of Leuna.
Close to each borehole, a dynamic penetration test
was executed in order to obtain a more effective cor-
relation between test results and visual inspection. An
example is presented in Fig. 7, in which the DP results
are compared with the information contained in the
corresponding borehole report.
From the in-situ and laboratory investigation cam-
paign the following geotechnical profile was derived.
The first layer consisted of loose to medium-dense
sandy fillings, with a variable thickness of 0.0–2.2 m
from the ground level (up to 2.64 m close to B5413).
The second level, containing a larger fine frac-
tion with respect to the first, was composed of Loess,
glacial marl, or a less weathered gravel terrace depend-
ing on the considered site area. The cohesive part had
a soft to hard consistency, while the granular material
was characterized by a loose to medium-dense state.
Some organic formations of soft to stiff consistency
were encountered between 2.2–6.1 m from the surface.
Figure 4. MIDOS pile tested in Germany in 2013 in silica Sand and gravel of the alluvial terrace, prevalently
sand. at a loose to medium dense state, formed the third
level. In some DP tests, NDP was over 60 blows per
site in Leuna (Germany) in order to limit the ground- 10 cm. These high values were probably related to the
water contamination in the surrounding area. The presence of boulders or cobbles. The thickness of this
wall was 450 m long (6400 m2 ) and 203 panels were layer was variable throughout the jobsite area and was
necessary for its completion. between 6.1 m and 10 m below the ground level. On
In addition to the cut-off wall, a retaining wall was the north (B5407 to B5409 and B5500), the surface
produced for a separate test purpose. quaternary deposits described so far were found to lie
directly on the bedrock.
3.2 Geotechnical characterization of the site
In the remaining area, a tertiary layer was located
A comprehensive geotechnical investigation was in between. The tertiary deposits incorporated a
accomplished in advance to the CSM wall construction. sequence of non-carbonate sands, silts, and clays with
94
Figure 7. Dynamic penetration test results obtained from
the preliminary ground investigation at the jobsite of Leuna –
Borehole B5418.
Table 1. Mix design at Leuna – Cut-off and structural wall.

Figure 5. Grain size distribution curves obtained from the Parameter Cut-off wall Structural wall
soil samples collected at the site.
Cement (CEM III) (kg) 373 709∗
Water (l/m3 ) 858 682
Bentonite (kg) 40 20
w/c (water/cement) 2.3 0.96
w/b (water/bentonite) 21.45 34.1
∗
: initially set equal to 909 kg.
3.3 Mix design, production data, and results

The cut-off CSM panels had a maximum depth of 16 m
below ground level and a thickness of 0.64 m. The
construction was performed following the one-phase
system and the back-step procedure (fresh-in-hard),
with the realization of 2.8 m wide primary panels and
2.0 m wide secondary panels, i.e. with an overcut of
0.4 m at each side.
As far as the cut-off wall production concerns, the
mix parameters provided in Tab. 1 were adopted at the
site. In order to introduce into the ground an amount
Figure 6. Classification of the natural subsoil at the jobsite of cement corresponding to 152 kg/m3 of natural soil,
of Leuna by means of Casagrande chart. a flow rate of 858 l/m3 was assumed. The cement used
was a CEM III 32.5 N NW/HS (85% slag and 15%
OPC). The retaining wall, on the contrary, was cre-
ated with a higher amount of cement to ensure the
a thickness ranging from 13.5 m to 41.0 m. The cohe- achievement of a predefined strength level (Tab. 1).
sive formations had a stiff to solid consistency, which The same flow rate of 858 l/m3 was adopted to intro-
led, in some cases, the DP tests to refusal. The tertiary duce into the ground an amount of cement correspond-
sands were denoted by a medium dense state increasing to about 550 kg/m3 of natural soil. Furthermore,
ing in density with depth. In some boreholes, local the same binder as in the cut-off wall was used.
organic formations of lignite were identified. Hydraulic conductivity values of the order of mag-
Finally, the weathered and disaggregated sandstone nitude of 10−9 to 10−11 m/s were obtained from
bedrock with a stiff to semi-solid consistency was permeability tests performed on cut-off wall sam-
reached at a variable depth, which in the North-western ples. Unconfined compressive strength (UCS) values
area of the jobsite was approximately comprised ranging from 0.5 MPa to 6 MPa were measured on
between 5.0 m and 7.0 m. samples collected from the structural wall.
95
4 THE MIDOS TECHNOLOGY FOR OFFSHORE
PILED FOUNDATION APPLICATIONS
4.1 Laboratory tests in calcareous and silica sands

To investigate the feasibility of using MIDOS in cal-
careous sands, a laboratory study was required to
assess both the geotechnical performance of the pile
and the structural behavior of the Mixed-in-Place
(MIP) grout, which forms the pile body. Doherty
et al. (2015; 2016) showed that the silica sand tested
(Blassington sand) was chosen as a control or refer-
ence sample. This was deemed particularly important
because the MIDOS system has already been shown
to be an effective pile installation technique in sil-
ica deposits, with the load carrying capacity shown
to be in-line with predicted values for conventional
D&G piles. As a result, direct comparison of the silica
and calcareous sand behavior was deemed essential
for this study. The testing regime focused on both the
potential geotechnical and structural failure modes. A
carbonate sand from Dog’s Bay (west coast Ireland)
was used. Portland CEM II cement was used through-
out the testing regime and a water/cement (w/c) ratio
of 0.4 was adopted for the tests as suggested by ISO
19902 for fixed steel offshore structures. Different Figure 8. Comparison of UCS for grouted samples of sil-
cement-to-sand ratios (c/s) were used for the tests, ica and carbonate sand (top); comparison of tensile force
i.e. 15-25-35%. These values were chosen to simu- for grouted samples of silica and carbonate sand (below)
late a high installation (15% c/s) and low installation (modified after Doherty et al. 2015).
time (35% c/s), where the cement would have more
time to be mixed with the soil. The 25% c/s is a com- Table 2. Silica sand soil model properties (modified after
promise value between the low and high installation Spagnoli et al. 2015)
time. A series of tensile tests and UCS tests were per-
formed at 3 days, 7 days and 28 days and across the Diameter Pile
range of c/s ratios noted above. The shear rate for the chamber length
Model no Soil (m) (m)
tensile test was 50 N/s whereas for the compression
test was 0.5 MPa/s (Doherty et al. 2015). The results
1 Silica sand 2 20
demonstrated identical trends in all cases, regardless of 2 2 40
c/s ratio and curing time. This is particularly evident 3 2 60
in Fig. 8, which directly compares the compression 4 3 20
and tension tests of grout samples with the same test- 5 3 40
ing time and c/s values (Doherty et al. 2015). Over 6 3 60
the entire range of parameters considered, the grout 7 Calcareous sand 2 20
samples demonstrated comparable behavior, suggest- 8 2 40
ing that the structural performance of the MIDOS pile 9 2 60
10 3 20
body should be similar in both calcareous and silica
11 3 40
deposits. 12 3 60
4.2 Finite element analysis

In tension the shaft resistance is the primary resis-
Spagnoli et al. (2015) showed MIDOS pile models tance component. The radial effective stress at failure
were developed with Plaxis 3D to analyze a range of is highly dependent on the initial horizontal effective
pile length to diameter ratios in silica and calcareous stress (at rest condition) σrc , as the MIDOS pile aims
sand deposits. In order to investigate the behavior of to maintain a constant horizontal lateral stress dur-
MIDOS piles in calcareous sand, suitable soil parame- ing installation the simplistic “wished-in-place” model
ters were calibrated in order to reflect the load bearing construction should provide a reasonable initial instil-
behavior of this type of soil more precisely. lation stress condition – compared to that of driven
The range of model dimensions and the corre- piles where the driving stresses can be impossible to
sponding model numbers are presented in Table 2. Six accurately model. The soil models used in this study
geometry models were analyzed for each soil type, are assumed to be normally consolidated and the lateral
to cover the range of diameter/length combinations earth pressure coefficients are automatically deter-
presented. mined by Plaxis as Konc = 1 − sin ϕ . In practice the
96
initial soil stresses would be assessed and inputted
directly to Plaxis using the results of in-situ geotech-
nical testing such as CPT, to simulate the specific
conditions encountered at a real site. As a simplified
method of allowing for some strength reduction a “soft
soil” layer was modelled underneath the chamber to a
depth equal to the chamber diameter. The soft soil was
given stiffness parameters equal to half of the regular
sand parameters and a friction angle equal to the con-
stant volume friction angle of the soil in the respective
model. This has only been considered in the silica sand,
as the soil beneath the chamber in calcareous sand is
modelled using the crushed parameters. The silica and
calcareous sand models were calibrated using triaxial
tests. For the calcareous sands, triaxial tests conducted
at varying effective confining stresses (100 kPa and
1000 kPa) found that sand stiffness deteriorated signif-
icantly at higher confining stresses, subsequent grain
size distribution on those samples showed clear evi-
dence of particle degradation during the 1000 kPa
triaxial tests. Due to the limitations of the hardening
soil model, the soil degradation and stress could not be
calibrated into a single soil model. Therefore two sets
of soil properties were adopted and the soil proper-
ties were optimized using triaxial tests on uncrushed
and crushed samples. A limit stress of 400 kPa was
selected as the pressure at which the sand will begin
to deteriorate. Once this stress was reached, the soil
properties were changed to that of the crushed sand to
model the stress softening behavior of the calcareous
sand. This process required the model to be first run
using the intact soil only. The resulting soil stresses
were then examined and the depth of soil where the
stresses exceed the 400 kPa determined. The soil body
below this depth was assumed to be crushed, while
above this depth, the non-crushed soil model was used.
Therefore, in the next stage of analysis, a mixed soil
model was generated, where the soil properties at
depths below the identified depth were modified to Figure 9. Numerical vs. analytical capacity assessment of
represent the behavior of crushed soil, and the model the piles in silica sand for the models no. 4, 5 and 6.
was re-run. This methodology was found to provide a
more realistic response than adopting a fully crushed 5 CONCLUSIONS
or non-crushed approach (Spagnoli et al. 2015). Com-
parison of the tension and compression performance Offshore driven piles are the most common foundation
of the various pile geometries showed some scatter type. However, drilled-and-grouted piles, although
in the results obtained, however in general the sim- time consuming, are used where driven piles reach
ulations highlight similar performance between the their limits. The Deep Mixing Method (DMM) is
piles in silica and calcareous sand (Spagnoli et al. an in-situ soil treatment technology whereby binding
2015). materials are added and blended with soils in order
The results of numerical analysis of the piles in to improve their hydraulic and mechanical properties.
silica sand were also compared with those predicted This paper analyzed two case histories, where this tech-
using the API method, see API (2014), and a CPT- nology successfully was used for an onshore and for a
based approach, see Igoe et al. (2014), assuming potential offshore application.
three different values for the average qc of the seabed
(see Fig. 9). The comparison shows that the Plaxis
models predict significantly lower capacities than the REFERENCES
analytical predictions. It should be noted that Finite Abbs, A.F., Waterton, C.A., Bell, R.A., Khorshid, M.S. &
Element Modelling is not particularly suited to large Jewell, R.J. 1988. Evaluation of pile friction from con-
deformation analysis and convergence issues may stop ductor load tests. In R.J. Jewell & M.S. Khorshid (eds),
the model before the ultimate limit state capacity, as Proc. Int. Conf. Calcareous Sediments, Perth, 2: 607–616.
predicted in the analytical methods, is reached. Rotterdam: Balkema.
97
American Petroleum Institute. 2014. Planning, Design- Gerwick, B.C. 2007. Construction of Marine and Offshore
ing, and Constructing Fixed Offshore Platforms-Working Structures. Boca Raton: CRC Press.
Stress Design. API RP 2A-WSD:2014-11. Igoe, D., Spagnoli, G., Doherty, P. & Weixler, L. 2014. Design
Barthelemy, H.C., Martin, R., Le Tirant, P.C., Nauroy, J.F. & of a novel drilled-and-grouted pile in sand for offshore
Cipriano de Medeiros Jr, J. 1987. Grouted driven piles: oil&gas structures. Marine Structures, 39: 39–49.
an economic and safe alternate for pile foundations. Infield 2013. Global Perspectives Fixed Plaftorms Market
Proc Offshore Technology Conf, Houston, OTC5409, Report to 2017. London: Infield Systems Limited.
http://dx.doi.org/10.4043/5409-MS. King R.W., van Hooydonk, W.E., Kolk, H.F. & Windle, D.
Bruce, D. 2000. An introduction to the deep soil mixing 1980. Geotechnical investigations of calcareous soils on
methods as used in geotechnical applications. Tech. rep., the Northwest Shelf, Australia. Proc Offshore Technol-
Federal HighWay Administration (FHWA). FHWA-RD- ogy Conf, Houston, OTC3772, http://dx.doi.org/10.4043/
99-138. 3772-MS.
Clarke, J. 1993. Large Scale Pile Tests in Clay. London: Lee, C.Y. & Poulos, H.G. 1991. Tests on model instrumented
Thomas Telford Ltd. grouted piles in offshore calcareous soil. Journal of
Colliat, J.L., Foray, P.Y. & Nauroy, J.F. 1999. Behaviour of Geotechnical Engineering-ASCE, 117(11): 1738–1753.
carbonate sands and design of driven piles. Proc. Off- Le Tirant, P. & Nauroy, J.F. 1994. Foundations in Carbonate
shore Mediterranean Conf, Ravenna, March 24-26 1999: Soils. Paris: Editions Technip.
171–183. Mitchell, J. 1981. Soil improvement – state of the art report.
De Mello, J.R.C., Amaral, C.D.S, Maia da Costa, A., Rosas, Proc. 10th Int. Conf. Soil Mechanics and Geotechnical
M.M., Decnop Coelho, P.S. & Porto, E.C. 1989. Closed- Engineering. Stockholm (Sweden): 509–565.
ended pipe piles: testing and piling in calcareous sand. Murff, J.D. 1987. Pile capacity in calcareous sands: state
Proc Offshore Technology Conf, Houston, OTC6000, of the art. Journal of Geotechnical Engineering-ASCE,
http://dx.doi.org/10.4043/6000-MS. 113(5): 491–507.
Digre, K.A., Brasted, L.K. & Marshall, P.W. 1989. The Poulos, H.G. 1988. Marine Geotechnics. London: Unwin
design of the Bullwinkle platform. Proc Offshore Technol- Hyman Ltd.
ogy Conf, Houston, OTC6050, http://dx.doi.org/10.4043/ Randolph, M.F. 1988. The axial capacity of deep foundations
6050-MS. in calcareous soils. In R.J. Jewell & M.S. Khorshid (eds),
Doherty, P., Spagnoli, G. & Doherty, M. 2015. Laboratory Proc. Int. Conf. Calcareous Sediments, Perth, 2, 837–857.
investigations to assess the feasibility of employing Rotterdam: Balkema.
a novel mixed-in-place offshore pile in calcareous Spagnoli, G., Doherty, P., Bellato, D. & Weixler, L. 2014.
deposits. Ships and Offshore Structrures. DOI: 10.1080/ Latest technological developments in offshore deep mix-
17445302.2015.1056579. ing for piled oil and gas platforms. Proc. Int. Conf. on
Doherty, P., Spagnoli, G. & Bellato, D. 2016. Mixed-in-place Ocean, Offshore and Arctic Engineering, OMAE, doi:
response of two carbonate sands. Geotechnical Engineer- 10.1115/OMAE2014-23045.
ing, 169(2): 153–163, http://dx.doi.org/10.1680/jgeen.15. Spagnoli, G., Doherty, P., Murphy, G. & Attari, A. 2015. Esti-
00058. mation of the compression and tension loads for a novel
Dutt, R.N. & Cheng, A.P. 1984. Frictional response of mixed-in-place offshore pile for oil and gas platforms in
piles in calcareous deposits. Proc Offshore Technol- silica and calcareous sands. Journal of Petroleum Science
ogy Conf, Houston, OTC4840, http://dx.doi.org/10.4043/ and Engineering, 136: 1–11.
4838-MS. Topolnicki, M. 2004. In situ soil mixing. In M.P. Moseley &
George, P.J. & Wood, D.M. 1976. Offshore Soil Mechanics. K. Kirsch (eds). Ground Improvement: 331–428.Oxon:
Cambridge: University Engineering Department. Spon Press.
98
A comparative study of wind farm potential in shallow-waters in

Brazil and Scotland
G. Alves & S.B. Mickovski

School of Engineering and Built Environment, Glasgow Caledonian University, Glasgow Scotland, UK
ABSTRACT: Facing the predicted climate changes and the potential measures to reduce carbon dioxide emis-
sions globally, there is an increased use of clean renewable energy sources. In Europe, the wind power has often
been used in the last decades and, more recently, a trend for offshore utilization has been noticed. However,
there is a lack of knowledge on the specific planning and engineering considerations relevant to construction of
wind farms in shallow waters. The aim of this study is investigate the potential for wind power plants in shallow
waters at sites with contrasting bio-geo-climatic characteristics. Scotland is well known for the potential offshore
wind energy and in Brazil, where this technique is not yet in use, there are only initial studies with indication
of governmental interest for future implantations in great lakes. The focus of this study is on the environmental
impact assessment as well as the impact on coastal communities and public participation in the decision-making.
To this end, this study reviews the current practice in site determination, planning, and construction while inves-
tigating the potential for construction of wind farms at two shallow waters locations in Scotland and Brazil. A
comparative study between the two countries for offshore wind turbines geotechnics with proper structure at
low depths of water, taking into accounts the environmental characteristics and needs along with suitable wind
towers construction is outlined. The results show variation of construction models used by both countries regard-
ing the environmental impact assessment and a convergent tendency for projects planning acceptance by the
communities. The study identifies a potential transfer of knowledge in specific areas of planning, development,
and construction of offshore wind farms in shallow waters. Additionally, this study recognizes the importance
of geo-environmental site investigation and classification within a sustainability framework.
Keywords: wind power, offshore geotechnics, shallow water, environmental impact assessment, planning,
public participation
1 INTRODUCTION in the de-carbonization of the EU energy system. For

the UK it is expected that by 2020, 15% of the gener-
In many countries, natural gas, oil, coal and nuclear ated energy (electricity, heating and transport) will be
sources are the main source of energy. These are recog- coming from renewable sources (DECC et al. 2015),
nized as highly harmful to the environment for energy and a commitment from public and private sector will
production, either by their extraction, use/waste burn- be needed.
ing and returned to the environment and also its rarity Wind energy has been very important in clean
nature, not being renewable. Policies for reduction car- energy production for being one of the most mature
bon emissions in the atmosphere have been adopted technologies in use, especially in shallow waters (Koh,
and have given impulse to development and energy Ng 2016), but offshore wind projects can present
production from renewable energy sources and not pol- environmental problems such as noise, landscape alter-
luting (Ochieng et al. 2014). With individual targets for ation, mortality of wildlife especially birds, shadow
reduction of emission of greenhouse gases to be met, flicker, in addition to deal with the constant require-
countries are committed to limiting the emission of ments of the people living around the enterprises and
certain gases to the environment so as to combat the feel affected in some way (Aitken 2010). Hence, to
greenhouse effect. ensure the viability of projects with perceived potential
The renewable energy is a term that refers to the for environmental threat, it is necessary to include pub-
source of energy that is not exhausted by the long- lic consultations for the feasibility of projects (BWEA,
term use, and regenerates in relatively short time such 2002).
as the wind energy. Commonly recognized as environ- Wind farms have a tendency to be expensive when
mentally friendly, it has been playing an important part compared to other forms of power generation (Levitt
99
et al. 2011), and considering the offshore energy gen- designed for the country, Firth of Forth, in the region
erated, the expenses are even higher with increasing of Edinburgh, will have more than 700 wind turbines
distance from the coast and depth at which the struc- in relatively shallow depth (50 m; (O’Keeffe, Haggett
tures will be installed have being the main factors 2012). It is expected that the foundation structure to
that influence the high value of the investments (Hig- be used in this project is the same used in the Energy
gins, Foley 2014). Typically, in remote off-shore the Park Life demonstration tower already installed.
wind speeds tend to be larger (Cavazzi, Dutton 2016). In Brazil, through the NPCC – National Policy on
Therefore, more robust structures are used with greater Climate Change (MMA – Ministerio do Meio Ambi-
demand for materials and there is a greater need for ente, Brazil 2009), the country ensures efforts to build
services and logistics (Sovacool, Enevoldsen 2015). a low-carbon consumption savings in electric power
However, the notion of depth is according to the evo- generation. Thus, as it is established by the Kyoto Pro-
lution of technology. In shallow waters, the technology tocol, Brazil voluntarily assumes the commitment for
for construction of structures is already well known, reduction of greenhouse gases emission projected until
and the foundation is the main differential being 5 2020. Currently, Brazil already has its energetic and,
types of foundation structures usually used (Koh, Ng more specifically, its electric matrix based on renew-
2016). The need for applying each of the structures able energy. Hydropower is the largest electric power
varies with the depth that will be accessed: Monopile supplier in the Brazilian electrical system where the
(0–25 m), Gravity based (0–25 m), Jacket (20–50 m), non-renewable energy source is only 26.9% in 2014. In
Tripod (20–50 m), Tripile (30–50). the same year the wind energy production had 85.6%
In the meantime, the costs of projects of this nature, increases over the previous year, reaching 12.210 GWh
offshore, tend to have up, often, three times the cost of (EPE Brazil 2015).
the projects to be carried out onshore and this is due At the moment, Brazil still does not use off-
not only the distance from the coast or depth, but also shore technology for wind power generation; however,
for all special treatment that the structures need to sup- potential sites for the development have been investi-
port the abrasive conditions of the sea (Wu et al. 2014). gated (Schubert et al. 2014). In general, the areas in
However, this is only an assumption, since the compar- which the studies of viable and favorable winds are
ison of values between offshore projects is relatively located are the sites with the lower depths, in areas
complex, especially when considering the different near the coast. This favors the installation of wind tur-
physical and environmental characteristics, and eco- bines with more mature and inexpensive technology.
nomic reality of the construction sites (Dismukes, Offshore wind farms usually have large dimensions,
Upton Jr. 2015). and in view of the necessary capital investment for
generation, financial details are quite important.
According to the complementarity theory of
2 BACKGROUND hydropower system in Brazil (Silva et al. 2016) the
wind energy can provide an alternative clean source
Currently, the alternatives of offshore wind energy of electric energy during the periods of lower level of
to be explored are not only related to the generation water storage due to decreased rainfall. Wind energy
capacity of the wind turbines but also to installation has great potential to be explored favoring the main-
sites (Le et al. 2014). Some of the sites with favorable tenance of the level of reservoirs of rainfall shortage
wind for the power plants can present challenges to the periods (ONS Brazil 2015).
implementation. However, the governments of Brazil It was approved in 2014 by CONAMA – National
and Scotland are driving efforts to develop the wind Environmental Council (Brazil) rules that regulate the
energy technology in theirs territory, once this source environmental licensing and establishes criteria for
has a huge potential to be explored in both countries. onshore electric plants and establishes that for enter-
Via the decisions of public policies, both countries, prises with low level environmental impact, the need
Brazil and Scotland, are driving their energetic sys- for public consultation is eliminated. Knowing that
tems to use mainly renewable energy sources in the offshore wind turbine development may affect marine
composition of their energetic matrix. Among these life negatively (Toonen, Lindeboom 2015), and many
efforts, it is the increasing of security of supply of other positive and negative effects are still unknown, it
energy sources from renewable alternatives sources, is necessary to have more studies to identify and assess
notably wind farms. the level of offshore wind plants impact.
Scotland decided to expand the UK targets and The rules that will regulate the environmental
aims to ensure that whole electric power generated in licensing for offshore power plants in Brazil would
the country can be provided from renewable energy be developed after the elaboration of standards for
sources (O’Keeffe, Haggett 2012), with planned at onshore projects, and has not been published up to
least 4.5 GW from offshore wind energy (DECC this date. It must be considered essential for a coun-
et al. 2015). A demonstration project (7 MW turbine try owning the regularization statutory process for
founded on jacket foundation; Renewable UK 2015) large enterprise development procedure as the imple-
was used to prove the concept, and a number of projects mentation of offshore wind farms. It ensures legal
are currently at different stages albeit none at construc- security and presents the projects more attractive to
tion phase. One of the biggest wind generation parks investors.
100
Investments are extremely important in supplies Patos, has a considerable number of indigenous com-
coming from alternative energy systems to the existing munities and areas of integral protection. On the other
hydroelectric system in Brazil. The greatest potential hand, in Scotland, it is possible to encounter a number
for hydroelectric power in the country is in use, leav- of ship wrecks in the areas for potential development.
ing the ones with greater environmental complexity The above can contribute negatively for the perception
(Eletrosul 2008) to be developed. These systems could of the offshore enterprises by the communities.
lower the dependence of the present hydropower sys- According to the review of several studies
tem and contribute to the sustainable development of (Betakova et al. 2015, Ladenburg, Lutzeyer 2012,
region accessed. Westerberg et al. 2015), the factor with the greatest
influence on the decision to supporting or not the wind
projects would be the loss of landscape or its alteration
3 SOCIAL ACCEPTANCE OF OFFSHORE
thereof. The landscape alteration can bring social reac-
WIND FARMS
tions and, if we consider large developments, effects
on the local fauna (Hammar et al. 2014). The environ-
According to surveys assessing the tendency of Euro-
mental impacts related to construction for wind power
pean citizens to fund and pay the price by advances
generation largely depend siting and the local envi-
in clean energy (Haggett 2011), if it could reflect in
ronmental (Leung, Yang 2012). The above shows that
the energy bill, it is still not clear with studies report-
there is a gap between public acceptance and local
ing conflicting results (Walker et al. 2014). Therefore,
opposition, and it can be better understood “when con-
it should be considered that not everyone is willing
sidering the environmental trade-off and global gains”
to pay for such development (Wüstenhagen, Bilharz
(Hall et al. 2013).
2006). Developers also assume the task of trying to
keep the accessible costs of the offshore wind farms,
since there is the unpopularity caused the increase in
energy bills. It is believed that developers would be 4 POTENTIAL ISSUES WITH WINDFARM
more willing to seek lower costs if this factor could CONSTRUCTION IN SHALLOW WATERS
raise the number of public supporters (Hooper et al.
2015). Shallow waters are those included in up to 50 meters
During the licensing process for offshore wind depth and the areas that comprise this water lamina
projects in UK, public consultation is necessary height are the most targeted for development projects
according to the standards set by The Crown Estate (Raadal et al. 2014). The tendency is that the depth
with presentation of environmental impact assess- maritime increase according to the distance from the
ment (EIA) studies for discussion with the public. coast. Thus, the shallow waters, ideal for enterprises
At this stage a range of stakeholders are considered, with low cost of construction are located in areas where
together with other interested parties for clarifications the visual impact is considered a factor to decrease
and agreements that may be necessary and possible support the execution of the projects. Direct impacts
financial compensation or mitigating environmen- related to the public would be linked to the changes
tal measures caused by the construction process are presented in the landscape. The divergence of opinion
sought. as the landscape alteration caused by wind structures
Environmental issues, such as landscape change, is still considered an area of research that needs further
can clearly change public perception (Dai et al. 2015). clarification.
The study of public participation becomes neces- The foundations design and construction for these
sary since the lack of support for development from projects can also be regarded as environmental impact
the public (e.g. NIMBY – Not in My Back Yard; factor. The installation of the structures can provoke
Bidwell 2013) can seriously derail offshore projects. disturbances in the marine environment, even being
Companies seek to compensate residents affected by insensitive to the human being, at the first moment; it
environmental problems what cause in many partici- can become harmful to marine species. Basically, all
pants the feeling of bribery. (Walker et al. 2014) and structures require a step of drilling the seabed, with
the developers must consider the socioeconomic real- exception of gravity base foundation. Such founda-
ity of the region affected during planning. Factors such tions that require drilling for installation may suspend
as age, income and educational level of the affected seabed sediments becoming harmful to life the fish and
population can interfere with understanding capacity even their reproduction. Depending on the structure of
of the proposed projects and even consider the possi- the soil in chosen sites, it is possible that more harmful
bility of the respondents individuals possess different alternatives need to be used, as the use of explosives.
interests when suggested projects. The impact sound emitted by construction works of the
The connection of the communities with the land projects may also be mentioned as an environmental
and cultural heritage can be determining factors for threat factor. Some studies try to identify if there are
decision support or opposition (Lombard, Ferreira environmental threat related to grid connection by the
2014). Indigenous communities in several areas in emission of possible electromagnetic waves.
Brazil have deep cultural connection with the envi- Considering the continental shelf, the water depths
ronment. One of the study areas for offshore wind in the coastal areas of southern Brazil and coast bor-
farm implementation in southern Brazil, the Lagoa dos dering the Lagoa dos Patos are relatively shallow. The
101
area has favourable winds for the siting of a wind farm. offshore. The application of questionnaire should be
The lagoon depth ranges from 7 m into the lagoon to performed during field work in order to collect and
up to 50m in oceanic areas bordering the shore of set the sample that will be accessed. The interview
the lagoon. Due to this and because of marine habi- will have semi structured questions to allow accessed
tats, gravity foundation structure can be an option have the possibility, at some point, openly express their
more financially viable and better adaptable to the concerns. Research will also be done by observation
local environment. The use of more complex structures and informal conversations that allows the partici-
needing higher value for construction are disregarded pants to be more confidents to express their concerns.
in this case, once the towers anchored that are con- The questionnaire based on an embracing literature
siderably a new technology for deep waters (Leung, review should be built considering the place attach-
Yang 2012), which ensures the project is designed in ment and knowledge of the region, common public
that region access a technology to foundations and concerns, assessment of level of knowledge technolo-
structures with more mature technology and better gies proposed and expectations for the respondents.
economic value. For deep water, with water depth over For this analysis should be considered the evaluation of
50 meters, the structures begin to provide high value cultural heritage by the participants related to the envi-
and floating structures present a good solution (Koh, ronment, and socio-demographic characteristics as age
Ng 2016). and education, as the level of understanding of the
For Scotland, a concern reported by fishing entities level of involvement and complexity of projects that
is the issue of navigation in and adjacent to wind tur- is important factor for supporting the constructions.
bines (Schillings et al. 2012). It must be considered It is intended to separate the participants into groups
that the North Sea, in the case of Scottish waters is in order to achieve equality among socioeconomic
indeed quite busy, with oil production, gas, military factors accessed.
activities and navigation routes (Toonen, Lindeboom
2015). Co-location of offshore wind farms and fish-
6 CONCLUSIONS
eries are studied, once artificial reefs are created from
the turbine foundations. However, in many cases, it is
The present paper understands as necessary a continu-
not possible. The safety zones for the boats are 500
ous study of public acceptance of energy enterprises,
meters away from the structures (Hooper et al. 2015).
mainly the offshore wind farms. It considers the great
Such restrictive measures it is necessary for safety
number of energy production planned coming from
purposes. Thus, to practice co-location between com-
this source for the next years in United Kingdom and
panies and fisheries would be necessary application
the possibility for future implementation in Brazil.
of insurance, training and open direct communication
Assuming the evolution of energy production sector is
channel between the company and fishermen.
been targeting the generation of clean energy to avoid
the carbon emission on the atmosphere, this evolution
5 PROPOSED METHODOLOGY is a tool for achieve the common welfare, for society
and environment. This new focus for energy supply
The measurement of social acceptance of offshore combats the global warming and climate changes, and
wind farms has not been completely explained in the it is necessary to include the public participation, opin-
past, while the low level of support to the enterprises ion and understanding of the enterprises. Researches
gives a bad popularity to the technology and, there- have been showing that the level of acceptance of the
fore, makes the offshore wind farms less attractive for wind turbines can increase with the better understand-
investments. Considering the number of offshore wind ing of the enterprises, planning and benefits for the
farms planned for Scotland and the increased possibil- local communities.
ity of offshore wind farms in Brazil begin to be planned There are cases of public manifestations for no sup-
in areas where environmental issues are relevant, the porting for implantation of wind onshore parks and
assessment of social acceptance can provide a tool for the same kind of view can be replicated for the off-
best guideline of development in agreement with the shore structures, and these movements can generate
public concerns. For this study, we assessed the cur- unplanned expensive for the enterprise, once the com-
rent research on the public participation and the effects panies will have to deal with unsatisfied communities.
on the technology for offshore wind farms in shallow However better explanations and involvement of the
waters, contrasting Scotland and Brazil. communities located near the structures perhaps can
For the next stage on this research, the data collec- provide an improvement of the understanding of the
tion and analysis of the effects accessed responses, this offshore wind parks and show the importance for the
study will be carried out by interviews with the stake- whole society about the support for this source of
holders based on a standardized questionnaire survey energy supply. The enterprise near the coast might pro-
for merchants and institutions possible involved in the vide a tourist attraction linking the region to sustain-
activities undertaken at the site, which will provide able development, reflecting the ideology of providing
some quantitative data. The sites of research inter- clean energy and progress thought focused on human
est are Lagoa dos Patos, in Brazil, and the coast of well-being, or even, contributing for the local economy
Aberdeen, in Scotland. Both sites have been showing when possible, as the example of co-location between
suitable for implementation of wind energy, mainly fishers and structures.
102
Second few researches, the NIMBY’s do not explain Haggett, C., 2011. Understanding public responses to off-
satisfactorily the intention of not supporting wind shore wind power. Energy Policy. 39(2), pp. 503–510.
farms. The main reasons for decreasing and increas- Hall, N., Ashworth, P. and Devine-Wright, P., 2013. Soci-
ing of support of the enterprises are still conflicting, etal acceptance of wind farms: Analysis of four common
themes across Australian case studies. Energy Policy. 58
requiring more studies in the field of acceptance of pp. 200–208.
offshore and onshore wind farms. The benefits com- Hammar, L., Wikström, A. and Molander, S., 2014. Assessing
ing from the wind power plants should not be felt as ecological risks of offshore wind power on Kattegat cod.
bribery and the outcomes prevenient from the enter- Renewable Energy. 66(0), pp. 414–424.
prises in the communities makes necessary an exten- Higgins, P. and Foley, A., 2014. The evolution of offshore
sive study to clarification and, therefore, contribution wind power in the United Kingdom. Renewable and
for the development of technology in accordance with Sustainable Energy Reviews. 37(0), pp. 599–612.
population interest. Hooper, T., Ashley, M. and Austen, M., 2015. Perceptions of
fishers and developers on the co-location of offshore wind
farms and decapod fisheries in the UK. Marine Policy. 61
ACKNOWLEDGEMENTS pp. 16–22.
Koh, J.H. and Ng, E.Y.K., 2016. Downwind offshore
wind turbines: Opportunities, trends and technical chal-
The authors acknowledge CNPq (Conselho Nacional lenges. Renewable and Sustainable Energy Reviews. 54
de Desenvolvimento Científico e Tecnológico – pp. 797–808.
Brazil) for the financial support. Ladenburg, J. and Lutzeyer, S., 2012. The economics of visual
disamenity reductions of offshore wind farms—Review
and suggestions from an emerging field. Renewable and
REFERENCES Sustainable Energy Reviews. 16(9), pp. 6793–6802.
Le, T.M.H., Eiksund, G.R., Strøm, P.J. and Saue, M., 2014.
Aitken, M., 2010. Wind power and community benefits: Geological and geotechnical characterisation for offshore
Challenges and opportunities. Energy Policy. 38(10), pp. wind turbine foundations: A case study of the Sher-
6066–6075. ingham Shoal wind farm. Engineering Geology. 177(0),
Barlow, E., et al., 2015. Exploring the impact of innovative pp. 40–53.
developments to the installation process for an offshore Leung, D.Y.C. and Yang, Y., 2012. Wind energy development
wind farm. Ocean Engineering. 109 pp. 623–634. and its environmental impact: A review. Renewable and
Betakova, V., Vojar, J. and Sklenicka, P., 2015. Wind turbines Sustainable Energy Reviews. 16(1), pp. 1031–1039.
location: How many and how far?. Applied Energy.151 Levitt,A.C., et al., 2011. Pricing offshore wind power. Energy
pp. 23–31. Policy. 39(10), pp. 6408–6421.
Bidwell, D., 2013. The role of values in public beliefs and atti- Lombard, A. and Ferreira, S., 2014. Residents’ attitudes to
tudes towards commercial wind energy. Energy Policy.58 proposed wind farms in the West Coast region of South
pp. 189–199. Africa:A social perspective from the South. Energy Policy.
Bwea – British Wind Energy Association, 2002. Best pratice 66 pp. 390–399.
guideline: Consultation for offshore wind energy devel- MMA – Ministerio do Meio Ambiente, Brazil., 2009.
opments. [database online]. London: BWEA. [viewed Política Nacional sobre Mudança do Clima. [database
October, 2015].Available from: http://www.energy.ca.gov/ online].Brasilia: Brazil, 29/12/2009, ISBN 12187.
windguidelines/documents/other_guidelines/BWEA-BPG- O’keeffe, A. and Haggett, C., 2012. An investigation into the
OFFSHORE.PDF. potential barriers facing the development of offshore wind
Cavazzi, S. and Dutton, A.G., 2016. An Offshore Wind energy in Scotland: Case study – Firth of Forth offshore
Energy Geographic Information System (OWE-GIS) for wind farm. Renewable and Sustainable Energy Reviews.
assessment of the UK’s offshore wind energy potential. 16(6), pp. 3711–3721.
Renewable Energy. 87, Part 1 pp. 212–228. Ochieng, E.G., et al., 2014. Future for offshore wind energy
Decc et al., 2015. Energy in Scotland 2015. [database in the United Kingdom: The way forward. Renewable and
online]. The Scottish Government Website: The Scot- Sustainable Energy Reviews. 39(0), pp. 655–666.
tish Government. [viewed June, 2015]. Available from: ONS Brazil, 2015. ONS – Operador Nacional do Sistema
http://www.gov.scot/Topics/Statistics/Browse/Business/ Elétrico – Situacao dos Reservatorios, Energia Armazeda
Energy/EIS2015. por Região. [database online]. ONS. [viewed April, 2015].
Dismukes, D.E. and Upton JR., G.B., 2015. Economies Available from: www.ons.org.br.
of scale, learning effects and offshore wind develop- Raadal, H.L., Vold, B.I., Myhr, A. and Nygaard, T.A.,
ment costs. Renewable Energy. 83 pp. 61–66. 2014. GHG emissions and energy performance of
Eletrosul., 2008. Relatorio Ambiental Simplificado do Com- offshore wind power. Renewable Energy. 66(0), pp.
plexo Eolico Coxilha Negra – Modulos V, VI e VII. 314–324.
AmbioTech Consultoria ed., Curitiba, BR: Eletrosul Cen- Renewable UK, 2015. Offshore Wind Project Timelines
trais Eletricas. 2015. [database online]. London: Renewable UK.
EPE Brazil, 2015. Brazilian Energy Balance 2015 Year [viewed September, 2015]. Available from: http://www.
2014. [database online]. Rio de Janeiro: Empresa de renewableuk. com/en/publications/reports.cfm/Offshore-
Pesquisa Energética. [viewed January, 2015]. Available Wind-Project-Timelines-2015.
from: https://ben.epe.gov.br/downloads/Relatorio_Final_ Schillings, C., et al., 2012. A decision support system for
BEN_2015.pdf assessing offshore wind energy potential in the North Sea.
Firestone, J. and Kempton, W., 2007. Public opinion about Energy Policy. 49(0), pp. 541–551.
large offshore wind power: Underlying factors. Energy Schubert et al., 2014. Atlas Eolico: Rio Grande do Sul. 2nd
Policy. 35(3), pp. 1584–1598. ed. Porto Alegre: AGDI.
103
Silva, A.R., Pimenta, F.M., Assireu, A.T. and Spyrides, England. Energy Research & Social Science. 3(0),
M.H.C., 2016. Complementarity of Brazil’s hydro and pp. 46–54.
offshore wind power. Renewable and Sustainable Energy Westerberg, V., Jacobsen, J.B. and Lifran, R., 2015. Offshore
Reviews. 56 pp. 413–427. wind farms in Southern Europe – Determining tourist
Sovacool, B.K. and Enevoldsen, P., 2015. One style to build preference and social acceptance. Energy Research &
them all: Corporate culture and innovation in the offshore Social Science. 10 pp. 165–179.
wind industry. Energy Policy. 86 pp. 402–415. Wu, J., Wang, Z. and Wang, G., 2014. The key technologies
Toonen, H.M. and Lindeboom, H.J., 2015. Dark green elec- and development of offshore wind farm in China. Renew-
tricity comes from the sea: Capitalizing on ecological mer- able and Sustainable Energy Reviews. 34 pp. 453–462.
its of offshore wind power?. Renewable and Sustainable Wüstenhagen, R. and Bilharz, M., 2006. Green energy mar-
Energy Reviews. 42(0), pp. 1023–1033. ket development in Germany: effective public policy
Walker, B.J.A., Wiersma, B. and Bailey, E., 2014. Com- and emerging customer demand. Energy Policy. 34(13),
munity benefits, framing and the social acceptance pp. 1681–1696.
ofoffshore wind farms: An experimental study in
104
A new backfill material enhancing axial bedding of district heating pipes
B. Wagner & T. Neidhart

Ostbayerische Technische Hochschule Regensburg (OTH.R), Regensburg, Germany
ABSTRACT: District heating cost reduction efforts are amongst others focusing on the use of a new backfill
material. The OTH.R is investigating this material in regard to the engineering of buried district heating pipes.
The main focus of this soil-structure interaction problem is axial bedding to reduce thermal induced elongations.
The backfill material is characterized and an axial non-linear spring-based calculation method is described.
Centerpiece of the method is the contact-working-resistance-line (CWRL). A new Rod-Shear-Test based testing
device Re-SIST to investigate the CWRL for the new backfill material is presented. The investigated CWRL is
described and compared to the CWRL for sand, which is the conventional backfill material. Solutions using the
spring-based calculation method are presented and a comparative example of calculation shows the difference
in axial bedding when the new backfill material is used.
1 INTRODUCTION 1 a general axial non-linear spring-based calculation

method to solve the axial bedding problem.
District heating (DH) delivers sustainable energy, 2 a new Rod-Shear-Test-based testing device Re-
especially when it is based on combined heat and SIST in order to investigate the CWRLs of TFSB.
power generation, which considerably reduces car- 3 the deduced trilinear CWRL for TFSB as the result
bon dioxide emissions. Cost reduction efforts in order of an extensive series of tests using the Re-SIST.
to increase competitiveness against other heat supply 4 the adoption of the axial non-linear spring-based
technologies are amongst others focusing on the use calculation method, incorporating the identified
of a new backfill material. The material uses the spoil CWRL and comparative calculations.
or aggregate and cement, optionally bentonite and in
All information is related to the first DHP warm-up
any case surplus water, requiring no compaction, sim-
stage.
ilar to Controlled Low Strength Material (CLSM). The
material will subsequently be referred to as Temporary
Flowable, Self-compacting Backfill material (TFSB),
which would be an appropriate translation of the 2 CHARACTERISATION OF BACKFILL
German umbrella term Zeitweise selbstverdichtende, MATERIAL
fließfähige Verfüllbaustoffe (ZFSV).
The main focus of the engineering of district heating If sand is used, the required grain size range is from 0
pipes (DHP) is axial bedding by means of the back- to 2 mm. No fines are allowed and the maximum grain
fill material with regard to the reduction and, finally, size (oversize particle) is limited to 4 mm (Wagner
compensation of thermal induced elongations. State of et al. 2013).
the art is the use of non-cohesive soils, mainly sand, TFSB consist of spoil or aggregate and cement,
using a(n) (elastic-)plastic interface friction approach optionally bentonite and in any case surplus water.
for engineering calculations. The centrepiece of cal- Such material mixtures are in a somewhat similar form
culation methods involving interface contact is the not unknown if one thinks of slurry walls. One sub-
contact-resistance-working-line (CRWL) of the mate- stantial difference is the amount of added cement and
rials in contact. Little is known with regard to the bentonite, which usually ranges significantly lower
axial bedding and corresponding calculation basics between 25 kg m−3 –50 kg m−3 . Another fact is that
and methods when using TFSB, especially nothing no bentonite slurry is used. All solid matter is mixed
about the CWRL. This is hindering the usage, which (immediately) before the addition of water. Table 1
is, however, for the above mentioned reasons, in (high) shows the composition of one of the TFSB used in the
demand by public and private sector utilities. Similar tests. According to the USCS the spoil raw material is
problems and interests exist in regard to high-voltage classified as SC.
underground power cables, as these are also subject to Tests were also conducted using a prefabricated dry
thermal induced elongations. The research described mortar consisting of ground limestone and cement
in this paper therefore deals with only. Both materials generally show the same CWRL,
105
Table 1. Composition of one TFSB used in the tests.
Quantity
Component kg m−3
Spoil SC (dry) 1308

Bentonite 40
Cement 38
Mixing water 479
Figure 2. Superposition of strains and displacements and

resulting strains and displacements.
where εT (x) = thermal strain at coordinate x of a rod

of length l, αT = linear coefficient of thermal expan-
sion and T = change in temperature from the current
state.
where uT (x) = displacement due to thermal strain at

coordinate x and l(x) = length of rod at coordinate x.
If the rod surface is in (full) contact with another
material, thermal strains will induce (contact) shear
stresses. These shear stresses result from friction and
adhesion if present. The induced shear stresses τ(x)
generate mechanical strains εm (x) in the opposite
direction of the thermal strains, hindering these and
Figure 1. Principle of the wall-effect.
generating normal compressive forces in the rod. This
indicating that the addition of cement is one of the process can be described as axial bedding. If the rod
decisive factors. This shows in the process of setting, is long enough, finally a state along the rod will be
solidifying and curing of the material once it has been reached for which εT (x) = − εm (x) is valid. At this
poured in place. and smaller coordinates no more elongation is hap-
The boundary surface of TFSB to other materi- pening and the rod will be at rest. εT (x) and εm (x)
als is always without observable grains. This can be can only be calculated separately from each other,
explained taking concrete technological findings into if τ(x) is at a given constant value and not dependent
account: At contact surfaces to other materials the on the displacement between the rod surface and the
powder suspension prevents direct contact between the contact material. Figure 2 gives an idea of the super-
bigger grains and the surface of the other material. This position of strains and displacements and the resulting
is known as the so-called wall-effect, leading further- strains ε(x) = εT (x) + εm (x) and displacements u(x) =
more to deviating material composition and abilities at uT (x) + um (x) if τ(x) is at a given constant value.
the interface. (Geisenhanslüke 2008, Deutscher Beton If this is not the case, the opposing strains cannot
und Bautechnik-Verein e.V. 2004) Cement and ben- be determined separately as εT (x) influences εm (x) and
tonite are part of the powder (grain ≤ 0,125 mm), vice versa. In this case, starting from the normal force
respectively powder suspension. Figure 1 gives an in the rod, we have
illustration. The application of this concept regarding
TFSB seems to be justified and has to be taken into
account regarding interface behaviour.
where N (x) = normal force in the rod at coordinate
3 THERMAL ELONGATION AND (x), E = Young’s modulus of rod, A = cross-sectional
AXIAL BEDDING area of rod and u (x) = first derivation of resulting
displacement u(x).
Unhindered thermal elongation of a rod and the cor- Figure 3 shows the balance of forces if we take a
responding displacements can be calculated using look at a differential rod element and yields Equation 4.
equations 1 and 2. For symmetrical reasons only one
half of the rod of total length L = 2 × l needs to be
considered.
with U = circumference of rod. Derivating Equation

3 and combining it with Equation 4 one gets the
106
Figure 4. Experimental interface test setup using a Simple
Shear device. (Weidlich 2008, slightly modified)
4 TESTING DEVICE RE-SIST
There are several laboratory devices on hand in order

to investigate CWRLs like Direct Shear Test devices,
Figure 3. Balance of forces for a differential rod element. Simple Shear Test Devices, Ring Torsion Devices
and Rod Shear Test devices. An overview is given,
for example, by Mooney (1998) and Musharaf &
governing homogenous linear differential equation of Arumugam (1995).
second order with constant coefficients. For natural soils Direct Shear and Simple Shear Test
devices are commonly used, as they are available in
most testing laboratories and the necessary modifica-
with u = second derivation of resulting displacement tions for interface testing can be done at a moderate
u. Equation 5 was in general imparted before. For effort. Figure 4 shows a typical experimental setup
example in the context of friction pile engineering using an altered Simple Shear device.
(Alpan 1978). Assuming that E, A and U are of con- A large number of tests can rapidly performed
stant value along the rod, the solution is governed by when natural soils are tested against other materials, as
τ and its displacement dependent relationship of acti- contact-time-dependent effects are negligible in most
vation. This can be described by means of a CWRL, cases or waiting times are rather short. When it comes
giving the value of τ for a certain (interfacial) displace- down to TFSB because of the wall-effect the powder
ment w between the rod and the utilized other material. suspension is the leading material constituent at the
Figure 6 shows amongst others a typical CWRL for interface. As setting, solidifying and curing are caused
sand (Beilke 1993) in contact with the surface of a by the components of and processes within the pow-
DHP, representing the rod in this case. der suspension, interface behaviour is most likely to
One may note that the interfacial displacement is be time-dependent to a large extent, too. This implies
named w(x) to point out that it has nothing to do with at least two boundary conditions in regard to interface
the resulting displacements u(x) or the displacements tests with TFSB:
um (x) due to mechanical strains. Additional explana- 1 long-term time-dependent testing needs to be done
tions will be given in paragraph 7. In this context one 2 contact between TFSB and the other material needs
has to be clear on the fact that the CWRL must not to be established permanently before setting of the
include any strains of the rod to reflect the true interfa- TFSB starts
cial displacement-force behaviour as this behaviour is
decisive on the activation of mechanical strains εm (x) This would require a large number of modified
by thermal strains εT (x). The CRWL must therefore laboratory testing devices, which are usually not on
be known to solve the problem analytically, but its hand for financial reasons. But even then problems
displacement-force behavior cannot directly be found like drying, setup of a shearing gap, reduction of
in any given final force-displacement state of the axi- shearing area would have to be solved. In order to over-
ally bedded rod, which comprises displacements of the come these shortcomings a new Rod-Shear-Test-based
rod from different causes. If we furthermore assume testing device Re-SIST (Regensburger Stab-I nterface-
that displacements in the soil material in contact with Scher-T est) has been designed and realised. One of the
the rod are strictly limited to the interface surface, the major advantages of the Re-SIST is that it serves as a
material around the rod can be represented by a soil separate storable testing container which is placed in
spring and the CWRL is then the corresponding non- the load frame only on the day of the test.
linear load deflexion curve. This simplification can be The Re-SIST device consists of a cylindrical pipe
regarded as valid in the case of TFSB due to the wall with two square end-plates (all made of rigid-PVC)
effect and the mechanical cementation of the solid which are screwed together by means of threaded
matter by and within the powder suspension. All in bars at the corners of the plates. The plates contain
all, one may speak of an axial non-linear spring-based a centric hole with auxiliary milled slots in order to
calculation method. install and hold the rod (simulating the DHP) in place.
107
the TFSB. The required force and displacement of
the rod are logged. In order to determine the post-
fracture interface behaviour tests are performed using
displacement-control at 0.05 mm min−1 . The selected
velocity tends to minimize the influence of curing
during the tests. All tests and post-testing inspec-
tions showed localization directly at the interface and
no changes in the structure of the TFSB. The prime
explanations are the powder suspension as the leading
material constituent at the interface, dominating inter-
face behaviour and the cementation of the solid matter
in general. As localization is concentrated directly
around the interface significant changes in the pore
water distribution are regarded as insignificant in
continuation of the underlying considerations.
The scope of conducted tests comprised mostly
time-series (2 samples at 7 d, 14 d, 28 d, 56 d and
112 d) at a selected radial pressure of 50 kN m−2 .
Additional tests at a sample age of 28 d but at other
selected radial pressures of 12.5 kN m−2 , 25 kN m−2
and 100 kN m−2 complemented the time-series. Fur-
thermore, creep-tests have been executed and at the
moment cyclic testing is being done. In the interest of
brevity a detailed discussion cannot be included here
Figure 5. Re-SIST (Regensburger Stab-I nterface-Scher- and the presentation of the results will be limited to
T est) device. general findings.
The holes are a bit larger in diameter than the rod

used in order to allow the unhindered development 6 TRILINEAR CWRL FOR TFSB
of a shear plane to a certain extent. Centrical fixa-
tion of the rod is ensured by additional measures. The Presentation and evaluation of test results yield the
inside of the plates are covered with a smooth drainage following major findings, partly in comparison to sand
fleece to reduce plate end friction and allow dewater- (Figs 6, 7):
ing by suitable water outlets in the plates. In order to
apply controllable radial pressure, simulating the lay- 1 TFSB show a (idealized) trilinear and not bilinear
ing depth, a custom-built latex membrane is placed CWRL like sand, including elastic, softening and
inside the pipe and clamped between the pipe and the plastic regions.
plates before screwing them together. 2 The peak elastic contact resistance stress is several
To prevent arching an additional fleece is placed times the plastic contact resistance stress, time-
between the membrane and the pipe. As the surface dependent and can be described by means of a
of the objective DHP is made of polyethylene pur- logarithmic best-fit curve.
chasable polyethylene rods with a non-ferrous metal 3 Plastic contact resistance stress is time-dependent,
core to minimize rod strains by means of high axial but only to a small extent.
stiffness were used. Rods with a stiffer metal core 4 Contact resistance stress is displacement-dependent
or DHPs with suitable diameter were not available. for the elastic and softening regions.
Immediately after mixing, the container is filled with 5 Contact resistance stress is not displacement-
TFSB and the filling holes in the top plate are sealed. dependent for the plastic region.
After 24 h of setting radial pressure is applied by means 6 The peak elastic contact resistance stress displace-
of compressed air in the annular space between the ment is significantly smaller than for sand.
latex membrane and the inner pipe surface. The water The reasons for these findings will be discussed and
outlets are opened simultaneously. Figure 5 shows a explained later in another paper (Wagner, in prep.).
schematically sketch of the device according to the The area below the CWRL indicates the work poten-
preceding description. tial of the CWRL. There is a difference in the areas
when comparing the two CWRLs up to a certain dis-
placement. This means that one of the CWRLs does
5 TEST PROCEDURE, IMPORTANT more work that is conveys more energy on the rod for
OBSERVATIONS AND SCOPE OF TESTS the same displacement, due to higher contact resis-
tance stresses. Taking a look at Figure 6 it becomes
On the day of the test, the pressurized Re-SIST clear that TFSB show a significantly additional work
device is placed into a conventional load frame with potential compared to sand, indicated by the shaded
a suitable base spacer and the rod is pushed through area. In regard to thermal elongation and axial bedding
108
Figure 6. Idealized CWRLs for TFSB and sand.
Figure 7. Temporal development of τel,TFSB and τpl,TFSB with
as described in paragraph 2 it becomes clear that logarithmic respectively linear best-fit curve.
thermal strains εT (x) will therefore generate higher
opposing mechanical strains εm (x) by inducing higher
shear stresses τ(x). For the objective first DHP warm- the axially bedded system the distribution of τ(x) along
up stage this means that a rod of equal length will the rod will be nonlinear: Combining Equations 5 and
sooner be at rest and show lower displacements when 7 and solving the resulting Equation will lead to a non-
TFSB are used as backfill material instead of sand. As linear function describing u(x). Therefore, the elastic
τ(x) is displacement-dependent εT (x) and εm (x) can- linear region of the CWRL can be stretched over a long
not be calculated separately from each other and the section of the rod.
solution has to be determined using Equation 5, which For each region of the CWRL a different general
will be done in the next paragraph. solution to Equation 5, depending on the term describ-
ing τ(x), exists in the first instance, also dividing
the rod in different regions. The individual solu-
7 AXIAL NON-LINEAR SPRING BASED tions to the problem ensuring the required continuity
CALCULATION METHOD of displacements and forces along the rod are then
achieved by correct constants of integrations. These
Equation 5 is the homogenous linear differential equa- are governed by Dirichlet- and Neumann-boundary
tion of second order with constant coefficients and has conditions including the coupling points. Dirichlet-
been presented by Alpan (1978) and utilized for DH by boundary conditions depend on the CWRL and this
Beilke (1993) in regard to sand as backfill material. In is where the values of w(x) like wel at the region tran-
regard to the different regions of the CWRL one has to sition points of the CWRL become relevant. In this
distinguish between τ(x) = f (u(x)) and τ(x) = f (u(x)). paper only the solution for a simplified CWRL for
Please do note that here u and not w as the overall dis- TFSB is presented as shown in Figure 9. The full solu-
placement is now relevant as Equation 5 is the solution tion will be presented later in another paper (Wagner,
to the axial bedding problem. This can be explained as in prep.).
follows: Figure 8 shows the equivalent statical system.
If we take a look at the elastic linear region of the The T -dependent displacement zero-point (εT (x) =
CWRL (Fig. 6), the soil spring rate c is described by −εm (x)) along the rod can be represented by a fixa-
Equation 6: tion. Vertical, horizontally moveable supports are not
depicted.
With Equations 8 and 9
For a certain overall displacement u(x) ≤ wel the

effective shear stress τ(x) can be calculated using
Equation 7
with a = decay ratio of axial stiffness.
and that is the term describing τ(x) = f (u(x)) in regard

to Equation 5. The image is that even though the with e = ratio of shear stresses the solution to Equa-
CWRL clearly shows a linear activation of τ(x) within tions 5 and 3 are as follows for the elastic region of the
109
“plastic” length and passed in the “elastic” length at
the coupling point. This “plastic” compensation is rep-
resented by the term a2 × e × wel × lpl in Equations 10
and 11. This concept can easily be understood if εm (x),
along a rod induced by a normal force at one end of
the rod, is considered: εm (x) is then a constant value,
just like εT (x) but in the above case of contrariwise
direction. The required plastic length is determined
by the fact that at the fixation εT (x) = −εm (x) has to
be valid. As εT (x) increases with increasing T lpl
increases, too, as we are dealing with a linear self-
adjusting system. This is a main difference to the
solution proposed by Beilke (1993), who assumed that
the “elastic” length has to be determined iteratively.
Other differences are
– the explicit use of shear stress and circumference
instead of already circumferential integrated forces.
Equation 8 gives a term that represents a stiffness
decay ratio as the diameter dependent stiffness of
the soil spring and the rod are contrasted. a is a
Figure 8. Equivalent statical system (state: T > governing factor if we look at Equations 10 to 13.
Tel,grenz ). – the use of the shear stress ratio e (Equation 9), which
yields a solution that covers not only a continuous
CWRL (Equations 10 and 11): elastic-plastic CWRL but all possible elastic-non-
continuous-plastic CWRLs.
Furthermore, Equations 10 and 13 show that the
change in temperature from the current state T is
the movement and normal force provoking cause. If
T = 0 there are no resulting displacements or nor-
mal forces as wel and terms including wel only become
relevant if T > Tel,grenz which would be contradic-
where lel,grenz = length of the rod along which u(x) ≤ tory to the precondition T = 0. In contrast to piles
wel and lpl = length of the rod along which u(x) > wel . no external loads are required to provoke movement,
The solutions to Equation 5 and 3 are as follows for displacements and normal forces.
the plastic region of the CWRL (Equations 12 and 13):
8 COMPARATIVE EXAMPLE OF
CALCULATION
Comparative calculations using sand respectively a

TFSB (Table 1) as backfill material have been done
lel,grenz is a fixed value and decribes the “elastic” length for a rectilinear DHP made up of a plastic jacket
of the rod along which the displacement u(x), which compound pipe DN 40/125 representing the rod. The
also represents the soil spring deflexion, is smaller or plastic jacket compound pipe is the most common
equal to wel . lel,grenz can be determined using Equations DHP in Germany (Espig 2012). The outer diameter
14 and 15. of the plastic jacket is 125 mm and 48.3 mm for the
steel medium pipe with a wall thickness of 2.6 mm.
The space between the medium pipe and the plastic
jacket is filled with polyurethane resin mainly for the
purpose of heat insulation. The stiffness (E and A)
of the pipe is dominated by the medium pipe. Shear
strains between the different components of the plas-
where Tel,grenz = change in temperature for which tic jacket compound pipe are usually neglected. The
the displacement u at the free to move end of the rod laying depth has been assumed to be 2.6 m equaling
will exactly be wel . a vertical overburden stress of approximately 50 kN
This means that once Tel,grenz is exceeded the m−2 . The CWRLS for the backfill materials are shown
“elastic” length of the rod is “frozen” and simply slips in Figure 9 (standardized with sand values) and have
to the left as the “plastic” length increases with increas- been determined for sand by the relevant rules and
ing T . The additional constant thermal strain along standards (AGFW 2007, DIN e.V. 2010) respectively
the “elastic” length resulting from T > Tel,grenz is by TFSB Re-SIST tests, assuming the first warm-up
compensated by the normal force generated in the stage is taking place 28 d after backfilling.
110
Figure 9. Calculation example: CWRLs for TFSB (simpli-
fied) and sand; standardized with sand values.
Figure 11. Normal forces along DHP (= rod) for TFSB and
sand, T = 130K > Tel,grenz .
As theoretically discussed in paragraph 7 the calcu-

lations show that a significantly smaller length of the
DHP shows displacements when the backfill material
is TFSB. The DHP is sooner at rest. The displacements
are furthermore smaller compared to those when sand
is used, but the generated normal forces are equal (as
the condition of the equilibrium is full compensation
of thermal elongation). The elastic length for TFSB
is noticeably smaller but the normal force generated
along the elastic length is higher. This is due to the
fact that although compared to the CWRL for sand the
elastic displacement wel to reach the higher maximum
elastic shear stress τel is smaller higher mechanical
strains εm are generated within the elastic length. This
Figure 10. Displacements along DHP (= rod) for TFSB and
illustrates the higher elastic work potential of TFSB
sand, T = 130K > Tel,grenz .
compared to sand, resulting in a higher Tel,grenz .
Therefore and due to the higher plastic shear stress, the
required plastic length is smaller compared to sand.
The CWRL for the TFSB includes necessary adjust- The overall view clearly shows that the new back-
ments in regard to the strains of the testing rod (see fill material TFSB enhances axial bedding of district
paragraphs 3 and 4) by means of a developed appropri- heating pipes. This is done by reducing the pipe
ate algorithm. The CWRL for the TFSB has in addition length showing displacements and by reducing these
been simplified neglecting the softening region as displacements. This reduces the need for additional
already indicated in paragraph 7. To make up for technical measures like expansion joints.
this neglect and in regard to findings of additional
large-scale test, the plastic value of τ has been raised.
Starting from a stress and strain-free state and the same 9 CONCLUSIONS
system temperature the required pipe lengths, the dis-
placements and the normal forces along the pipe for The main focus of District heating engineering is axial
T = 130 K have been calculated. Practical questions bedding to reduce and compensate thermal induced
like limitations of steel stresses are not considered elongations. For cost reducing Temporary Flowable,
within the example. Self-compacting Backfill material (TFSB) the corre-
Figures 10 (displacements) and 11 (normal forces) sponding state of knowledge is small.
show the calculation results. A possible way to calculate displacements and nor-
In the subsequent discussion of results the term mal forces of axially bedded district heating pipes
“elastic” specifies that shear stresses from the elas- (DHP) for the first warm-up stage is an axial non-linear
tic region of the CWRLS are relevant and the term spring-based calculation method based on a differen-
“plastic” that the shear stresses from the plastic region tial equation. The soil spring represents the interface
of the CWRLs are relevant. behaviour between the DHP and the backfill material
111
and is described by the contact-resistance-working- Alpan, I. 1978. Das Last-Setzungsverhalten des Einzelpfahls.
line (CRWL). The CRWL has to be investigated Bauingenieur 53: 293–298.
by interface tests. For several reasons and bound- Beilke, O. 1993. Interaktion des Bauwerks “Fern-
ary conditions standard common interface tests are wärmeleitung – Bettungsmaterial”. Hannover: n.p.
Deutscher Beton und Bautechnik-Verein e.V. (ed.) 2004.
less suitable for TFSB. A new Rod-Shear-Test-based DBV-Sachstandsbericht. Betonoberfläche – Betonrand-
testing device Re-SIST (Regensburger Stab-I nterface- zone – Fassung November 1996, redaktionell überarbeitet
Scher-T est) allows the investigation and identification 2004. Berlin: n.p.
of the trilinear (linear, softening, plastic) CWRL for DIN e.V. (ed.) 2010. DIN EN 13941. Auslegung und Instal-
TFSB. Based on the CWRL the solution to the differ- lation von werkmäßig gedämmten Verbundmantelrohren
ential equation can be determined using appropriate für die Fernwärme; Deutsche und Englische Fassung
Dirichlet- and Neumann-boundary conditions. Com- EN13941;2009+A1:2010. Berlin: Beuth.
parative calculations show that the new backfill mate- Espig, F. 2012. Schadensstatistik KMR 2010 des AGFW.
rial TFSB enhances axial bedding of district heating Euroheat & Power 41 (5): 32–35.
Geisenhanslüke, C. 2008. Einfluss der Granulometrie von
pipes due to its different CWRL compared to sand Feinstoffen auf die Rheologie von Feinstoffleimen. Kassel:
as the standard backfill material. The CWRL shows kassel university press.
a significantly additional work potential. The main Mooney, D.T. 1998. Experimental and numerical study of
effects are the reduction of the pipe length showing dis- the Rod Shear Test for determining steel-sand interface
placements and the reduction of these displacements, behavior. Arizona: University of Arizona.
whereas the same normal forces are generated. Musharaf, Z. & Arumugam, A. 1995. Soil-Structure Inter-
face: Experimental Aspects. In A.P.S. Selvadurai, & M.J.
Boulon, (eds), Studies in Applied Mechanics 42: Mechan-
ACKNOWLEDGEMENTS ics of Geomaterial Interfaces: 127–145. Amsterdam:
Elsevier.
This research was funded by the Federal Ministry for Wagner, B. et al. 2013. Einsatz fließfähiger Verfüll-
Economic Affairs and Technology, support codes FKZ baustoffe zur KMR-Verlegung. Euroheat & Power 42 (9):
03ET1063B and FKZ 03T1063D, whose support is 54–56.
greatly appreciated. Wagner, B. in prep. Ein Beitrag zur axialen Bettung von
warmgehenden Leitungen, speziell Kunststoffverbund-
mantelrohren des Fernwärmeleitungsbaus, in Zeitweise
REFERENCES fließfähigen, selbstverdichtendenVerfüllbaustoffen (work-
ing title).
AGFW (ed.) 2007. Arbeitsblatt FW 401 – Verlegung und Weidlich, I. 2008. Untersuchungen zur Reibung an zyk-
Statik von Kunststoffmantelrohren (KMR) für Fern- lisch axial verschobenen erdverlegten Rohren. Hannover:
wärmenetze. Frankfurt: AGFW. IGBE.
112
Assessment of geothermal snow-melting system used

in cold region area
I.-Hsuan Ho & Mark Dickson

Harold Hamm School of Geology and Geological Engineering, University of North Dakota, ND, USA
ABSTRACT: The improvement of pavement heating system has a great impact on transportation in cold
regions, such as highway and airport pavements. In the USA, Federal Aviation Administration (FAA) invested
money in developing new technologies in pavement heating system for runways and taxiways. Geothermal can be
one of the efficient energy resource for these engineering needs. The observed advantages of geothermal heating
system include: (1) Energy is renewable and reliable; (2) avoid using the chemical treatment; (3) increase the snow
removal safety; and (4) lower the CO2 emission. Moreover, by providing a sustainable temperature higher than
32◦ F, the maintenance cost can be reduced massively due to the frozen-thawed cycles. A few applications have
been demonstrated in some countries such as USA, Japan, Switzerland and Poland. However, the effectiveness
and limitations of the snow melting system depends on climatic conditions, properties of geomaterials and the
heat transfer mechanisms between the heat source and the pavement surface. In the state of North Dakota, the
annual snowfall ranges from 26 to 38 inches and the temperatures can be below −20◦ F for several months. If
the limitations can be overcome to increase the effectiveness of snow melting system using geothermal energy,
the technique can also be adopted in these areas with the similar climatic conditions. In this paper, three cases
of snowmelting design for pavements were reviewed, the snow melting equations using geothermal energy were
revisited and the limitations on snow melting system design in ND are discussed. The strategy to increase the
effectiveness of a snow melting system using geothermal energy is also addressed.
Keywords: Pavement heating system, geothermal energy, CO2 emission, heat transfer mechanisms, snow
melting system
1 INTRODUCTION The snow-melting methods have been very attrac-

tive to researchers. Several types of heating or snow-
The least thing people in cold region would do is to removal methods have been used and developed in
drive on the highways. The presence of snow and ice the past decades, such as vehicle removal, chemi-
on the roads is more than just a nuisance, but a genuine cal removal, electric snow removal, hydronic heating
safety concern. There are over 200,000 car accidents system water spray and geothermal heating system.
every year caused by snow and sleet every year in the Using salt and other chemical treatments to prevent
United States. This is not only costly, but causes hun- ice buildup requires the recurring costs of the chem-
dreds of deaths every year. In order to compensate icals, as well as maintaining the trucks that distribute
for the dangerous weather conditions, drivers must the chemicals. Salts such as NaCl, MgCl2 , and various
reduce their speed, which can cause traffic conges- glycols are often used (Ekolist, 2001). The frequent use
tion and millions of dollars in reduced productivity. of these chemicals can weaken and damage asphalt and
Snow buildup on airport runways can make an airport concrete over time, decreasing the longevity of roads
inoperable, creating hazardous landing conditions and and pavement (Melcher, 2001). After the snow melts,
costly delays. This is why so many public and pri- the chemical treatment runoff can cause local envi-
vate organizations invest in snow removal and snow ronmental damage, harming plant life and polluting
melting systems. These systems remove snow and ice local water systems. Expensive electric coils or water
buildup by chemical treatment, ploughing, or appli- boiler heating systems can also be used to heat the
cation of heat. Unfortunately, many of these systems concrete and melt snow. These systems can respond
can be expensive, inefficient, harmful to the envi- fairly quickly to snow fall, but are expensive to install,
ronment, or too slow to respond to the snow in an have high recurring energy expenses, can be very inef-
appropriate amount of time. Moreover, the frozen- ficient in colder climates, and consume large amounts
thawed cycle to the pavement materials increase the of either electricity/fuel which create additional CO2
cost for maintenance. emissions.
113
Global climate change (NOAA, 2006) is an increas- several limitations. However, it still can provide the
ing concern, so companies and individuals are increas- good estimation of heat requirement. The limitations
ingly turning towards renewable energy such as can be overcome when a finite element analysis is
geothermal. In comparison to electric or boiler based introduced.
systems, geothermal systems use about 2–10% of the
energy. (3) The Kyoto Protocol target CO2 reduc- 2.1 Transient weather conditions
tion is 4.2% (Olivier et al., 2011) over the next 5
Most snowfall events happen over the course of sev-
years, so this system more than exceeds that require-
eral hours, and during that time precipitation rates,
ment. Meanwhile, some geothermal applications have
ambient air temperature, humidity, wind speed, and
been conducted in Japan (Nagai et al. 2013), Poland
solar radiation can change enough to affect a systems
(Zwarycz, 2002) and Switzerland…etc.
performance. An ideal system would melt snow and
In the state of North Dakota, the average temper-
ice as soon as it forms on the surface, using the least
ature is 19◦ F for four months data (Nov. to Feb.)
amount of energy as possible. Some systems use mois-
during winter. The available fluid temperature cir-
ture and temperature sensors to automatically activate
culated to melt the snow was estimated to be 45◦ F
in response to freezing conditions and precipitation,
(Gosnold, 2012). The energy required to melt the
while others systems are on a timer or are manually
snow is 25% compared to Alaska (Rees et al., 2002;
activated.
Ho and Dickson, 2015). The weather conditions are
When a snowfall event begins, the rate of snow-
unique compared among other states in the U.S. Also,
fall usually increases until the peak snowfall rate is
North Dakota’s shallow depth geothermal resources
reached.A system can be designed so that it melts snow
are abundant, especially in the western region of North
as it falls by heating the pavement to prevent snow from
Dakota. Several studies have focused on geothermal
forming on the surface. Preventing snow from falling
snow melting systems, however, very little research on
has several advantages; it insures that driving condi-
snow-melting systems conducted in this area, with its
tions will be safer throughout the snowfall event, and
unique climate and geological environments, particu-
it prevents the snow from creating an insulating layer
larly using low-temperature heating fluids. In order to
that interferes with the melting process. Heavier rates
overcome the limitations of using geothermal energy
of snowfall, lower temperatures, and several other fac-
directly or through heat pumps to melt snow and bal-
tors affect the amount of energy to most effectively
ance the high heat requirement, a detailed parametric
keep the surface snow and ice free. The water content
study is needed prior to a further pilot experimental
of the snow also affects the efficiency of the system, as
study.
not all inches of snowfall require equivalent amounts
of thermal energy to melt them.
2 PARAMETRIC STUDY
3 CLIMATIC CONDITIONS OF
In order to design a geothermal snow melting system, NORTH DAKOTA
several conditions must be considered. Previous stud-
ies on snow melting systems have been performed, The climatic conditions are the most important influ-
many of which propose designs based on steady state ential factor when designing a snow melting system.
conditions. Variables such as the snowfall rate, ambi- The air temperature, rate of snow fall and frequency
ent air temperature, and wind speed, are often derived of snow fall, moisture content of snow, wind speed,
from weather data compiled over several years, and and amount of sunlight all control the rate at which
usually these values represent a typical or upper value snow will be melted. The Data for the climatic condi-
one would expect during a snow fall event. The equations in North Dakota used for snow melting system
tion below by Chapman and Katunich (1956) assumes design is from the National Oceanic and Atmospheric
steady state conditions and is used to calculate the Administration. The data used was averaged over the
amount of heat needed per square foot to adequately last 15 years from 1999 to 2014. The snow melting
melt snow. system needs an insight of climatic conditions over
Chapman and Katunich Equation is derived in the the months of November to February the next year.
following form: These winter months typically have the greatest snow-
fall and lowest temperatures over the year, where the
snow is likely to stick to pavement and affect trans-
portation safety. Most of the data was collected from
where qo = total heat flux per unit area of the sur- the National Oceanic and Atmospheric Administration
face, Btu/h · ft2 (W/m2 ); qs = total sensible heat flux, and the World Data Center for Meteorology.
Btu/h · ft2 (W/m2 ); qm = melting load, Btu/h · ft2 Figure 1 presents the averaged fifteen-year of air
(W/m2 ); Ar = snow-free area ratio; qh = sum of the temperatures in cold months from 1999 to 2014. The
convection and radiation losses, Btu/h ·ft2 (W/m2 ); and coldest month in a year is in January and the snow usu-
qe = evaporative losses, Btu/h · ft2 (W/m2 ). ally starts from November. The snow melting design
This equation can be useful for roughly approxi- should take account of various climatic conditions
mating the heat load required to melt snow, but has to optimize the use of available geothermal energy.
114
Figure 1. Average temperature during cold months in North
Dakota, USA.
Apparently, the worst climatic conditions should not

be the first priority in the snow-melting system design.
However, the principles of a design should consider a Figure 2. Subsurface temperature distribution.
system which can provide the most convenient period
in a year.
The other significant climatic factors such as wind
speed, snowfalls rate, humidity and moisture content
also need to be considered properly in a design. As for
the magnitude used, a guideline should be proposed
and it’s basically depends on the significance of the
infrastructure.
4 GEOTHERMAL RESOURCE
The subsurface temperature data shown in Figure 2

was collected from a well near Grand Forks area in
North Dakota, USA, where temperature data was gath-
ered from November 1st 2011 to February 13th 2012. Figure 3. Geothermal resource distribution in the US (Refer
Although heat measurements have been taken from to DOE, USA).
several different wells throughout North Dakota, How-
ever, the data is very limited, the data gathered through- seasonal energy storage system is recommended to
out the year, allowing for a seasonal comparison of optimize the use of shallow geothermal resource.
underground temperatures. In the United States, over 2000 wells of heat flow
According to the data, the ground temperature sta- data have been released. The thermal energy distribu-
bilizes at about 7 degrees C at a depth of 5–7 meters, tion based on milliwatts per square meter is shown on
or 16–23 feet. Although there is a slight drop in ground Figure 3. In the US, the abundant geothermal resources
temperature during the coldest months, the overall the are mainly concentrated in the Western United States.
temperature is still much more stable than atmospheric On the contrary, the available geothermal resource can
weather conditions. This means that the shallow under- be used is relatively low in the rest of area, especially
ground temperature could be used as a heat source in cold regions. The available shallow geothermal
for pavement heating during cold season. The amount resources become significant if a snow melting system
of energy that can be extracted out of the earth is for the pavement has to be designed.
limited by the temperature difference between the
underground and the surface.
Another important thing to note about shallow 5 SNOW MELTING SYSTEM DESIGN
geothermal energy is that it is a renewable resource;
not an unlimited one. If a greater amount of energy Based on the climatic conditions, the available
is pulled out of the ground by the system than can geothermal resource and the subsurface temperature
be replaced, the reservoir can be depleted. This is in the state of North Dakota, USA. The examples
why supplementing the system with a heat pump or of snow-melting system are presented. The design
115
considers the climatic conditions such as air tempera- where q is the heat flow rate; k is the thermal conduc-
ture, wind speed, snowfall rate, humidity and moisture tivity; A is the cross-sectional area; d is the thickness
content in heat requirement calculations and then of the material; T1 is subsurface temperature; and T2
balanced by the available geothermal heat resource. is the surface temperature.
Chapman’s equation (1952, 1956) was used in cal- The pipe designs below were modeled in Excel
culating the energy requirements based on climatic using the one-dimensional heat flow equation
data in North Dakota. Outlined below are the designs described above. Three scenarios are estimated by
for the numerical model of a geothermal snow melt- assuming the ratio of spacing (S) to diameter of pipes
ing system. It shows the weather conditions and heat (D) equal to 2, 4 and 6, respectively. When assuming
requirements according to the Chapman and Katunich heat flow rate per pipe is 115.27 BTU/hr-ft2 , thermal
Equation 1952. This equation can be used to estimate conductivity is 5.26 BTU/hr-ft2 , cross-sectional area,
the overall heat needed in terms of Btu/h-ft2 . per sec- 0.1667 ft2 , subsurface temperature, T1 is 40◦ C and
ond in order to melt snow at a specified snowfall rate. surface temperature, T2 is 0◦ C, the heat requirements
for pipe designs, S/D equal to 2, 3 and 6 are 230, 345
5.1 Pipe design for hydronic/geothermal snow and 690 BTU/hr-ft2 , respectively.
melting system
The pipe designs described below were modeled using
spreadsheet calculations in Excel. The equation is 6 DISCUSSION
the one-dimensional heat flow equation derived by
Chapman (1957). After calculating the heat requirements based on avail-
The obvious advantage of having a denser pipe spac- able weather data, the system must be capable of
ing is more heat per square foot will be supplied, but it supplying at least 309 BTU per hour, per square foot
increases installation costs and pulls more water and during peak snow fall times, and be able to efficiently
heat out of the geothermal reservoir. On the other hand, supply 200–250 BTU per hour, per square foot dur-
having fewer pipes requires warmer water, meaning the ing the regular snowfall events. The lower the water
heat pump must apply additional heat to compensate, temperature that can be used, the greater the energy
making the system less energy efficient. efficiency of the system. At 40 degrees Celsius, the
Based on 1-in diameter of pipes, the spacings geothermal system is supplying approximately 30% of
between pipes were assumed to be 6, 4 and 2 inches, the systems thermal energy. This thermal energy sup-
respectively. Water temperature used in the calcula- ply increases to approximately 45-60% during lighter
tions is 40◦ C, as that is what most low temperature and more typical snowfall events.
systems utilize. The low temperature is 0 ◦ C or 33 ◦ F The spacing design of S/D=2.0 could compromise
as that is the required temperature to melt snow. The the structural strength of the pavement and is most
embedment of pipes is 3.25 inches below the surface; likely not realistic, and supplies far more heat than
some systems embedded pipes as shallow as 2 inches is necessary. On the contrary, the design of S/D=6
but that runs the risk of damaging the pipes. The ther- cannot make the snow-melting system meet the heat
mal conductivity was assumed to be 0.84 w/m-k which requirements and work efficiently, unless a higher tem-
converts to 5.26 Btu/hr-ft-F. A thermal conductivity for perature of water (> 40◦ C) is available. This would
asphalt and concrete often falls within these values. mean the heat pump is supplying such a large per-
The heat requirement calculations are based on centage of the overall heat energy that it would not be
Chapman’s equation (1952, 1956) and the class III practical to utilize a geothermal heat resource.
system recommended by American Society of Heating The ratio, S/D=4 is most likely can supply the heat
and Air-Conditioning Engineers (ASHRAE). against the worst scenario by using 40◦ C of water in
this area. This design also can work efficiently to melt
the snow under the normal snowfall condition in winter
5.2 Heat requirements season which required the heat ranging between 200–
The heat requirement for the air temperature, 0◦ F; 250 BTU per hour, per square foot most often.
wind speed, 15 m/s; free area ratio of snow, and rate of
snowfall, in/hr, is 309 BTU/hr-ft2 . This estimated heat
requirement is based on Class III recommended by 7 SUMMARY AND CONCLUSIONS
ASHRAE for the worst scenario. The usual target heat
requirement is mainly between 200 to 250 BTU/hr-ft2 The shallow geothermal energy is a form of sustain-
in North Dakota. able and renewable energy. The heat can help the cold
regions to melt the snow for pavements in winter.
5.3 Shallow geothermal resource Hence, a proper design is needed to optimize the use
The heat flow transfer mechanism includes three gov- of this renewable energy. Several conclusions can be
erning factors: conduction, convection, and radiation. made as follows:
The heat flow rate based on the following equation: (1) When designing a snow-melting system, a proper
design to consider the heat requirement can melt
the most events in winter is more practical.
116
(2) The geothermal heat should work with heat pump Region.” 1st International Conference for Geo-Energy
to optimize the pipes design under the pavement. and Geo-Environmental, GeGe 2015, Hongkong.
However, the more direct contribution from the Lund, J. W. (2000). Pavement snow melting. Geo-Heat Center
geothermal resource is the optimal design. Quarterly Bulletin, 21(2), 12–19.
Melcher, K. (2001): Winter road maintenance spreadings in
(3) The geothermal heating system for the pavement the Czech Republic and in EU countries.
not only can increase the safety for road users but Nagai, N., Miyamoto, S., Osawa, Y., Igarashi, S., Shibata,
can extend the life for the roads and reduce the K., & Takeuchi, M. (2013). Numerical simulation of snow
maintenance cost in cold regions. melting using geothermal energy assisted by heat stor-
age during seasons. Heat Transfer-Asian Research, 42(8),
ACKNOWLEDGMENT 724–744.
NOAA (2006): NCRFC Climate and Topography, National
Oceanic and Atmospheric Administration
This research is supported by NSF ND EPSCoR Olivier, J. G. J. et. al. (2011), Long-term trend in
project. global CO2 emissions; 2011 report (PDF), The Hague,
Netherlands: PBL Netherlands Environmental Assess-
REFERENCES ment Agency; Institute for Environment and Sustainabil-
ity (IES) of the European Commission’s Joint Research
Chapman, W. P., (1952). “Design of Snow Melting Systems, Centre (JRC), ISBN 978-90-78645-68-9 PBL publica-
Heating and Ventilating (April): 95 and (November): 88”. tion number 500253004. JRC Technical Note number
Chapman, W.P., and S. Katunich. (1956). Heat require- JRC65918.Rees, S. J., Spitler, J. D., & Xia, X. (2002).
ments of snow melting systems. ASHAE Transactions 62: Transient analysis of snow-melting system performance.
359–372. ASHRAE Transactions, 108, 406.
Ekolist, FHWA Report (2011): How do Weather Events Zarling, John P. “High Capacity Intersection Thaw System
Impact Roads, Federal Highway Administration. “School of Engineering University of Alaska Fairbanks,
Gosnold, W. (2012): Geothermal Well Data Grand Forks Fairbanks, Alaska 99775.
North Dakota. Zwarycz, K. (2002). “Snow melting and heating systems
Ho, I. H. and Dickson, M. (2015). “Assessment of Pavement based on geothermal heat pumps at Goleniow Airport”
Snow-Melting System Using Geothermal Energy in Cold Poland Transportation.
117
Difficult behavior of young continental “loess – like” clayey

soil deposits of mud-volcanogenetic heritage in Azerbaijan
E.E. Vicente
Consultant, A-Mehr, California, USA
A.M. Krumb & F. Ulbricht

Fugro Consult GmbH, Berlin, Germany
ABSTRACT: These soil deposits, while in their native unsaturated condition are relatively stiff to very stiff in
consistency, reflected in their relatively low compressibility and moderate shear strength. These soils, however,
also exhibit “collapsible” behavior upon wetting, a sudden drastic loss of their bearing capacity, namely, softening
and losing much or most of their initial “dry” shear and compressive strength. In some cases, these soils collapse
even under its own weight, i.e., without any increase of external or surface loads. These soils may lose a
significant fraction of their original (unsaturated) undrained shear strength, in some cases more than 50%
and even up to 95%, with potentially serious consequences on the bearing capacity and settlement of shallow
foundations.
These soils also contain sufficient amount of active clay minerals to be considered potentially “expansive” upon
wetting (subject to a low to medium degree of heave or “swelling”) which may seem an apparent contradiction.
Since those phenomena are both triggered by the addition of water they would occur simultaneously, and it
may be difficult to identify and quantify contributions of each of them. Further, clay components of these soils
are also considered slightly to moderately “dispersive” (i.e., exhibiting deflocculating tendencies) in a Sodium
environment (rather than flocculating as in a Calcium environment). Dispersive clay (as opposed to “ordinary”
clays) particles can easily and rapidly dislodge in the presence of flowing water, which added to the presence of a
large fraction silt-size particles, make these clayey soils “erodible” (quite susceptible to soil particle detachment
and migration) in presence of flowing water.
An exploratory drilling and CPT sounding program, combined with a variety of in-situ and geotechnical,
chemical laboratory tests and geologic analyses were conducted to characterize the behavior of this material of
difficult behavior, along coastal Caspian Sea areas (CSA) of Azerbaijan.
1 INTRODUCTION to the effects of one or more of the above listed

characteristics.
1.1 Genesis A history of earthwork-related problems includes
excessive settlement and bearing capacity failure of
Relatively young soils deposits (recent, ongoing
shallow foundations (collapse upon wetting), failure
soil forming processes) of mud-volcanogenic her-
of road pavements and scour erosion requiring signif-
itage deposited within an alluvial sedimentological
icant maintenance efforts and resources. Specifically,
environment that can be described as weathering
this type of soils has caused significant damage to
products of volcanic muds with components of dacitic-
existing roadways and foundations of energy infras-
andesitic origin, covering extended areas of southeast
tructure facilities in and around the main study area of
Azerbaijan.
roughly 10 km2 along coastal Caspian Sea areas about
40 kilometers south of Baku, Azerbaijan.
1.2 Problematic soils
1.3 Objective of this study
These surficial unsaturated “loess-like” soil deposits
soils are known to be very sensitive to, and adversely Consequently, this study has focused on: a) under-
affected by, the presence of water in its porous spaces standing soil behavior; b) evaluating cost-effective
and to the flow of water. These fine-grained soils earthwork requirements by performing compaction
are known to have caused serious difficulties, upon trials and cost-effective earthwork utilizing these
wetting, to civil projects during and after construc- locally abundant soils, both freshly-excavated (FE)
tion, with costly unintended or premature repairs, due as well as re-excavated spoil heap (SH) stockpiled
119
soils; c) assessing deep and shallow foundation mineralogical content of these soils indicate quartz ≈
requirements for their adequate long-term and cost- 30%, chlorites ≈ 7%, muscovite/illite ≈ 26%, smectite
effective performance. mixed-layer ≈ 8 to 14%, plagioclase ≈ 13%, calcite ≈
10%, gypsum ≈ 1 to 2% and Anatas ≤1%. Concentra-
tions of Chlorides and Sulphates range below oceanic
1.4 Investigated site conditions
seawater, approximately meet the value of the Caspian
The main study site is underlain by a thickness of typ- Sea; no crystalline salts and only 1 to 2 percent gypsum
ically 2 to 12 meters of this type of “loess-like” soil were identified; hence the material could be classified
deposits of mud volcano-genetic origins. Field explo- as “non-saline” soil.
ration of this site primarily included performing 119 Chemical analyses show a calcium carbonate con-
Cone Penetration Test (CPT) soundings to maximum tent of about 10 percent and total dissolved salts (TDS)
depths of 25 meters, and 70 boreholes to maximum between 5 and 20%. Measured pH-values range from
depths of 60 meters, 40 test pits, downhole as well neutral to slightly alkaline which reflects an environ-
as surface geophysical surveys (shear wave propaga- ment ideal for the formation of chlorites and smectitic
tion velocities, and electrical and thermal resistivity), clay minerals. Smectite mixed-layer minerals were
constructing “test embankments” (compaction trials), identified in a volume mass of 10 to 15 mass percent;
in addition to conducting an extensive geotechnical and the high sodium absorption ratio (SAR > 8%) and
laboratory test program. exchange sodium percentage (ESP ≈ 9 to 24%) leads
Where groundwater is deep, for instance deeper to the conclusion of that clay minerals are dominated
than 10 meters, soils are mostly unsaturated (herein by Sodium type. Clays rich in sodium smectite are well
called “dry” or “unsoaked” for simplicity). However, known as bentonite.
due to poor surface water drainage conditions, swampy
areas also existed the study site which resulted in
wet and softened soils (and very difficult to access
2 NATURAL SOIL BEHAVIOR, INDEX AND
to exploration locations). These soft and swampy
ENGINEERING PROPERTIES
areas were also explored using CPT soundings and/or
dynamic penetration probing light (DPL).
Geotechnical Properties for the “loess-like” CSA soil
deposits are summarized in Table 1b, including prop-
1.5 Geotechnical characterization erties in their unsaturated (“dry”) as well as “wet”
conditions. Their natural or in-situ properties have
At a large study site, these materials are mostly fine been studied using a variety of field exploration, in-
grained, and generally consist of lean clay (CL) to silty situ and laboratory testing methods in their initially
clay (CL-ML), and occasionally fat clay (CH) with “Dry” (Unsaturated, or unsoaked) and “Wet” (soaked
variable amounts of sand. or nearly fully saturated) soil conditions, as described
In their natural unsaturated (“dry”) condition, these in the following paragraphs.
soils generally exhibit a stiff to hard consistency, A sudden drop in their original (unsaturated)
namely, SPT-N ≈ 20 to 70 blows/per 0.3 m; and strength and bearing capacity was measured and doc-
CPT-qc ≈ 2.5 to 18 MPa. In-situ measurements of umented in the field by performing cone penetra-
shear wave velocity, from downhole velocity logging tion test (CPT) soundings, with correlated undrained
at 10 exploratory borings typically were, Vs ≈ 250 shear strength after wetting on the order of 2 to
to 360 m/s. Typical ranges (excluding highly infre- 15% of the unsaturated value. A similarly drastic
quent or extreme values) of particle size fractions are: drop in resistance was observed using various other
median grain size, D50 , commonly ranges from 0.002 geotechnical laboratory testing techniques. For exam-
to 0.034 millimeters, with clay-size fraction (minus 2 ple, the resistance of pavement’s native subgrade soils
micra), or CF ≈ 15 to 45%, plasticity index, PI ≈ 10 to was measured in the laboratory by performing CBR
30, and clay “activity” index CAI=PI/CF ≈ 0.4 to 0.9. (California Bearing Ratio) tests, as well as triax-
A summary of Soil Classification and Index Prop- ial unconsolidated-undrained compression (Tx-UU)
erties is provided in Table 1a; a summary of their tests. Often, “postwetting” strength and stiffness val-
Geotechnical Engineering Properties for both natural ues were as low as 2 to 10% of those for unsaturated
unsaturated condition and after wetting is provided in soils. Similarly, a significant drop on effective shear
Table 1b. Tables 1a and 1b provide “maximum” mea- strength parameters (roughly 65% and 15% of their ini-
sured ranges of the soil properties or parameters of tial unsaturated values, respectively, for friction angle
natural soils; however “typical” ranges were also pro- and cohesion intercept) were measured by performing
vided in the following paragraphs, which exclude what direct shear (DS) tests. Furthermore, drastic changes in
were deemed as outliers, namely, highly infrequent or compressibility parameters (stiffness) were measured
extreme values of these properties. in laboratory by performing double (1-D) oedometer
tests on specimens of initially “dry” (unsaturated or
1.6 Mineralogy and geologic characterization “unsoaked”) as well as initially “wet” (nearly saturated
or “soaked”) soils, and also combined cases (initially
X-Ray diffraction (XRD) analysis performed (at T.U. “dry” then “soaked”) at different 1-D compression
Freiberg, Germany) on two specimens to identify the loads.
120
Table 1a. Soil classification and index properties.
Unsaturated (“Dry”) soils

Natural soils
Parameter, Property Number of tests Measured range
Particle size distribution

Clay-Size Fraction 135 5 to 45
[<2 µ], CF (%)
Fines Fraction 135 48 to 99
[<75 µ], FF (%)
Silt-Size Fraction 135 45 to 70
[2 µτo75 µ], SiF (%)
Sand Fraction 135 3 to 25
[4 mm to 75 µ], SF (%)
Gravel Fraction 135 ≤2
[>4 mm], GF (%) Median Grain Size, 53 0.002 to 0.034
D50 (mm)
Atterberg Limits, Water and Organic Matter Content
Plastic Limit, PL (%) 147 13 to 24
Liquid Limit, LL (%) 147 25 to 56
Plasticity Index, PI (%) 147 5 to 35
Water Content/Liquid Limit Ratio, w/LL 123 0.06 to 0.63
Liquidity Index 132 −0.2 to +0.17
[(w-PL)/PI], LI
Clay Activity Index [PI/CF], CAI 109 0.37 to 0.90
Organic Matter, OC (%) 104 1 to 7
Calcium Carbonate Content, Ca (%) 10
Geologic Description and Classification
Alluvial soils with components of mud volcanic origin
Soil Classification Lean and fat clays (CL and CH,
per ASTM D2487)
Unit Weight and Volumetric-Gravimetric Soil Phase Relationships
Total Unit Weight (kN/m3 ) 84 14 to 21
Dry Unit Weight (kN/m3 ) 85 13 to 20
Moisture content (%) 278 2 to 31
Specific Gravity, Gs 120 2.58 to 2.77
Void Ratio, e 84 0.26 to 1.0
Porosity, n 84 0.21 to 0.5
Saturation Degree, S (%) 79 13 to 82
A description of geotechnical properties of these range of 0.04 to 0.15 and rebound (Cr ) indices in
soils is presented in the following paragraphs for both the range of 0.004 to 0.017, respectively (excluding
unsaturated conditions and upon wetting. extreme values);
b) At near-saturation (wet) conditions when inundated
2.1 Compressibility at the initial load of about 5 or 12 kPa, their com-
pression (Cc ) indices increase significantly, and
Compressibility upon loading and unloading was pri- are roughly in the range of 0.21 to 0.29, and
marily evaluated from laboratory oedometer (one- rebound (Cr ) indices are in the range of 0.006 to
dimensional consolidation) test data. The tests were 0.040, respectively. Their coefficient of consolida-
performed on relatively undisturbed specimens of tion ranges from approximately 3 to 43 m2 /year.
natural soils. Soil compressibility indices and poten- c) Inundated at loads between 200 and 400 kPa, a col-
tial for relatively sudden volumetric change behavior lapse index, Ie , generally about 1.6 to 6% percent
(reduction or increase) upon wetting, were measured at 200 kPa, and up to 7.5 percent at 400 kPa.
for the following cases: a) At in-situ moisture content
and density conditions (initially unsaturated, or “dry”), The degree of specimen collapse may be classified
without addition of water; b) Inundated at initial as “slight” when the collapse index, Ie , is less than 2
seating load; and c) Inundated after reaching a percent; “moderate” when ranging from 2 to 6 per-
cent, “moderately severe” when ranging from 6 to 10
selected load, usually at pressures ≈ 100, 200, or
percent, and “severe” when greater than 10 percent.
400 kPa An example is given in Figure 1.
The highest measured collapse value at overburden
a) At in-situ moisture content (“dry”) soils are very pressures other than 200 kPa was approximately 12.5
stiff, with compression (Cc ) indices typically in the percent, on a specimen inundated at 100 kPa.
121
Table 1b. Summary of geotechnical engineering properties for natural unsaturated and wet condition.
Unsaturated (“Dry”) soils “Soaked” (or “Wet”) soils

Parameter, Property
In-Situ Conditions Number of tests Measured range Number of tests Measured range
Field Consistency
BS 1377, ASTM standards Very stiff to hard Very soft to soft
CPT-qc – Tip Resistance (MPa) 119 soundings 5 to 10 (typical) 3 0.5 to 1.5
2.5 to 18.5 (max.)
CPT-Rf – Friction Ratio (%) 119 soundings 3 to 5 3 2.5 to 5
SPT-N value (blows/0.3 m) 24 borings 16 to 69
Shear Wave Velocity, Vs (m/s) 10 downhole tests 250 to 360
Compressibility (Oedometer tests)
Compression Index, Cc 12 0.016 to 0.15 24 0.084 to 0.29
Recompression Index, Cr 12 0.001 to 0.017 24 0.015 to 0.062
Average Cc /Cr Ratio 12 9 24 5.7
Coefficient of Consolidation, Cv (m2 /year) – – 21 1.5 to 88
Volumetric Changes Upon Wetting
Collapse Potential, at 100 kPa, Ic (%) – – 11 0.8 to 12.6
Collapse Index, at 200 kPa, Ic (%) – – 2 1.5 to 6
Free Swell at 1 to 12 kPa (%) – – 11 0.1 to 3.3
Swell Pressure, SP (kPa) – – 11 3.5 to 35 (max)
7 to 15 (typical)
Undrained Shear Strength Parameters (Peak Values)
From Laboratory Tx-UU Tests, Su (kPa) 24 120 to 1085 (max) 11 30 to 142
200 to 700 (typical)
From Laboratory Tx-UU tests, 24 1.2 to 8.5 (max) 11 0.24 to 2.7 (max)
Undrained Shear Strength Ratio Su /σ3 1.5 to 5.5 (typical) 0.35 to 0.95 (typical)
From Laboratory Tx-CIU Tests, Su (kPa) – – 12 φ ≈ 12 ± 1 degrees;
c ≈ 6 ± 3 kPa
Correlated from in-situ CPT soundings,
CTP-Srmu (kPa) – 500 ± 200 (natural) 10 to 20 (typical)
600 ± 150 (compacted) 10 to 70 (overall)
Drained Shear Strength Parameters (Peak Values)
From Laboratory DS Tests; Internal Friction, φ 25 φ ≈ 35 ± 3 degrees; 25 φ ≈ 24 ± 2 degrees;
(degrees); Cohesion, c (kPa) c ≈ 50 ± 15 kPa c ≈ 7 ± 5 kPa
From Laboratory Tx-CIU Tests; Internal – – 12 φ ≈ 27 ± 2 degrees;
Friction, φ (degrees); Cohesion, c (kPa) c ≈ 2 ± 1 kPa (Post-
hydrocompression)
Saturated Hydraulic Conductivity (Permeability)
Natural (in-situ) conditions (undisturbed) – – – 9 4.1 × 10−10 to
Permeability Coefficient, K (m/s) 4.2 × 10−7
Clay Dispersivity – Physical Tests
Pinhole Test Classification Group – – 5 ND1 to ND3
Double Hydrometer – Percent Dispersion, PD (%) – – 29 7 to 93
“Crumbs” Test – – 20 Grades 2, 3 and 4
(25%, 70% and 5% of
test data, respectively)
Clay Dispersivity - Geo-Chemical Tests
Exchangeable Sodium Percentage, ESP (%) 32 18 to 54
Sodium Absorption Ratio, SAR (meq/L) 32 7 to 22
Sodium in Saturation Extract, PSSE (%) 32 14 to 48
Total Dissolved Solids, TDS (%) 29 5.7 to 24
2.2 Swell potential of natural soils upon wetting swell pressures were typically in the range of 7 to
15 kPa, with a high value of 35 kPa.
Some of the clayey soils also showed some “free”
swelling potential when wetted at low confining pres-
2.3 Shear strength properties
sures, often ranging from about 1 to 7%. Free swell
and swell pressure testing were performed to mea- They were initially estimated from field measure-
sured unrestrained swelling deformation upon wetting ments (CPT soundings) based on published empir-
(under only the initial seating vertical pressure or 5 to ical correlations between CPT-qc (tip resistance)
12 kPa), and pressure necessary to restraint swelling and undrained shear strengths. Subsequently, shear
upon wetting (or zero vertical deformation). Measured strength parameters were directly measured in the
122
Effective internal friction angle, φ ≈ 27 degrees
and an effective cohesion intercept, c’ ≈ 2 kPa, from
a data set of 12 individual tests (4 test series).
Total strength parameters were: friction angle,φ ≈ 12
degrees and cohesion intercept, c ≈ 6 kPa.
Effective shear strength parameters from DS tests
indicated an effective friction angle, φ ≈ 35± 3
degrees and cohesion, c’ ≈ 50 ± 15 kPa; while for
“wet” condition the internal effective friction angle
drops to φ’≈ 24 ± 2 degrees and the effective cohesion
to c’ ≈ 7 ± 5 kPa.
2.4 Saturated hydraulic conductivity properties

Soil permeability properties were primarily derived
using a flexible membrane (triaxial cell) testing
apparatus or permeameter. In general, coefficients
of permeability (saturated hydraulic conductivity)
of natural soil deposits were found to vary from
about 4.1 × 10−10 to 4.2 × 10−7 meters/second (m/s),
Figure 1. Results of laboratory oedometer tests – illustration with an average from 9 laboratory tests of about
of swell and collapse behavior. 1.2 × 10−7 m/s and a coefficient of variation of
approximately 1.3.
laboratory by performing unconsolidated-undrained 2.5 Potential for dispersive clay behavior
triaxial (Tx-UU) and consolidated-undrained triaxial
(Tx-CIU) compression tests as well as direct shear Surficial CSA soils, were evaluated for potential “dis-
(DS) strength tests on relatively undisturbed speci- persive” clay behavior. Obtaining more specific fac-
mens. Both total and effective shear strength parame- tual information on the conditions that would possibly
ters were evaluated for soils in their natural moisture make these soils dispersive should help in selecting
condition (i.e., unsaturated, with degrees of satura- appropriate mitigation measures. No single physical
tion generally less than 60%), as well as in a near full or chemical test type by itself is known to be 100 per-
saturation (or “wet”) condition. cent reliable in all cases, applicable to all geographical
Undrained Shear Strength, Su , was initially derived (worldwide) locations, or to provide sufficient infor-
from 119 CPT soundings performed throughout the mation for design. The current standard of practice is
study site using an average cone factor, Nk ≈ 15. CPT- to perform a suite of various physical and geochemi-
based undrained shear strength estimates were derived cal tests. Consequently, the following sets of physical
from about 30,000 individual CPT data points, extend- and geochemical tests were performed: a) pinhole; b)
ing to depths of 10 meters. This correlation resulted double hydrometer; c) “crumbs”; and d) EFA (Erosion
in Su ≈ 500 ± 200 kPa for unsaturated conditions. Function Apparatus, developed by the University of
Conversely, “wet” natural soils exhibited significantly Texas at Austin).
lower values, namely, Su ≈ 10 to 70 kPa. Geochemical tests included: e) Exchangeable
Approximately 80 undrained triaxial compression Sodium Percentage (ESP, or Exchangeable Na/CEC);
tests (Tx-UU) in the laboratory. Tx-UU tests on “dry” f) Sodium Absorption Ratio (SAR, or Na/[(Ca +
soils indicated: Su ≈ 120 to 700 kPa; while those on Mg)/2]2 ); g) Percent Sodium in Saturated Extract
“wet” soils were drastically lower: Su ≈ 30 to 140 kPa. (PSSE); h) Total Dissolved Solids (TDS), and i) Cation
Observed stress-strain-strength soil behavior was Exchange Capacity (CEC, or Na), as summarized in
often (but not always) relatively brittle for “dry” soils. Table 1b. In general, “Sodic” Soils conditions are
For instance, peak undrained shear strength values of considered to exist when ESP >6%; and “excessive”
700 kPa were measured at an axial strain of about 1 Sodium conditions exist when ESP > 15%, or SAR
to 2.5%, dropping initially to post-peak value of about >10 meq/L. Average ESP value was approximately
200 to 300 kPa at about 5 to 8 percent, and to 120 kPa 35%, and average SAR ≈ 12 meq/L for the 10 Km2
at 20% strain. study area.
Total and Effective Shear Strength Parameters:
Total and effective shear strength parameters for undis-
3 COMPACTED FILL SOIL BEHAVIOR AND
turbed natural soils were independently assessed from
PROPERTIES
Tx-CIU triaxial tests performed on initially saturated
specimens with pore water pressure measurements.
3.1 Compaction trials
Tx-CIU tests performed on initially saturated soil spec-
imens also helped describe the stress-strain-strength An initial pair of test embankments were per-
behavior of soils and provided both undrained and formed to collect in-situ and laboratory test data on
drained sets of shear strength parameters. field-compacted specimens using Modified Proctor
123
Table 2a. Soil classification and index properties of field compacted material.
Compaction trials
Engineered fill parameter, property Number of tests Measured range
Field Consistency (Compaction Trials)

CPT-qc – Resistance 8 soundings 5 to 13 [3 pass.]
(MPa) [No. of passes] 6 to 16 [7 pass.]
CPT-Rf – Friction Ratio 8 soundings 3 to 4
(%)[No. of passes]
Unit Weight and Phase Relationships
Total Unit Weight (kN/m3 ) 222 13.5 to 22.5
Dry Unit Weight (kN/m3 ) 222 12.5 to 20.5
Moisture Content (%) 344 4.5 to 20
Specific Gravity, Gs 53 2.61 to 2.85
Void Ratio, e 222 0.30 to 0.99
Porosity, n 222 0.24 to 0.72
Saturation Degree, S (%) 216 13 to 82
Moisture-Density Relationships of Compacted (Fill) Soils
Maximum Dry Unit Weight 57 18.5 to 20
(kN/m3 ) – Modified Proctor
Maximum Dry Unit Weight 17 16.7 to 18
(kN/m3 ) – Standard Proctor
Optimum Moisture Content 57 10 to 15
(%) – Modified Proctor
Optimum Moisture Content 17 14 to 19
(%) – Standard Proctor
compaction efforts; a second pair of test embankments the advantage of allowing a more uniform distribution
used the Standard Proctor compaction efforts. of moisture content throughout the compaction layer.
Relatively undisturbed field-compacted specimens A second set of embankment tests was conducted to
were collected for: 1) freshly-excavated (FE) and com- further test different combinations of compacted layer
pacted; as well as 2) re-excavated from available spoil thickness (0.25 m, 0.20 m, 0.15 m and 0.10 m, number
heap (SH) stockpiles and re-compacted soils. These of passes of vibratory sheep foot roller (VSFR), vibra-
two options (compacted FE and SH fill soils) were con- tory smooth roller (VSM), and comparing Modified
sidered to address concerns by local engineers inclined versus Standard Proctor compaction energy, as well as
not to use of this type of “bad” soils. These soils had the VSFR speed (about 2 km/hour).
a long history of cases of excessive settlement and Note: the use of Standard Proctor test was also
bearing capacity failure of shallow foundations, and considered and suggested to provide a practical bal-
excessive repairs to roadways and pipeline utilities. ance between expansive and collapsible tendencies
Therefore, both compacted FE and SH fill soils in the soil upon wetting. With higher compaction
were tested after 3, 5 and 7 passes of a 14-ton vibratory moisture contents and lower dry densities, the result
sheep-foot roller per 0.25-meter-thick of each com- is a lower expansion potential, possibly at the expense
pacted soil layer, for a total thickness of each test of a slightly higher soil collapse potential.
embankment of approximately 2.5 meters, to com- A description of geotechnical properties of com-
pare their geotechnical properties as described in pacted soils is presented in the following paragraphs
subsequent paragraphs and summarized in Table 2a. for both unsaturated conditions and upon wetting.
A relative compaction (or degree of compaction) of
RC ≈ 95% and near-optimum molding moisture con-
3.2 Compressibility
tent had initially been targeted (namely, −1% to +3%)
per Modified Proctor (MP) compaction test. Laboratory oedometer (1D consolidation) tests
That range of moisture content was intended to were performed on field-compacted and laboratory-
simulate generally desired earthwork specifications compacted (reconstituted) fill soil specimens.
during construction. These soils were, however, usu- Compacted fill soil properties, summarized in Table
ally field-compacted slightly on the dry side of the 2b, for three tested moisture conditions, were: a) At
optimum moisture content because of great difficulties near-optimum moisture content, compacted soils are
in soil-water mixing of this soil type in the field with- very stiff, with compression index, Cc ≈ 0.02 to 0.12
out a special soil-water mixing equipment or plant. and rebound index, Cr ≈ 0.002 to 0.011, respectively;
This is about 10 to 15 percent moisture content per MP b) At near-saturation conditions, when inundated at
test. Using layer or sublayer thickness of 0.10 m had the initial seating load of about 5 or 12 kPa, their
124
Table 2b. Summary of geotechnical engineering properties for compacted fill soils.
Compaction trials (Drive-cylinder laboratory-tested specimens

field compacted at near-optimum moisture content and RC ≈ 95%)
Near-optimum moisture Near-saturation moisture
Engineered fill Number of Measured range Number of Measured range

parameter, property tests tests
Compressibility (Oedometer tests)

Compression Index, Cc 39 0.020 to 0.158 20 0.04 to 0.12
Recompression Index, Cr 38 0.002 to 0.020 29 0.003 to 0.032
Average Cc /Cr Ratio 38 9.8 29 4.8
Volumetric Changes Upon Wetting
Collapse Index at 100–800 kPa, Ic (%) – – 4 1.4 to 3.8
Free Swell [at 1 to 12 kPa], FS (%) – – 10 1.1 to 6.9
Undrained Shear Strength Parameters – Peak Values
From Tx-UU Tests, Su (kPa) 49 100 to 680
From Tx-CIU Tests, Su (kPa) – – 9 φ ≈ 19± 1 degrees;
c ≈ 5± 3 kPa
Correlated CTP- Su (kPa) 9 300 to 750 [3, 7 passes]
Drained Shear Strength - Peak Values
From Laboratory DS Tests 16 φ ≈ 31 ± 2 degrees;
c ≈ 50± 15 kPa
From Laboratory Tx-CIU Tests 9 φ ≈ 30 ± 2 degrees;
c ≈ 6 ± 3 kPa
(Post-hydrocompression)
Saturated Hydraulic Conductivity (Permeability)
Field-Compacted, K (m/s) – – 13 1.9 × 10−10 − 1.4 × 10−9
Laboratory-Compacted, K (m/s) – – 8 6.5 × 10−11 − 1.6 × 10−9
Pavement Subgrade Static Resistance (California Bearing Ratio, CBR)
In-Situ CBR (%) - Field 41 18 to 93
Laboratory CBR (%) – Laboratory 17 22 to 48 10 1.4 to 2.6
Plate Load Test (PLT) – After 3, 5 and 7 passes of 14-ton sheepfoot roller
Deformation Modulus, Loading 93 10 to 29
Phase, Ev1 (kN/m2 ), 20 to 42
29 to 55
Deformation Modulus, Re-Loading Phase, 93 32 to 73
Ev2 (kN/m2 ) 58 to 88
67 to 110
Moisture-Density Relationships of Compacted (Fill) Soils – Modified and Standard Proctor (MP & SP) Tests
Maximum Dry Unit Weight (kN/m3 ) 57 18.6 to 20 [MP]
Maximum Dry Unit Weight (kN/m3 ) 17 16.7 to 18 [SP]
Optimum Moisture Content (%) 57 10 to 15 [MP]
Optimum Moisture Content (%) 17 14 to 19 [SP]
compression indices increase significantly to Cc ≈ 3.3 Swell potential upon wetting

0.04 to 0.10, with rebound, Cr ≈ 0.007 to 0.025,
Laboratory-compacted clayey FE and SH soils also
respectively; c) When inundated at loads between 100
showed some swelling potential when wetted at rel-
and 400 kPa, their compressibility curves prior to and
atively low confining pressures. At seating loads of
after wetting also tend to be somewhat similar (near-
about 10 to 12 kPa, free swell averaged approximately
parallel) to those of “dry” and “wet” soils, respectively.
3.3 to 4 percent, respectively. Ranges of free swell for
Even though these soils had been previously com-
compacted FE and SH soils were about 1 to 7 percent
pacted to RC ≈ 95%, a residual collapse generally
and 2 to 5 percent, respectively. Overall, no significant
about 1.5 percent was never-the-less still observed at
difference was found between FE and SH soils.
an inundation load of 200 kPa, (down from an average
of 3.5 percent for natural soils). At inundation loads
3.4 Undrained shear strength
between 100 and 800 kPa, for samples from another
compaction trials area, collapse potential values mea- Undrained shear strength data were collected from
sured in the laboratory were generally less than 4 field-compacted FE and SH fill soils during com-
percent. paction trials, using Cone Penetration Test (CPT)
125
soundings in the field, and triaxial tests on relatively Total strength parameters were: friction angle φ ≈ 19
undisturbed soil specimens in the laboratory. degrees and cohesion intercept, c ≈ 5 kPa.
CPT soundings provided undrained shear strength Direct Shear (DS) tests were performed on undis-
data at four locations through the 2- and 3-meter- turbed field-compacted (unsaturated) samples of CSA
thick test embankments, plus about 3.5 to 4 meters soils. The specimens were moisture-conditioned in the
of their underlying natural soils to a total depth of field to about 8 to 11 percent, which is 2 to 5 per-
approximately 5.5 to 6 meters. cent less than the optimum moisture content, based
Compacted FE fill soils exhibited undrained shear on the Modified Proctor (MP) compaction tests. The
strength of, Su ≈ 500 ± 100 kPa and ≈650 ± 150 kPa, tested specimens were not initially inundated in the
for 3 and 7 passes of the 14-ton vibratory sheep- laboratory prior to shearing, and they remained unsatu-
foot roller (per 0.25-meter-thick compacted soil rated throughout DS testing. For this testing condition,
layer), respectively. Median strength, Su(50) ≈ 500 the derived shear strength parameters were: effective
and 650 kPa were obtained for 3 and 7 passes of friction angle, φ ≈ 31 degrees an effective cohesion,
the 14-ton vibratory sheepfoot roller, with typical c ≈ 50 kPa.
ranges of Su(50) ≈ 350 to 650 kPa and 400 to 900 kPa, For comparison purposes, the internal friction angle
respectively. is about 4 degrees higher than what was derived
Similarly, for compacted SH fill soils, four CPT from Tx-CIU tests for natural soils initially saturated
soundings were conducted through the 2-meter-thick in the laboratory. It is anticipated that for the inun-
test embankment, plus about 3.5 to 4 meters of their dated (or “wet”) condition, DS-derived shear strength
underlying soils to a total depth of approximately 5.5 parameters will be significantly lower.
to 6 meters.
The undrained shear strength of compacted SH 3.6 Saturated hydraulic conductivity
soils for 3 passes ranged from less than 300 to over (permeability)
850 kPa, with occasional localized peaks in excess of
1,000 kPa. For 7 passes, the undrained shear strength Field-Compacted Soils: The permeability coefficient
ranges typically from about 400 to 900 kPa. Statistical measured in the laboratory for FE and SH field-
(cumulative frequency distribution) analyses indicate compacted soils ranged from about 1.9 × 10−10
a median strength, Su(50) ≈ 530 and 560 kPa, suggest- to 1.4 × 10−9 m/s, with an average of about
ing also that no significant densification of soils takes 8.9 × 10−10 m/s from 13 laboratory tests, with a stan-
place after 7 passes. This seems to be the case for both dard deviation of 3.64 × 10−10 m/s and a coefficient
compacted FE and SH fill soils. of variation of nearly 0.5. It should be noted that the
Tx-UU Tests: Undrained shear strength from undis- moisture contents of the field-compacted specimens
turbed specimens of field-compacted FE and SH fill collected from compaction trials were often 1 to 5%
soils was directly measured from Tx-UU compression below the optimum moisture content (based on MP
tests in the laboratory. Field-compacted specimens of test) because of significant difficulties in soil-water
FE and SH fill soils were tested at confining pres- mixing in the field.
sures of 50 and 400 kPa, respectively. Undrained shear Laboratory-Compacted Soils: The permeability of
strength results ranged from about 130 to 230 kPa and laboratory-compacted specimens of engineered fill
from 500 to 590 kPa, respectively, at those confining soils remolded to near-optimum moisture content and
stresses. 95 percent relative compaction (MP), ranged from
Similarly, undrained shear strength results for com- about 6.5 × 10−11 to 1.6 × 10−9 m/s, with an average
pacted SH soils tested ranged from about 130 to of about 4.2 × 10−10 m/s from 8 laboratory tests, and
380 kPa and from 490 to 680 kPa, for confining pres- a standard deviation of 5.6 × 10−10 m/s.
sures of 50 and 400 kPa respectively. This appears
to suggest that compacted SH soils are not signifi-
cantly stronger and are slightly more variable than FE 3.7 Subgrade resistance and deformation
compacted soils. modulus
Subgrade resistance for pavements and deformation
modulus of engineered fill soils were evaluated during
3.5 Effective and total shear strength parameters
compaction trials by performing a series of: a) Cali-
Triaxial CIU Laboratory Tests were performed on fornia Bearing Ratio (CBR) tests both in the field and
field-compacted FE soil specimens during compaction in the laboratory, as well as b) field plate load tests
trials to help evaluate the stress-strain-strength behav- (PLT) on FE and SH field-compacted soils.
ior of engineered fill soils. Field-compacted soil Pavement Subgrade Resistance from CBR tests:
specimens were initially saturated in the laboratory, The typical range of “unsoaked” field CBR values on
achieving Skempton’s B-value of 0.95 or greater, to compacted FE fill, performed at near-optimum mois-
measure pore pressures during testing and to pro- ture content, was 39 to 61%, with a maximum range
vide both undrained and drained sets of shear strength of 18 to 93.
parameters. Effective internal friction angle φ ≈ 30 Both unsaturated or “unsoaked” and “soaked” CBR
degrees and an effective cohesion intercept c ≈ 6 kPa, tests were performed in the laboratory. Optimum mois-
were derived from a data set of 8 tests (2 test series). ture content values were actually achieved in the
126
laboratory. The typical range for “unsoaked” labora- moisture content (≈ +1% to +3%); and d) full-time
tory CBR values was 22 to 48. continuous program geotechnical engineering mon-
Since soaking duration time is also known to have itoring of earthwork during construction including
an effect on measured CBR, compacted soils were in-situ and laboratory testing for conformance to
“soaked” in the laboratory for two different soaking project construction quality assurance specifications
periods: 4 days and 20 days. Measured laboratory throughout the engineered (compacted) fill area.
“soaked” CBR values ranged from 4 to as little as 0.5, Foundation Support: use of deep foundations for
namely, drastically lower than (or roughly 2 to 10 per- critical heavy and settlement-sensitive structures, or
cent of) the “unsoaked” laboratory CBR values, for any structures that have connections to other structures
specimens remolded at near-optimum moisture con- with low-movement tolerances. Consideration may be
tent and a relative compaction (RC) ≥ 95 % (per MP given to, for instance, drilled and cast-in-place rein-
tests). forced concrete piles (with diameters ranging from
Deformation Modulus from field plate load tests: 0.45 to 0.9 meters, depending on the load). The use
were measured for both field-compacted FE and SH of driven precast concrete or steel piles in very stiff
soils, including the initial loading (Ev1 ) and reloading to hard unsaturated soils may not have significant
(Ev2 ) phases. advantages and result in potential installation diffi-
For field-compacted FE fill: The median moduli culties or higher cost. Small, lightly-loaded structures
for 3, 5, and 7 passes of a 14-ton vibratory sheepfoot that can tolerate some settlements could be sup-
roller were: Ev1 ≈ 26, 32 and 40 kN/m2 , and Ev2 ≈ ported on shallow foundations or mats resting on
68, 72 and 80 kN/m2 , respectively. Furthermore, the engineered fill.
stiffening rate of compacted FE soils, was measured by “Loess-like” CSA soil deposits at the Azerbai-
the ratio of the reloading and initial loading modulus jan study area primarily consist of lean clay (CL)
of deformation, Ev2 /Ev1 ≈ 2.7, 2.2 and 2.0 for 3, 5, and to silty clay (CL-ML) and occasional fat clay (CH)
7 passes of compaction equipment, respectively. with variable amounts of sand. This soil unit in its
For field-compacted SH fill: The median mod- native unsaturated (“dry”) condition is generally stiff
uli were: Ev1 ≈ 24, 30 and 35 kN/m2 , and Ev2 ≈ 67, to very stiff in consistency, with carbonate bonding.
74 and 80 kN/m2 , respectively. The stiffening rate of The results of the index tests in the region indicate
compacted SH soils, as measures by the modulus of that these soils are generally “collapsible” and soften
deformation ratio, was Ev2 /Ev1 ≈ 2.8, 2.5 and 2.3 for considerably upon wetting. These soils are consid-
3, 5, and 7 passes of compaction equipment, respec- ered potentially “expansive” and subject to volumetric
tively. In general, compaction trials therefore sug- changes upon moisture changes (namely, heave or
gested fairly similar results of deformation modulus “swelling/shrinkage”), and also slightly to moder-
for both field-compacted FE and SH fill soils. ately “dispersive” in nature and therefore may cause
difficulties during and after construction.
The degree of collapse potential within the study
4 MITIGATION MEASURES – SITE area ranged between slight to moderate. Expansion
PREPARATION, EARTHWORK AND potential was generally estimated as low to medium.
FOUNDATIONS Since those phenomena are both triggered by the addi-
tion of water, they will likely occur simultaneously,
When avoidance of these soils for earthwork and and it may be very hard to identify contributions from
shallow foundations is not feasible, it is consid- each of the processes. Nonetheless, with a good under-
ered especially important to mitigate these conditions standing of the degree of severity of each process, a
and anticipated consequences. This may include, for balanced approach may be adopted for treating this
instance, cases of: structures straddling on shallow type of soil deposits within a specific area.
foundation supported by or into these soil deposits, The following discussions are general guidelines
therefore subject to high potential for differential set- (namely, not meant to be site-specific or project-
tlements or loss of bearing capacity resulting from specific recommendations, but) to try to assist in
soil wetting. Active pipelines that may be particularly initial planning and preliminary design activities for
vulnerable to differential settlements at roadway cross- site preparation, earthwork and foundations construc-
ings and that would require protection against these tion in similar CSA “loess-like” soil deposits, and to
settlements. minimize associated risks as much as possible.
Site Preparation and Earthwork Measures: Imple-
mentation of design and construction provisions are
advised, which may generally include: a) adequate sur- 4.1 Expansive soils
face and subsurface water drainage control measures Based on the results of the Atterberg Limits tests and
(and impermeabilization or water barrier); b) a fairly the available volume change data, these native “loess-
thorough soil-water mixing (moisture-conditioning) like” soils above the groundwater level are judged to be
procedure for engineered fill soils, which may include somewhat expansive with a low to medium degree of
the use of an effective and advanced soil-water mixing shrink/swell potential, which would generally be a pri-
plant (for compaction of fill soils at a large site); c) mary consideration for grading and foundation design.
soil compaction in the field slightly wet of optimum With seasonal weather and associated soil moisture
127
content changes, and longer-term periods of drying the localized depressions caused by collapses or resist
and rewetting, soil shrink and swell could occur. Those expansion forces and the resulting differential move-
soil volume changes can damage slabs, pavements, ment without causing cracks and distress to the super-
and shallow foundations. The general approaches that structure. Reinforced concrete slabs with a system of
can be used individually or in combination in order underlying rigid cross-beam grids or post-tensioned
to deal with expansive soils and reduce the risk slabs-on-grade have been used successfully on col-
of structural damage and cracking are described in lapsible and expansive soils. In general, shallow spread
Section 4.3. footings are not used for critical structures in a
collapsible and expansive soil environment.
4.2 Collapsible soils For collapse: Where they are used, however, tech-
niques such as deepening and widening the footings
Based on the results of the Atterberg Limits tests and
and increasing the reinforcement around the perimeter
the available volume change data, the CSA native
and into the floor slab to stiffen the foundations, are
unsaturated “loess-like” soils (above the groundwater
usually applied to decrease the bearing pressure and
level) are judged to have a slight to moderate degree
strengthen the foundations, therefore minimizing the
of collapse potential upon wetting.
potential for distress.
Note– Rodgers (1995) proposed the following defi-
For expansion: Where they are used, however, tech-
nition of this type of soils: “…a collapsible soil which
constituting parts have an open packing and which niques such as deepening and narrowing the width of
forms a metastable state that can collapse to form a the footings, and increasing the reinforcement around
closer-packed more stable structured of significantly the perimeter and into the floor slab to stiffen the
reduced volume. In most collapsible soils the struc- foundations are usually applied to increase the bearing
tural units will be primary mineral particles rather pressure and strengthen the foundations to minimize
than clay minerals. The collapse process that occurs the shrink/swell potential.
in these soils gives them a geotechnical significance.” Deep Foundations. Deepened foundation and piles
The collapse potential of these soils is another pri- may be used to transfer the structural loads to lower
mary consideration to be accounted for grading and strata and bearing zones where the potential expansion
foundation design. Soil collapse or additional set- and collapse is not an issue and adequate founda-
tlement could occur with seasonal moisture content tion support can be provided. The piles may either be
changes and infiltration of surface water due to rainfall designed as structural elements and directly support
or onsite activities. Soil volume changes can damage the structures or they may be used to strengthen and
slabs-on-grade, pavements, and shallow foundations. stabilize the collapsible soil zone with the structures
The following are the general approaches that can be supported on a shallow and mat system of foundation.
used individually or in combination in order to deal The deepened foundation and pile option may be con-
with collapsible soils and reduce the risk of structural sidered to be the most risk-free system; however, it
damage and cracking: may not be the most economical option.
Control of Potential Water Sources. Since the source
4.3 General approach to reduce effects of of the settlement and collapse or expansion is derived
collapsible and expansive soils from an increase in the moisture content of the sub-
ject soil, one method for controlling the collapse or
Both effects Collapse (or hydrocompression) and expansion would be to control fluctuations in mois-
Swelling are triggered by an increase of moisture content. It is nearly impossible to fully prevent
ture within the sensitive soil. Therefore, the general moisture fluctuation of the subsurface soils. However,
approach is to separate water and soil or to strengthen it is possible to control the rate of change and seasonal
or design the structure to deal with soil movements. fluctuations. The most commonly used technique is to
Removal and Replacement. In this method, the place horizontal and vertical moisture-barriers under
collapsible or expansive soils are removed to a spec- and around the foundations to some depth. The barriers
ified depth and replaced with engineered (moisture- are not fully effective, but are quite helpful in reducing
conditioned and compacted) fill consisting of import the edge effect and differential movements.
materials or materials excavated onsite. For best These barriers may consist of low-permeability
results, soil within the full depth of the “active” zone earth or geosynthetic (geomembrane) liners, and
(seasonal moisture fluctuation zone), which could combined “geomembranes” and compacted low-
range up to about 4 to 8 meters, would need to be permeability earth liners, as well as associated filters
replaced. However, a more practical approach may be and subdrain systems (such as geosynthetic drains
to limit the depth of removal and replacement to a wrapped in filter fabric or geotextile). However, per-
lesser depth and accept some level of risk, if at all meable erosion protection measures (or “geotextiles”)
feasible. Non-expansive fill may be selected so that should only be installed where collapse settlement of
it does not cause accumulation of subsurface water. CSA “loess-like” soils will not cause a problem for
Therefore, native “loess-like” soils are prevented from foundations or underground utilities, unless a tailored
being in contact with water. design (with specialty engineered details) is provided.
Special Foundation Systems. The slabs and foun- Other Miscellaneous Options. Other options that
dations may be designed stiff enough to span over are also used to deal with collapsible soils are ground
128
modification or improvement methods, such as deep alternative to compacted fill” (ACI, 2005). Several
dynamic compaction, stabilization with additives, and other names have been used to describe this mate-
areal water ponding and pre-saturation of the site, rial, such as flowable fill, unshrinkable fill, controlled
inducing collapse prior to foundation construction, if density fill, flowable mortar, plastic soil-cement, soil-
feasible. cement slurry, and cement-sand slurry grout.
Where future excavation of these materials will
not be required, CLSM would generally have an
4.4 Erodible soils
unconfined compressive strength of less than 8.3 MPa.
In general, erosion is defined as the natural process Where future excavation of CLSM is anticipated,
by which wind, moving water, ice, and gravitational a compressive strength of less than 2.1 MPa may be
forces displace the solid and particulate materials of used. Where low unit-weight materials are needed, a
the land. Determination of soil erodibility is a complex low-density LD-CLSM material may be used.
process that requires consideration of soil type and In general, surface water penetration into underly-
texture, among several other factors. The most vulner- ing soils must be minimized in areas where possible
able soils are non-plastic soils with a particle diameter collapse settlement could cause a problem. Therefore,
roughly on the order of 100 microns (µ), such as fine erosion protection measures must be carefully selected
sands and sandy silt. The erodibility of soils generally in order prevent or minimize surface water penetration
decreases for smaller (clay-size) and larger (sand- and in areas where collapse settlement is a potential issue.
gravel-size) particles, respectively.
These CSA “loess-like” mud-volcano-genetic soils
deposits within the study area have been found to have ACKNOWLEDGEMENTS
silt-size particle fraction on the order of 40 to 70%,
and other particle-size fractions as shown in Table 1a. Our deepest thanks are extended to all contributors
At the investigated location, soils are typically clas- of test data and advise that assisted the authors in
sified as clays in accordance with the plasticity charts. the preparation of this article, including Ines Rom-
In general, “ordinary” clay and clayey soil types mel, Roberto Quaas, Dr. Jens Krumb, Jens-Peter Ertel,
are cohesive (and plastic), and therefore would not Kemal Guerel, Farhad Boniadi, Massimo Deiana,
normally be expected to degrade (erode) rapidly. Shuai Wang and many others.
However, “dispersive” clays tend to deflocculate
in the presence of water even without the aid of
mechanical agitation or chemical deffloculants. At the
study site, clays generally exhibited a low to moder- REFERENCES
ate degree of dispersivity and are considered erodible American Society for Testing and Materials (2011), Annual
soils therefore requiring erosion protection measures. Book of ASTM Standards, Section 4 Construction, Vol-
The following are general approaches or measures that ume 04.08, Soil and Rock.
appear applicable mitigate this problem, and can be American Concrete Institute, ACI (2005), Controlled Low
used individually or in combination in order to min- Strength Materials; Report 229R-99.
imize and control soil erosion on sloping or level Atkins (2010), Geologic and Geomorphological Mapping of
ground: Coastal Strip (Azerbaijan).
BSI British Standard Institution (1990), “British Standard
• Soil grading measures to provide effective surface Methods of Test for Soils for Civil Engineering Purposes,”
water drainage, e.g., low slope gradients, slope BS 1377:1990 (Parts 1–9, with Amendments).
benches, diversion ditches, etc. DIN 4023, 18125-1, 18134: Geotechnical Investigations and
• Gravel or sand-filled cellular confinement system, Testing, Laboratory Tests, and Plate Load Testing.
such as Geocell™or Geowe™placed on the surface EN ISO 22475-1, 22476-1, -2 and-3 (2005): Geotechnical
Investigation and Testing.
or in multilayer horizontal pattern forming the slope
GOST 23161-78 (19768), Laboratory Method for Determi-
face, nation of Subsiding Characteristics.
• Riprap or rock fill underlain by filter fabric, Rogers, CDF (1995), Genesis and properties of collapsible
• Concrete aprons or facings, soils. Springer
• Roller-compacted concrete, SNIP 2.02.01-83, National Codes & Standards of Russia.
• Retaining structures such as gabions, crib walls, or
reinforced earth, and
• Soil-cement slurry or controlled low strength mate-
rials (CLSM).
A CLSM material is a “self-compacted, cemen-
titious material used primarily as a backfill as an
129
Wind farm development and the use of geogrid-stabilised unbound

platforms and floating access roads on soft ground
M.A. Horton, T.L.H. Oliver & J. Cook

Tensar International Ltd, Blackburn, UK
ABSTRACT: Onshore wind farms in Northern Europe are frequently located on land that is marginal in
terms of accessibility by conventional plant. They are frequently located in upland areas on soils of with little
recognisable shear strength at formation. These factors, coupled with increasingly onerous bearing capacity
and deformation demands for access track and working platforms for delivery and erection of turbines plus a
claims-conscious industry, make for challenging designs to deliver economical and safe construction.
This paper considers track and platform specifications and requirements, and examines the risks that these
contain in terms of safe and economical design and construction in upland areas. The paper examines design
methods incorporating the use of stiff hexagonal polymeric geogrids to stabilise access tracks and platforms.
The paper also discusses the advantages, limitations and problems involved with ‘floating’ working platforms to
create stiff platforms to limit movement on heavily loaded but very soft ground.
1 INTRODUCTION appropriate to this environment and able to detect any

sensitivity that natural site materials might possess.
1.1 Development in upland areas Investigations should be directed and interpreted by
appropriately experienced staff to ensure that design
Wind farm development in upland areas in the United
data passed to access roads and platform designers is
Kingdom has been conducted since the mid-1980s, ini-
reasonably accurate and reliable.
tially with small-scale developments but more recently
One of the issues with this type of land is that access
with much larger scale developments incorporating
can be poor. During ground investigations, the use of
many tens of turbines. In northern Europe, upland
low ground pressure plant may be the only means of
windfarm construction occurs on low-value land with
accessing much of the site to conduct activities as
relatively low productivity, commonly represented by
drilling and trial pitting; safe foot and plant access
moorland. Soils in such areas consist of peats and clays
might not be feasible on extremely poor ground. For
with very low stiffness interspersed with rock, the lat-
access roads, this is not always such an issue as rela-
ter frequently providing the source of the stone that is
tively shallow investigation (<1 m) at relatively close
used to construct the wind farm infrastructure.
spacing combined with deeper intrusive surveys at a
Construction of wind turbines in such areas can per-
wider spacing can provide adequate information for
mit dual land use, for example combined with forestry
design.
or sheep and deer farming and recreational activi-
For working platforms however, deeper investi-
ties. There is advice available in previously published
gation is frequently required due to the need for
documents that outline good practice with respect
adequate information for turbine foundation design.
to construction in these environments (SNH 2010,
Turbines are often micro-sited (i.e. moved from
Scottish Renewables et al 2010).
originally planned locations) after preliminary inves-
tigations to take advantage of better ground condi-
1.2 Ground conditions
tions. The revised locations can then permit access
Upland areas in Northern Europe are typified by gener- for heavier, non-specialised plant to conduct detailed
ally wet ground conditions and soils with poor bearing investigation.
capacity. Soils frequently demonstrate shear strengths
typically of 24 kN/m2 (approximately 1% CBR) and
less and typically comprise varying depths of peat
2 SPECIFICATIONS
and weathered glacial till clays. The latter can show
a marked tendency to lose shear strength when wet-
2.1 Development
ted and subjected to compaction loads, such that their
shear strengths can fall to less than 50% of that which As little as a decade ago, specifications for wind farm
might be obtained from a static soil test. It is therefore access roads were few and far between, depending
important that ground investigation techniques are upon the contractor building the site to determine
131
Figure 2. Main lift crane conducting a working platform
performance test.
Figure 1. Automated Plate bearing test being conducted on
a working platform using an excavator as a reaction load. Typically, requirements for access roads usually
specify an allowable bearing pressure of 180–220 kPa
or an EV 2 of 80–100 MPa.
design and construction, usually using traditional
methods.
More recently, wind turbine manufacturers have
3 ACCESS ROAD DESIGN
formed their own specifications, commonly specify-
ing minimum access and working platform dimen-
3.1 Road types
sions, required geometry, minimum bearing capacities
for road and working platforms and sometimes com- There are essentially three road types in common use:
bined (or replaced by) requirements for Ev 2 and Ev 2 dig out and replace, floating and displacement.
to Ev 1 ratio (to determine adequate compaction). In Dig out and replace roads involve removing sur-
nearly all cases, verification of the roads and working face vegetation and soft superficial materials down to
platforms is required in terms of a plate bearing tests a more competent horizon, typically up to 1m depth
(e.g. BS 1377 Part 9 or an EV2 test (e.g. DIN 18134). (but can be more depending on specific site location
However, it is pointed out that the dimensions of the and conditions) and replacing with competent, usually
test plate for these tests (even for the largest test plates) granular, material to form a road, usually bringing the
is not the same as, say, the outrigger bearing plate on road up to or above the original ground level. Such
a 750t-lift crane; whilst they can provide a degree of roads require the removal and dumping of excavated
comfort, such a test does not replicate the dimensions material and, in the case of peat and allied materials,
and depth of influence of the real working load. this can make this construction less environmentally
On some sites, plate bearing testing on working plat- friendly than other techniques due to peat desiccation
forms is not the only testing undertaken. Where time and degradation.
permits, the main lift crane can do a more realistic bear- Floating roads (a focus for this paper) are utilised
ing test by raising the boom of the crane and aligning where deep peat is encountered and/or digging out
the counterweight load over each of the outriggers and and replacing would be uneconomic or undesirable.
spreader plates in turn. This places a load (of up to The usual method of forming these roads is not to
200t) over the outrigger for several hours allowing the strip-off the vegetation but to place the road directly
platform to be monitored. This represents a far more upon the existing surface. On a peat bog, for example,
realistic test of both loads and the loaded area for this the vegetation and root mat may be the stiffest part
application. of the formation as these deposits frequently show a
The use of EV 2 specifications (essentially a Ger- reduction in stiffness and shear strength with depth as
man highways-based specification with the blacktop the peat becomes less ‘rooty’ and more amorphous in
removed) has limited applicability on peaty windfarm consistency.
sites and those sites where the Ev 2 at formation is Floating roads can use ‘brash’ (tree branch trim-
less than 5MPa (approx. 1% CBR). In such sites, mings from forestry operations) or logs but more
alternative methods, including shear strength/CBR commonly nowadays use geogrids or a combination
assessment are required and alternative design can proof geogrids and brash in very poor conditions and
duce a more economical ‘floating road’ and platform especially over forestry furrows (Fig 3).
construction rather than by using such techniques as Floating roads nearly always settle after construc-
chemical stabilisation or cement mixing (which may tion. The weight of the placed geogrid and stone upon
be environmentally undesirable). the surface causes consolidation of the peat and leads
132
3.2 Geogrid/geotextiles and mechanisms
Geogrids and geotextiles have been used for several
years to improve the performance of access roads and
working platforms on soft ground. These materials
work in two distinct ways:
Reinforcement: the reinforcement geogrid or geo-
textile acts as a tensile element within the ground
helping to support the aggregate and running sur-
face in a ‘tension membrane’ or hammock mechanism
that depends on a certain amount of strain in order
to develop the strength needed to support the loaded
surface. This mechanism can need a considerable
anchorage length of material buried in the ground
(which may not be available under the shoulders of
the road alone) and may need additional geogrid or
geotextile buried in the ground at the sides of the
access road in order to prevent the geogrid or geotex-
tile from pulling through the ground; it is not adequate
Figure 3. Geogrid-stabilised floating access road on deep to assume that the buried length along the road line is
peat in felled forestry constructed on brash bundles (‘facine sufficient to allow the geogrid or geotextile to strain
mats’). as a wheel load is imposed radially and will strain the
geogrid/geotextile in all directions.
Geogrids and geotextiles operating with such a
to an increase in shear strength and stiffness. At some mechanism usually need to strain to the order of 2% in
point this increase is sufficient to support the road with order to start functioning, although due to the nature of
only minor or insignificant further settlement. some materials, strain at full tensile strength can be of
Floating roads can cause relatively little damage to the order of 7–9%. This can lead to deformation at the
the peat and can allow it to remain relatively undam- surface of the road leading to additional rutting (over
aged under the road line. However, on slopes, the and above that imposed by the axle passes themselves)
aggregate in the road can act as a drain and the detail which will need to be remediated with additional mate-
design of such roads on slopes needs careful consid- rial, reflected deformation down through the aggregate
eration to ensure that the road does not damage peat layer onto the formation (this is where the strain in the
bogs at the top of the slope. geogrid is occurring to support the surface loads) and
The use of stabilisation geogrids can have an advan- ponding of water at both the surface and at the forma-
tage in that they can permit a marked reduction in the tion surface (within the aggregate layers), which can
thickness of aggregate that needs to be used. Typically, lead to softening of the formation.
on peat with a CBR of less than 1% this reduction can Stabilisation: at present this mechanism is restricted
be up to a third. This reduction in road thickness is an to a single manufacturer (Tensar International Ltd)
advantage not only in cost terms but the reduction in and a single innovative product (TriAx) which is
dead-load leads to a reduction in overall road settle- a polypropylene, punched and drawn, monolithic,
ment post-construction leading to better serviceability hexagonal geogrid with triangular apertures.
and less remediation during use. This product works as a stiff geogrid with inte-
Displacement roads use large quantities of dumped gral nodes between the ribs to interlock with suitably
rock pushed (or allowed to settle in very poor ground sized aggregates and markedly reduce their tendency
conditions) into the surface of peat and that essentially to move when trafficked. The aggregate stabilised by
displace it horizontally, building up a pyramid of stone the geogrid aperture (in a ‘fully confined’ zone), in
that can eventually be used to found the road upon, the turn, provides stability to the aggregate layers above
base having reached (or the load of the stone having (in a partially confined zone) and typically this can
created through consolidation of underlying strata) a be to a distance of 350–500 mm above the geogrid
firmer horizon. layer (Fig 4); if a design requires a thicker aggregate
This method of construction is usually only used in layer, then it may be necessary to introduce a sec-
very wet and extremely poor ground conditions with ond layer of geogrid to ensure that the whole of the
virtually no shear strength and where a floating road aggregate thickness is either fully confined or partially
would become unserviceable in a short time due to set- confined.
tlement. It is also used where there are floating bogs The incorporation of integral nodes joining the
(vegetation and root mass several meters thick sitting geogrid ribs is important in that the almost 100%
on water). It is relatively expensive and time consum- junction efficiency they impart ensures that aggregate
ing and can only be used where there is no risk of particles remain interlocked and confined when loaded
damage to infrastructure to the sides of the road line; with stone; broken junctions (such as with welded
some vertical displacement (heave) of peat can occur products placed on very soft ground) are unable to
in rapid filling operations. interlock and confine the aggregate meaning that
133
Figure 4. Model of confinement for TriAx stabilisation
geogrid.
reinforcement from the intact ribs (with the associated

deformation) is the only mechanism that can prevent Figure 5. Estimation of standard axle passes for a typical
rutting. aggregate delivery vehicle.
The calculation of access road thickness incorpo-
rating TriAx stabilisation geogrids depends on the use safe (or necessarily economical) to substitute another
of empirical testing of various aggregate thicknesses unrelated geogrid into such designs without reassess-
and different formation stiffnesses. This testing allows ing the design. This information should come from
the calibration of designs incorporating TriAx with adequate testing both in the laboratory and in the field
respect to the ground conditions, expected trafficking to corroborate the design values.
and acceptable rut depth. Such data cannot be derived The Ev 2 method can lead to the development of
theoretically and involves many trials using wheeled designs that can be overly thick for roadway design on
traffic loading in order to determine design data that such sensitive ground. The Ev2 method aims to limit
can be used reliably and is a method advocated by deformation at the top of the aggregate layer to avoid
Giroud and Han (2004). It has proven to be a reliable overstressing blacktop. However, as the vast majority
and economical method of access road design. of wind farm access roads have an unbound surface,
Crucially, the stabilisation mechanism is a low- the reduction in deformation has a limited scope; most
strain mechanism and does not depend upon the tensile geogrid stabilised wind farm, access roads on peat
strength of the geogrid at a specified strain; confining will deform (largely elastically) under the passage of
and stabilising the aggregate particles generates strain vehicles and it is often more important to limit the
levels of the order 0.5% for a trafficked, correctly cal- differential movement between the axles of the vehi-
ibrated and designed access road. This results in less cles (especially very long vehicles delivering turbine
deformation in terms of rutting (which can also be lim- blades).
ited based upon the requirements of the design), less CBR target methods follow a similar methodol-
deformation at the formation surface (as the geogrid ogy but would also require some form of calibration
is straining much less than with the tension membrane to ensure target values are likely to be met by a
mechanism) allowing the formation to retain its shear geogrid and thickness of aggregate on a given subgrade
strength and function to help to support the road. A stiffness .
main aim of design with the stabilisation mechanism An alternative method which is permitted by the use
is to prevent deformation at the surface and protect the of geogrids such as TriAx which have been calibrated
formation. as described in section 3.2, is to assess the type and
amount for trafficking that is likely to enter the site
3.3 Access road design (rendering these in terms of ‘standard axles – either
Design of wind farm haul roads can use several 80 kN or 100 kN) and using this data to produce a suit-
methods of design, all of which are of empirical ori- able road thickness consistent with allowable rutting.
gins. Many of these methods can also be used with Axles that are not ‘standard axles’ are converted
geogrids in order to restrict the thickness of aggregate using the ‘4th power rule’ (RRL 1970) to standard
used in their construction over an unreinforced and axles as follows (80 kN example):
unstabilised construction. Ns = (actual axle loads/standard axle load)3.95
The Ev 2 method for reinforced or stabilised float-
ing roads usually uses design charts (or computerised Where Ns = vehicle equivalency in terms of standard
versions of such charts) with modifications to take axles.
account of the reduction in aggregate thickness that So for a typical standard 20t payload road-going
can be effected by incorporating one or more layers of vehicle, the vehicle imposes 5 standard axles worth of
geogrid. These design charts need to be calibrated for damage (in terms of rutting potential) upon the road
the specific product to be used and it is not entirely (Figure 5).
134
It should be noted that the standard axle loads for 3.4 Additional checks
an off-road articulated hauler can be much higher for
In all cases, a check should be made for the worst-
each visit.
case wheel load on an expected tyre footprint to check
The trafficking takes two forms:
against bearing capacity failure in a similar way to
the loads being checked on a working platform (see
Construction traffic – the traffic required to actually section 4). In some deep peat deposits, it may also be
build the access roads and working platforms. This necessary to take into account possible circular slip
traffic, which can include heavily loaded aggregate failure through the platform and into the peat deposit
delivery vehicles (articulated haulers and road-going for stationary loads. However, site operational rules
trucks) is by far the most onerous loading on most that dictate that no vehicle should remain stationary
wind farms in terms of trafficking and standard axles. on any floating road section can help to minimize road
On very large sites, other traffic is almost negligible thickness and reduce the requirement to conduct such
in terms of total standard axles. analysis; it is difficult to obtain appropriate parameters
for such assessment in peat due to the soft nature of
In-service traffic: the traffic that uses the access the deposits and difficulty in sampling and testing in
roads to access and build other elements, such as the laboratory.
concrete and steel delivery trucks for turbine bases,
cranes for turbine construction and turbine mast, blade
and nacelle delivery vehicles. On smaller sites, such
vehicles (especially concrete delivery wagons) should 4 WORKING PLATFOMS
always be included in an assessment of trafficking as
they can form a higher proportion of the total traf- 4.1 Working platform types
fic than on larger sites and may increase the road and All working platforms will require a certain number
platform thicknesses as a result. Pick-ups and light of vehicle visits in order to construct them, in addi-
vans can be excluded from this assessment as they tion to the vehicles that will use them operationally,
impose relatively little damage in terms of standard such as piling rigs and cranes. The total number of
axles, vastly simplifying the assessment. axles needs to be taken into account in any trafficking-
All traffic that is expected to use the site is assessed related design assessment and added in to the road
as above in terms of standard axles, the total number of thickness calculations so that the additional traffic is
stared axles required is totaled and each section of road accounted for in the road thickness calculation.
is designed, based on the empirical trafficking testing There are generally two types of working platforms
conducted previously, to give an aggregate thickness used on wind farms, mainly at turbine lifting locations:
when incorporating an appropriate TriAx geogrid to Working platforms founded on a firm stratum:
suit this total. generally these will suffer from little in the way of
For some wind farms, the method can become settlement and can often be dig-out and replace, replac-
quite involved; the access roads are designed from the ing the soft, near surface material with aggregate or
furthest point away from the aggregate access point less compressible material with a geogrid stabilised
and then designed back towards the access points such top surface. Such platforms may need to be assessed
that the trafficking becomes cumulative – the parts for edge stability if surrounded by deep peat or similar
of the access road taking the most traffic are then soft deposits.
assured of having sufficient thickness to ensure their Floating working Platforms: these are more contro-
performance consistent with the total axle loads and versial and less commonly used. On very soft and deep
trafficking, yet the least trafficked sections (typically peat, it is difficult to justify their use for crane lifts as
spurs to single turbines) have less thickness to give edge stability and rotational failure of the platform
the most economical construction for the respective (which any incorporated geogrid will be hard-pushed
sections. to mitigate) may become dominant failure mecha-
It should be noted that all of these methods can only nisms leading to catastrophic failure of the platform
give the required performance with a certain minimum and a potential to lose the crane.
quality of aggregate. Poor quality, crushable, rounded However, on smaller wind farms with more lim-
and poorly graded aggregates can all lead to a road ited financial resources, it may be possible to build a
that underperforms when used in a geogrid-stabilised working platform, say, designed for the piling rig for
or reinforced road. If it is desired to use such mate- the ‘can’ for the base of the turbine and then use the
rial for economic reasons, the designer needs to be same piling rig at the outrigger positions for the lifting
informed before design commences so that allowances crane. However, it should be noted that this limits the
may be made for the sub-optimal aggregate, usually by type of crane that might be able to use the platform for
increasing the road thickness. any subsequent maintenance or turbine replacement; it
On very soft peat formations, the use of good will need to have outriggers in a similar arrangement
quality aggregate will ensure a minimal road to that of the crane that the platform was originally
thickness which is likely to reduce settlement of designed for. Piles also need to be designed in order
the road after construction and ensure long-term to deal with a lack of lateral support from peat in deep
serviceability. deposits and also that the platform may settle relative
135
of comfort; they do not represent a realistic test of
the loads and area that a large crane can impose on
an outrigger plate (which may be 200 t on an area of
2.5 m × 6 m), nor the depth of influence of such a load.
Tests conducted on some sites involve the main lift
crane raising an unloaded boom and rotating to place
the full counterweight load over each of the outrigger
in turn, often for several hours, and monitoring settle-
ment. This is a far more onerous and realistic test of
such a working platform and gives better confidence
of the performance during a lift.
5 CONSTRUCTION OF ACCESS ROADS AND

WORKING PLATFOMS ON SOFT GROUND
Where wind farm access and working platform con-

Figure 6. Working platform with piled outrigger positions struction takes place on very soft ground, cognizance
on very deep soft peat showing settlement around piles should be taken of the speed of construction.
caps due to consolidation. Geogrids are unable to prevent With peats, care should be taken to avoid load-
such movement but can stiffen the platform to minimize ing the surface very rapidly with the full construction
differential settlement. thickness, especially in very wet ground, or where the
vegetation and root mat is relatively thin, as this can
to the pile cap and pile, requiring remediation to ensure induce shearing of the vegetation and root mat near sur-
serviceability (Fig 6). face, which is often where much of the shear strength
Working platforms are required to have an lies. Geogrids are unable to compensate for this rapid
allowable bearing pressure of 200–260 or an EV 2 construction as the shear plane usually lies right on the
of 100 to 120 MPa in many turbine manufacturers’ edge of the road pyramid and outside the zone of influ-
specifications. ence of the geogrids themselves. However, on sites
where this has been observed by the authors, it does not
4.2 Working platform design appear to be detrimental to access road performance,
although the open shear plane can be disconcerting.
There are a number of methods for working platform In a similar fashion, rapid construction over sensi-
design. In the UK, guidance in the form of BR470 tive clays (especially near-saturated weathered glacial
(BRE 2004) produces safe but relatively uneconomic till clays) can lead to a marked reduction in shear
designs. A review of this document produced in 2011 strength (by up to more than 50%) and especially
suggested that alternative methods could be used pro- where compaction plant is used on relatively thin lifts
viding they were validated by appropriate laboratory of aggregate fill above the formation; such sensitivity
and field testing (BRE 2011). should be picked up during the ground investigation
Although EV 2 design methods can be used to stage. Geogrids are again unable to deal with such cir-
achieve a specified EV 2 at the top of the platform, cumstances as the loss in shear is a function of the
some geogrid manufacturers use their own in-house formation materials response to construction stresses.
methods to assess such platforms. This involves the This usually reveals itself by marked have during com-
use of a ‘load spread’ method where the expected pres- paction and under such circumstances, construction
sure on a specified loaded area is dissipated through a must be halted and the section of road left undisturbed
design platform thickness through an angle established until the excess pore-water pressure has reduced and
through appropriate laboratory and field testing to get some of the shear strength has been restored; further
a pressure at formation level that gives an acceptable work will only make the situation worse and requires
factor of safety against bearing capacity failure. a reassessment of the construction method. It is usu-
This is usually done by iterative calculation and can ally better to track-in at least the first few layers of
lead to an economical design that still maintains the aggregate instead of using formal compaction.
desired safety in such platforms.
4.3 Working platform testing

6 CONCLUSION
Working platforms are frequently tested using either
EV2 (DIN1834 2001) tests or BS1377:Part 9:1990 For in excess of 20 years, geogrids have provided a safe
(BSI 1990) plate bearing tests. Whilst such tests are and economical solution for the construction of wind
relatively easy to conduct and might be appropriate for farms on soft ground in upland areas. With the correct
tracked plant and very small outrigger bearing plates level of appropriate ground investigation data, access
on small cranes, even the largest plates commonly used roads and working platforms can be designed incorpo-
in such test (750 mm or so) can only provide a degree rating stabilisation geogrids such as TriAx to minimize
136
construction thickness consistent with trafficking and BRE, 2004. ‘Working Platforms for Tracked Plant’, UK
bearing capacity requirements. Building Research Establishment.
The use of these methods will ensure serviceability BRE, 2011. ‘Working Platforms for Tracked Plant – Use of
of the constructed areas throughout the construction Structural Reinforcement – A BRE Review Seven Years
On’ UK Building Research Establishment.
phase and later design life of the wind farm. The abil- DIN1834 2001-09, 2001. ‘Determination and Strength Char-
ity to minimize construction depths can also lead to acteristics of Soil by the Plate Loading Test’.
a significant saving in the amount of CO2 generated Cook, J., Dobie, M. & Blackman, D. 2016. ‘The Development
by such activities from the construction and quarrying ofAPT methodology in theApplivcation and Derivation of
plant as this plant is having to quarry, transport, place Geosynthetic benefits in Road Design’, Accelerated Pave-
and compact less material compared to non-stabilised ment Testing Conference, Costa Rica, September 2016 (in
construction. print).
Giroud, J. P. & Han, J. (2004) ‘Design Method for Geogrid-
Reinforced Unpaved Roads. Part II. Calibration and
Applications’, Journal of Geotechnical and Geoenviron-
REFERENCES mental Engineering.
RRL, 1970. ‘A Guide to the Structural Design of Pavements
SNH, 2010. ‘Floating Roads on Peat’, Scottish Natural
for New Roads’. UK Department of Environment Road
Heritage.
Note 29, 3rd Edition, UK Road Research Laboratory.
Scottish Renewables, 2010. ‘Good Practice during Windfarm
BSI, 1990 BS1377:Part 9:1990 ‘British Standard Methods
Construction’, Scottish Renewables, Scottish Natural Her-
of Test for Soils for Civil Engineering Purposes – In-Situ
itage, Scottish Environment ProtectionAgency & Forestry
tests’.
Commission Scotland.
137
Static analysis of Mansa Devi hill landslide using FLAC3D
S. Islam
Department of Civil Engineering, King Khalid University, Abha, KSA
Z.H. Rizvi
Department of Geosciences, University of Kiel, Germany
ABSTRACT: The present study mainly deals with the analysis of Mansa Devi landslide, Haridwar, Uttarakhand,
India. The area experiences local as well as regional slides every year. The experimental work were conducted to
determine the various mechanical property of rock mass. These properties have been used as input parameters
for the numerical simulation of slope using FLAC3D. The deformations in all the three direction and stresses
along xx and zz axis as in these direction movement take place have been calculated. Finally factor of safety of
study area is calculated. The study indicates that slope at the higher height along the slope is unstable and it is
stable at the lower height. The effects of instability have been thoroughly considered.
1 INTRODUCTION popular for slope stability analysis in situations where

the failure mechanism is not controlled completely
Over the last decade, we have had numerous land- by discrete geological structures (Monjezi and Singh
slides culminating in tragedies and it is no coincidence 2000, Sarkar et al. 2010, Singh andVerma 2007,Verma
that they all happened in hilly region. The fact is, and Singh 2010a,b).
the slopes are inherently unstable. While erosion in the Application of numerical simulation is increasing
form of soil runoffs occurs daily, slopes can collapse due to many advantages over limit equilibrium method
and result in destructive landslides. Such landslides (Griffiths and Lane 1999; Duncan 1996). Griffiths and
are common problems in the mountainous regions Lane (1999) summarized the results of a slope stability
particularly along the hill-cut slope of the Lesser analysis using FEM based on shear-strength reduc-
Himalaya. Landslide accounts for large number of tion technique and provided information concerned
human casualties and an enormous property loss, the advantage and disadvantage of FEM for use in a
and creates severe damage to natural ecosystems practical slope engineering problem. The discrete ele-
and human-built infrastructures (Dai and Lee, 2002; ment method (DEM) was introduced by Cundall and
Guzzetti et al., 2005). Many models for landslide Strack (1979) explained the mechanical behavior of
susceptibility mapping have been proposed with the assemblies of discs and spheres. Wang et al. (2003)
assumptions that landslide susceptibility is related to used the Particle Flow Code (PFC) to carry out a gen-
causal factors and can be evaluated as long as the eral study on the stability of heavily jointed rock slope.
causal relationship is known (Zhu et al., 2014) Sung and Chung (2005) did stability analysis of jointed
The stability of large rock slopes which are prone rock slopes using the Barton–Bandis (BB) constitutive
to rock mass (rotational) failures can be analyzed model using UDEC.
by traditional limit equilibrium methods, such as In the present article, three-dimensional finite dif-
Bishop’s method (Bishop1955), Janbu’s simplified ference methods are applied. The deformations and
methods (Janbu 1954), or the later improved sophis- stability of potentially vulnerable slope investigated
ticated methods (Janbu 1957, Spencer 1967) based and a quantitative measure for stability is provided in
on assumptions regarding the inclination and location terms of factor of safety.
of the interslice forces. In limit equilibrium method,
the state of stress in the rock mass can easily be
determined both before and after the formation of the 2 STUDY AREA
slope. The drawbacks of this method are that contin-
ues loading cannot be simulated easily and step loading The study was carried out on the east facing slope of
and deloading is not incorporated (Verma and Singh Mansa Devi hills in Hardwar, Uttaranchal . The area
2010a). comes under toposheet number 53 k/1 of the Survey
Numerical methods such as the finite element of India. The soil sample is taken from the location
method (FEM), the finite difference method (FDM), 29◦ 57 57.9 N, 78◦ 10 9.03 E, 395 m elevation.the
the boundary element method (BEM), and the distinct area is taken into consideration as it is regularly dam-
element method (DEM) are becoming increasingly aged and there is very high probability of landslide
139
Figure 1. Study Area.
Figure 3. Plot of FoS vs height at a distance of 192 m from
Toe of Hill.
parameters that are required by FLAC3D for Mohr-

Coulomb model are bulk modulus, shear modulus,
cohesion and friction angle.
The representative soil sample was collected from
the site and used for various test such as sieve analysis,
attenberg limits, compaction test as well as direct shear
test in order to determine input parameter for the anal-
ysis purpose. wet sieving procedure was performed by
taking 200g of oven dried soil, passing 4.75 mm I.S
sieve the result indicates that there is 39.2% sand and
60.8% of fine sand (Mittal 2008) . The liquid limit
come out to be 23.26% and plastic limit obtained after
oven drying was 14.93% and plasticity index come
out to be 8.33%.By comparing above result with IS
Figure 2. Geometry. 1498-1970 the soil is of type clay of low compress-
ibility(Prakash 2002). Further by compaction test it
was found that the optimum moisture content was 11
occurrence in future with possible loss of human lives and maximum dry density was 1.889 g/cc. The result
and property. of shear parameters by doing direct shear test obtained
as c = 4000 kg/cm2 and Ø = 31◦ .
3 MODELLING AND ANALYSIS

5 RESULT AND DISCUSSION
The geometry of the soil domain is created using no
of brick shape element and wedge element The size The results obtained from static analyses are summa-
of the Mansa Devi model is considered as x = 227 m rized in the following figures. The output obtained
y = 10 m and z varies from 0 to 136 m dependeng upon in all the analyses are in terms of displacement
variation of slope. The entire grid is discretized into magnitude contours, x-displacement, y-displacement,
7000 number of zones and 8481 number of grid points. z-displacement, stresses in xx and zz direction. Finally
For the static analysis, the boundary conditions speci- FOS showing variation along the height of the Mansa
fied were as follows. At the bottom plane of the grid all Devi hill at different location from the toe of the hill
movements are restrained. The planes at X = 0 m are has been plotted.
free to move in the Y and Z directions but not in the From the analysis of displacement magnitude con-
X direction. Similarly, the planes Y = 0 and Y = 10 m tour obtained from static analysis, it is evident that
are free to move in X and Z directions but not in the Y the maximum displacement of magnitude 94.57 cm
direction. occurs at the top of heel side of the study area. Also
from the figure of x-displacement contour it is clear
that maximum X-displacement occurs at the downs-
4 MATERIAL PROPERTIES lope just adjacent to the road. The maximum value of
X-displacement is observed to be 18.295 cm. More-
In the present analysis, Mohr-Coulomb plasticity over Y-displacement is found to be of negligible value.
model has been used for the soil medium. The input Also figure shows displacement vector from which it
140
Figure 4. Plot of FOS vs Height at a Distance of 164 m from Figure 7. Plot of FOS vs Height at a Distance of 93 m from
Toe of Hill. Toe of Hill.
Figure 5. Plot of FOS vs Height at a Distance of 142 m from Figure 8. Plot of FOS vs Height at a Distance of 45 m from
Toe of Hill. Toe of Hill.
From the figure of XX stress it is observed that

the tensile stress occurs along the top of the slope.
The maximum observed value of tensile stress is
26.019 kPa. Similarly it can be seen from the fig-
ure of ZZ stress contour that the tensile stress occurs
along the top of the slope whose maximum value is
5.469 kPa.
The contour of FOS wrt Elevation and Height is
plotted which indicates it is unstable at higher height
and is stable at lower height showing trends of circular
failure. Also the plots of FOS vs Height at various
position along X- axis from the toe of hill confirms
the same results.
6 CONCLUSION
Figure 6. Plot of FOS vs Height at a Distance of 138.5 m
from Toe of Hill. Our study analyses the instability of the most vulner-
able slope of our study area along the road slopes in
can be conclude that the soil at downslope is sliding the Mansa Devi Hill, (Hardwar) Uttarakhand region.
forward whereas settlement is observed at upstream The slope was investigated for the stability using three
slope. The maximum value of Z-displacement is found dimensional finite difference codes. The slope under-
to be 94.675 cm. goes circular failure mechanism as obtained by the
141
analysis. The factors of safety of the slope at some fourth international conference on soil mechanics, vol 2,
location are coming out to be less than one, which pp 207–212 Mittal, S., Shukla, J.P., (2008) “Soil testing
confirms its instability. The factor of safety along the for engineers” pp 21–45
slope is high at low heights and slope is quite stable. It Monjezi, M. and Singh, T.N., 2000, Slope instability in an
opencast mine. Coal International,8, pp. 145–147
then decreases due to positive shear strain rate and then Prakash, S., Jain, P.K., (2002) “Engineering soil testing”
again increases at higher heights due to negative shear pp 16–17
strain rate. This trend of the factor of safety confirms Sarkar, R.K., Singh, T.N. and Verma, A.K., 2010, A numerical
the path of circular failure. simulation of landslideprone slope in Himalayan region –
Hence what we can say on the wholistic point is that a case study. Arabian Journal of Geosciences, DOI
there are ongoing deformations at the Mansa hill and 10.1007/s12517-010-0148-8
the analyzed slope is critically unsafe to the inhabitants Singh, T., N., Gulati, A.,Dontha, L., Bhardwaj, V., (2008)
of the area. “Evaluating cut slope failure by numerical analysis- a case
study” pp 263–279
Singh, T.N. and Verma, A.K., 2007, Evaluating the slope
REFERENCES instability of the Amiyan Slide. In Rock Mechanics Meet-
ing Society’s Challenges and Demands, E. Eberhardt, D.
Bishop AW (1955) The use of the slip circle in the stability Stead and T. Morrison (Eds), pp. 993–998 (London: Taylor
analysis of slopes. Geotechnique 1:7–17 and Francis)
Cundall PA, Strack ODL (1979) A discrete numerical model Sung OC, Chung SK (2005) Stability analysis of jointed
for granular assemblies. Geotechnique 29:47–65 rock slopes using the Barton–Bandis constitutive model
Dai, F.C., Lee, C.F., 2002. Landslide characteristics and slope in UDEC. Int J Rock Mech Min Sci 41:469
instability modeling using GIS,Lantau Island, Hong Kong. Spencer E (1967) A method of analysis of the sta-
Geomorphology 42, 213–238 bility of embankments assuming parallel interslice
Duncan JM (1996) State of the art: limit equilibrium forces.Geotechnique 17(1):11–26
and finite-element analysis of slopes. J Geotech Eng Verma, A.K. and Singh, T.N., 2010a, Assessment of tunnel
122(7):577–596 instability – a numerical approach. Arabian Journal of
Frohlich OK (1955) General theory of stability of slopes. Geosciences, 3, pp. 181–192
Geotechnique 5:37–47 Verma, A.K. and Singh, T.N., 2010b, Modeling of a jointed
Guzzetti, F., Reichenbach, P., Cardinali,M., Galli, M., Ardiz- rock mass under triaxial conditions. Arabian Journal of
zone, F., 2005. Probabilistic landslide hazard assessment Geosciences, 3, pp. 91–103
at the basin scale. Geomorphology 72, 272–299 Wang C, Tannant DD, Lilly PA (2003) Numerical analy-
Griffiths DV, Lane PA (1999) Slope stability analysis by finite sis of the stability of heavily jointed rock slopes using
elements. Geotechnique 49(3):387–403 PFC2D. Int J Rock Mech Min Sci 40:415–424 (Elsevier
Hudson JA, Harrison JP (1997) Engineering rock mechan- Publishers)
ics: an introduction to the principles. Elsevier Science Zhu, A.X.,Wang, R.X., Qiao, J.P., Qin, C.Z., Chen, Y.B.,
Publishers, Oxford Itasca Consulting Group, Inc. (2003) Liu, J., Du, F., Lin, Y., Zhu, T.X., 2014. An expert
“FLAC3D user’s guide” knowledge-based approach to landslide susceptibility
Janbu N (1954) Application of composite slip surfaces for mapping using GIS and fuzzy logic. Geomorphology 214,
stability analysis. Proceedings of European conference on 128–138
stability of earth slopes, vol 3, pp 43–49 Zheng H, Liu DF, Li CG (2005) Slope stability analysis
Janbu N (1957) Earth pressure and bearing capacity by based on elasto-plastic finite element method. Int J Numer
generalized procedure of slices. Proceedings of the Methods Eng 64:1871–1888
142
The influence of natural convection on thermal performance

of ground heat exchangers
A. Bidarmaghz & G.A. Narsilio

The University of Melbourne, Parkville, Victoria, Australia
ABSTRACT: Ground-source heat pump (GSHP) systems efficiently heat and cool buildings using sustainable
geothermal energy accessed via ground heat exchangers (GHEs). Thermal performance of GSHP systems is
typically investigated considering either pure conduction in the ground or also accounting for hydro-geological
conditions (i.e., groundwater flow). However, in saturated soils, the temperature gradient in the ground induced
by the GSHP systems operation may result in natural movement of groundwater due to the changes in water
density, which leads to an emerging natural convective heat transfer in the ground, potentially influencing
thermal performance of GHEs. To capture and quantify this effect, a GHE-field installed in a fully saturated
soil is modelled using a state-of-the-art 3D FE model. In these simulations, groundwater flow in the ground
and the convective-conductive heat transfer and fluid flow in the fluid circulating in the pipes are coupled to the
convective-conductive heat transfer in the ground and the GHEs.
1 SHALLOW GEOTHERMAL 2 NUMERICAL MODEL DEVELOPMENT

TECHNOLOGIES
2.1 Governing equations
Most of the existing methods and studies to predict
In general, for geothermal systems in fully saturated
thermal performance of GHEs have considered purely
soil with no pressure driven groundwater flow, heat
conductive heat transfer in the ground or conductive-
conduction dominates in the soil, GHE backfilling
convective heat transfer with pressure-driven convec-
material (e.g., concrete) and high density poly ethy-
tion, which occurs when there is groundwater flow
lene (HDPE) pipe wall, and partially in the carrier
(Chiasson et al., 2000, Diao et al., 2004). However,
fluid (water) that these absorber pipes contain in the
even in the absence of ground water flow in satu-
GHEs. Heat convection occurs in the carrier fluid
rated soils, the temperature gradient induced by heat-
circulating in the pipes as well as in the ground
ing/cooling operations of GSHP systems may result in
due to the natural flow of groundwater. In this new
buoyant flow of groundwater due to density changes
model, the density driven flow in the ground and
of (ground) water with temperature and consequently
the convective-conductive heat transfer and fluid flow
induced natural heat convection in the ground may
in the fluid circulating in HDPE pipes are coupled
occur. To investigate the effects of natural convection
to the convective-conductive heat transfer that takes
on thermal performance of GHEs, a 3D numerical
place in the ground and the GHEs. To model the fluid
model is developed. The model considers different car-
flow inside the pipes, the continuity and momentum
rier fluid velocities in the GHEs and the efficiency of
equations for incompressible fluid are used as follow
GSHP systems subjected to natural heat convection
(Barnard et al., 1966):
are compared to the cases where only pure conduc-
tive heat transfer is considered in the ground. The
governing thermo-hydro equations are implemented
in the finite element package COMSOL Multiphysics
to couple the heat transfer and fluid flow in the ground
(porous medium) and the GHEs to the fluid flow and
heat transfer in the carrier fluid circulating in the pipes where A[m2 ] is the inner cross-section of the HDPE
of GHEs (Comsol, 2015). pipe, ρw [kg/m3 ] is the carrier fluid density, v1 [m]
This new model is validated against numerical represents the fluid velocity field, t shows the time,
data from a well-documented study for an extreme p [Pa] is the pressure, fD represents the Darcy fric-
condition of low hydraulic conductivity of the soil, tion factor and dh [m] is the hydraulic diameter
subsequentially relaxed (Lee and Lam, 2008). of the pipe. The energy equation for the fluid to
145
describe the convective-conductive heat transfer for the soil and the groundwater , Cp,f and Cp,m [J/(kgK)]
an incompressible fluid is (Lurie, 2008): are the groundwater and the soil specific heat capaci-
ties, n is the porosity and ρm [kg/m3 ] is the soil density.
Equations 7 and 8 are coupled to density driven flow
in the soil via Darcy velocity v2 .
2.2 Initial and boundary conditions

To solve the above coupled governing equations,
appropriate initial and boundary conditions must be
where Cp,w [J/(kgK)] is the specific heat capacity of provided. For this study these conditions are as follows
the fluid, λw represents the thermal conductivity of the (Fig. 1):
fluid [W/(mK)] and Qwall [W/m] is the external heat A constant initial temperature equal to undisturbed
exchange rate through the pipe wall and it is a function ground temperature, T0 is defined to the entire system
of the temperature of the pipe outer wall, T(m,pipe,wall) including the far field boundary of the ground and the
[◦ C], the temperature of the carrier fluid, T [◦ C], the carrier fluid inside the pipes (T = Tm = T0 = 20◦ C).
pipe thermal conductivity λpipe [W/(mK)] and diam- No flow boundary condition is defined for all the
eter dpipe [m] (Batini et al., 2015, Bidarmaghz, 2014, external boundaries (−n1 · ρ v2 ) = 0. where n1 is a unit
Narsilio et al., 2016). The above equations are solved normal vector.
for pressure, velocity field and temperature field in the Thermal insulation boundary condition is applied
carrier fluid and are coupled to the temperature field to the top and bottom of the model (−n1 · q = 0).
Tm [◦ C] obtained from the conductive heat transfer A constant heat flux of 3300 W (for the single bore-
equations solved for the soil, the GHEs and the pipes: hole) or a constant inlet temperature of 2◦ C (for the
borefield case) is assigned to the inlet pipe.
Numerically solving these systems results in tem-
perature, velocity and pressure distributions of the
ground, groundwater, GHEs, pipes and the carrier
where ρm [kg/m3 ] represents the solid material den- fluid. It should be noted that in all simulations the
sity, Cp,m [J/(kgK)] and λm [W/(mK)] represent spe- ground domain is large enough to avoid edge effects.
cific heat capacity and thermal conductivity of solid The key input parameters used in the numerical
material respectively. models in this work are shown in Table 1.
The single phase fluid flow in a porous medium
is usually described by Darcy’s law, which states that
the Darcy velocity field, v2 [m/s] is determined by the 3 NATURAL CONVECTION EFFECTS:
total head gradient, ∇(p2 − ρf gz), the fluid dynamic RESULTS AND DISCUSSIONS
viscosity, µf [Pa.s], and the structure of the soil:
Using the newly developed model, a 110 m long sin-
gle U-loop GHE is numerically modelled using the
governing equations summarised in Section 2. and
validated against a publish work with pure conduc-
where K [m2 ] is the permeability of soil, p2 [Pa] repre-
tion in the ground (Lee and Lam, 2008). The key input
sents pressure, ρf [kg/m3 ] is the groundwater density,
parameters used in this modelling were shown in Table
kh [m/s] is the porous medium hydraulic conductivity
1 (single). The model is solved for various ground
and g [m/s2 ] is the gravitational acceleration vector.
hydraulic conductivities, ranging from 0 to 10−2 m/s.
The continuity and momentum equation to model
Figure 1a shows details of the geometry and boundary
the flow in porous medium is as follows:
conditions used in these validations. Figure 1c shows
a schematic of the single U-pipe GHEs used in the
simulations.
Equation 5 and 6 are solved for Darcy velocity field Numerical results presented in Table 2 (columns
and pressure in the soil. To model heat transfer in the 2 to 4) and Table 3 suggest that for kh ≤ 104 m/s,
soil, an energy equation is used. It is assumed that the the soil convective heat flux is so small that can be
fluid and the soil phases are in thermal equilibrium. neglected in numerical simulations as the results match
Therefore, the energy conservation equation in the soil the published case with pure conductive heat trans-
is written as: fer (Lee and Lam, 2008). However, a porous ground
with kh ≥ 10−3 m/s shows up to 1.7◦ C lower tempera-
ture rise in the ground (taken from 1m radial distance
from the GHE) in cooling mode and this difference
increases to about 3.7◦ C when kh = 10−2 m/s. This
represents a significant amount of thermal energy,
which is simply ignored when natural convection in
where Tm [◦ C] is the temperature field in the soil, λm the ground is not considered. Therefore, accounting
and λf [W/(mK)] represents thermal conductivity of for the densitydriven flow and natural convection in
146
Table 1. Key input parameters used in numerical simulation.
λ Diameter
Single Borefield Cp ρ Single Borefield Spacing
W/(mK) J/(kgK) kg/m3 m m
Soil 3.5 1.5 1,000 2,160 – –

Concrete boreholes 1.3 1.3 1,100 2,300 0.11 0.13 7.5
HDPE (pipe) 0.4 0.4 – – 0.032 (SDR 21) 0.032 (SDR 21) 0.06
Carrier fluid 0.582 0.582 4,190 1,000 – – –
∗
Saturated soil porosity n = 40%.
Table 2. Comparison of average ground temperature rise for different hydraulic conductivities.
Average ground temperature rise at mid-depth of borehole (◦ C)
Conductive-convective heat transfer

Distance from Conductive only
borehole heat transfer∗ kh = 0 m/s kh = 10−5 m/s kh = 10−4 m/s kh = 10−3 m/s kh = 10−2 m/s
Borehole 10.0 10.6 10.7 10.8 9.0 6.5

1m 5.5 5.7 5.7 5.7 3.8 1.8
2m 4.3 4.5 4.5 4.5 2.6 0.9
5m 3.0 2.8 2.9 2.8 1.2 0.3
10 m 1.5 1.6 1.6 1.8 0.5 0.1
∗
Data from Lee and Lam (2008).
Table 3. Comparison of average borehole temperature rise

for different hydraulic conductivities.
Average borehole
temperature
Model rise (◦ C)
Pure conduction (Lee and Lam, 2008) 10.1

Conduction-convection kh = 0 m/s 10.7
kh = 10−5 m/s 10.8
kh = 10−4 m/s 10.9
kh = 10−3 m/s 9.0
kh = 10−2 m/s 6.5
the ground with kh ≥ 10−3 /s helps the GSHP system Figure 1. Geometry of (a) single, (b) multiple GHE field
to deliver the required thermal load (30 W/m) with a (borefield) and the associated 3D FE meshes and (c)
lower temperature change being induced by the sys- schematic of a single U-pipe GHE.
tem. The hydraulic conductivity of the borehole is kept
at 10−10 m/s in all models.
To investigate how carrier fluid velocity inside
the pipes potentially affects natural convection in the of ground hydraulic conductivities between 10−3 and
ground, a shallow GHE-field consisting of sixteen 10−5 m/s. Figure 2 shows the resulting temperature
30 m long single U-loop GHEs is built and solved in distribution around the GHEs at a plane at mid-depth
heating mode (i.e., extracting heat from the ground) of the GHEs for different soil hydraulic conductivi-
for a wide range of carrier fluid velocities (from lam- ties. It is observed that for higher kh , the volume of
inar to turbulent). Geometry and boundary conditions soil thermally affected by the borefield operation is
used in these simulations are shown in Figure 1b, c. significantly smaller and the required thermal load
Due to the existence of 2 planes of symmetry, only one is provided with a lower temperature change of the
quarter of the geometry is modelled to save computa- ground. However, when kh decreases, a larger soil vol-
tional time and memory.The key input parameters used ume around the GHEs will be thermally affected as
in these simulations were presented in Table 1 (bore- the ground requires a higher temperature change to
field). These simulations are conducted for a range transfer heat to the GHEs.
147
soils due to groundwater density variations induced by
ground temperature changes during GSHP operations.
The model is initially validated against data from a
well-documented study with good agreement between
the results for non-convective cases. Results show that
when kh ≥ 103 m/s, natural convection plays a signifi-
cant role in GHEs thermal performance. Moreover, in a
sixteen borehole GHE-field case study, the velocity of
the carrier fluid inside the pipes influences the induced
convective heat flux in the soil. The thermal capacity
Figure 2. Temperature distribution around the GHEs,
obtained from the GHEs increases more significantly
(a) kh = 10−3 m/s, (b) kh = 10−4 m/s and (c) kh = 10−5 m/s.
with increasing velocity and as soil hydraulic conduc-
tivity increases. The difference between the thermal
capacities obtained from soils with different hydraulic
conductivities is more pronounced for carrier fluid
velocities in the turbulent range. In summary, ignoring
the amount of thermal energy derived from the induced
natural convection in saturated coarse soils may result
in a significant over design of GSHP systems (GHE
length, diameter and number of boreholes). This letter
aims to highlight this finding and lay the foundations
for further research in this area.
ACKNOWLEDGEMENTS
Figure 3. Borefield (16 GHEs): GHEs thermal power
(per GHE). The authors would like to acknowledge the financial
support from the Australian Research Council (ARC
According to the results (Fig. 3), the effect of soil FT140100227) and The University of Melbourne.
hydraulic conductivity is more pronounced for higher
carrier fluid velocities (i.e., larger Reynold numbers
in the HDPE pipes). Moreover, the rate of change of REFERENCES
thermal power increases with increasing soil perme-
Barnard, A., Hunt, W., Timlake, W. & Varley, E. (1966) A the-
ability (up to 95% higher thermal load is obtained
ory of fluid flow in compliant tubes. Biophysical Journal,
from the GHEs when kh = 10−3 m/s in comparison to 6, 717–724.
the case where kh = 10−5 m/s). As Reynold number Batini, N., Loria, A. F. R., Conti, P., Testi, D., Grassi, W.
increases and with the inlet temperature prescribed at & Laloui, L. (2015) Energy and geotechnical behaviour
2◦ C, the average carrier fluid temperature inside the of energy piles for different design solutions. Applied
pipes decreases (as a result of a higher fluid veloc- Thermal Engineering, 86, 199–213.
ity inducing a lower temperature difference between Bidarmaghz, A., (2014) 3D numerical modelling of ver-
the inlet and outlet of the U-loop pipes). Therefore, tical ground heat exchangers. PhD, The University of
the temperature gradient between the GHEs’ pipes Melbourne, Melbourne, Australia.
Chiasson, A. D., Rees, S. J. & Spitler, J. D. (2000) A prelim-
and the surrounding ground increases which triggers
inary assessment of the effects of groundwater flow on
more pronounced groundwater density changes and closed-loop ground source heat pump systems. Oklahoma
results in a higher Darcy velocity in the groundwater State Univ., Stillwater, OK (US).
and a more significant natural convection emerging Comsol (2015) User’s guide. Version: 5.1.
in the ground. The figure also shows that increasing Diao, N., Li, Q. & Fang, Z. (2004) Heat transfer in ground heat
the carrier fluid velocity (from an initial Re = 870) exchangers with groundwater advection. International
for soils with higher hydraulic conductivity results in Journal of Thermal Sciences, 43, 1203–1211.
about 57% increase in thermal power obtained from the Lee, C. & Lam, H. (2008) Computer simulation of bore-
GHEs. However, this reduces to about 38% and 24% hole ground heat exchangers for geothermal heat pump
systems. Renewable Energy, 33, 1286–1296.
for kh = 10−4 m/s and kh = 10−5 m/s respectively.
Lurie, M. V. (2008) Modeling and Calculation of Stationary
Operating Regimes of Oil and Gas Pipelines, Weinheim,
Germany, Wiley-VCH Verlag GmbH & Co. KGaA.
4 CONCLUSIONS Narsilio, G., Bidarmaghz, A., Colls, S. & Johnston, I. (2016)
Geothermal energy, detailed modelling of ground heat
A FE model is developed to couple heat transfer and exchangers. Computer and Geotechnics, under review.
fluid flow in the GHEs and additionally to couple
these to heat transfer and fluid flow in the ground.
The model captures natural convection in saturated
148
Geological models as a tool for the planning of deep geothermal projects
F. Hese, K. Lademann, P. Offermann, A. Schaller & C. Thomsen

LLUR SH, Geological Survey of Schleswig-Holstein, Flintbek, Germany
ABSTRACT: The prerequisite for a successful and effective operation of a geothermal facility is the choice of
the proper location based on the knowledge of geological conditions and processes. The Geological Survey of
Schleswig-Holstein supports the planning of future geothermal projects by evaluating the geothermal potential
of the deep underground. The main focus is to study sandstone formations within the middle Buntsandstein,
upper Keuper and middle Jurassic time periods and deep reaching fault zones. A state wide geological model
has been developed to map areas, which may be suited for geothermal applications and to provide a basis for
subsequent studies and applications. Geological modelling and reservoir characterization is carried out within
the scope of the research projects GeoPower, GeotIS-StörTief and TUNB.
1 INTRODUCTION
The Geological Survey of the federal state Schleswig-

Holstein (LLUR SH, Dept. 6 – Geology and Soil)
aims to support the development of a sustainable
and renewable energy/heat production. One focus is
the investigation of the geothermal potential of the
deep underground to provide planning tools for future
geothermal projects. Further interests are to study
potential locations for the storage of energy (e.g. pres-
surized air in salt caverns) and in general to facilitate
the discussion on underground use regulations.
The onshore part of the federal state Schleswig-
Holstein covers the northern part of the North German
Basin, enclosing one of the deepest subsidence
centers – the Glückstadt Graben. The NNE-N trend-
ing Graben is reaching depths of 6 to 10 km and
is characterized by abnormal high thicknesses of
Mesozoic units, by intense salt diapirism and nor-
mal faulting (e.g. Baldschuhn et al. 2001, Maystrenko
et al. 2005, Hese 2012). It is bounded to the West Figure 1. GeoPower and GeotIS-StörTief project areas,
by the Westschleswig Block and to the East by the distribution of salt diapirs and main structural units.
Eastholstein-Mecklenburg Block, less complex and
relatively shallow horizontal layered areas. We focus
on sandstone formations, which are potentially suited modelling is always complemented by reservoir char-
for hydrothermal utilization. Most relevant in this acterization to gain more detailed information on
region are sandstones of the middle Buntsandstein, lithological and hydraulic properties at well locations.
upper Keuper and middle Jurassic (e.g. Kaufhold et al. The evaluation of the geothermal potential is the result
2011). Furthermore we concentrate on deep reaching of a combined interpretation of geometrical, lithologi-
fault systems, which may represent permeable path- cal and petrophysical datasets. This contribution aims
ways for hot circulating fluids, and on salt diapirs, to present a recently developed statewide geological
whose development is closely related to faulting and 3d model, showing the overall structure, geothermal
sedimentation and which have an effect on structural, reservoir complexes and fault zones of the Glückstadt
geochemical or petrophysical properties (e.g. temper- Graben and adjacent areas.
ature field) of adjacent formations of the rim syncline.
Within the scope of past and recent R&D projects 2 DATA
geological 3d models have been developed, covering
the deep underground of Schleswig-Holstein and The knowledge of the underground structure of
locally the near surface underground. Geometrical Schleswig-Holstein is mostly based on deep wells and
149
reflection seismic data, which have been produced pri- with input constraints. Such a boundary representation
marily for hydrocarbon exploration between 1870 and model can be used to construct volume models, e.g. 3d
2007. The onshore part of Schleswig-Holstein is cov- voxets, stratigraphic 3d grids or tetrahedral meshes.
ered by 3234 seismic profiles, 2 seismic 3d surveys and
about 1600 wells. Within the past decades the North 3.1 Interreg IVa project geopower: investigation of
German Basin has been investigated intensively by the North German basin in the Danish-German
various research projects (e.g. DFG Project SPP1135: border region for the utilization of geothermal
Dynamics of Sedimentary Basins under varying Stress energy
Regimes: The example of the Central European Basin,
The aim of the project GeoPower was the evaluation of
Littke et al. 2008; The Millenium Atlas, Evans et al.
the underground potential of the region Syddanmark /
2002; DEKORP/Basin96, Bayer et al. 2002; Geotek-
Schleswig for the use of deep geothermal energy. It was
tonischer Atlas von NW-Deutschland, Baldschuhn
funded by the European Community in the scope of the
et al. 2001). In the 80’s and 90’s Kockel and Bald-
INTERREG VI A program Syddanmark-Schleswig-
schuhn et al. (2001) developed the Tectonic Atlas of
K.E.R.N. and the European fund for regional develop-
Northwest Germany, using wells and reflection seis-
ment. Project partners have been the LLUR SH (Dept.
mic profiles. The Atlas consists of 14 depth contour
6, Geological Survey of Schleswig-Holstein), GEUS
maps of the base of lithostratigraphic units between the
(Geological Survey of Denmark and Greenland), the
middle Miocene and the Zechstein. Moreover it also
Institute for Geosciences of the University of Aarhus
contains residual thickness maps of selected forma-
and the Institute for Geosciences of the University of
tions, geological profiles, reports and more. The data
Kiel. The Geological Surveys of Schleswig-Holstein
of the Tectonic Atlas, wells and reflection seismic data
and Denmark/Greenland investigated seismic profiles
are the primary constraints for geological modelling
and correlated relevant reflectors across the border.
and reservoir characterization.
Furthermore, well reports and logs have been analyzed
For data processing, data analysis and 3d modelling
to gain information on the lithological composition,
we are using the Software GOCAD (Paradigm, Version
thickness, porosity and permeability of sandstone for-
2009.4 and 2015). Due to the fact, that the industry data
mations. The project focussed sandstone layers of the
has been mostly archived in an analog form and due
middle Buntsandstein (Bunter Sandstone Fm.), upper
to the huge amount of datasets, the digitization pro-
Keuper (Gassum Fm.) and middle Jurassic (Haldager
cess is not finished yet. We set up all well objects with
Sand Fm. absent in southern Denmark). Seismic data
horizon markers and relevant logs. Selected digitized
and the Tectonic Atlas have been used to develop a
seismic sections have been set up and interpreted in
geological 3d model of the project area. The Institute
the time domain together with time converted wells.
for Geosciences of the University of Aarhus used the
All datasets of the digital Tectonic Atlas are prepared
model to develop a 3d temperature model. Further-
for further applications as point data, line data, 2d
more, the Institute for Geosciences of the University
grids and voxets. These are depth contour and residual
of Kiel did geophysical measurements to close data
thickness maps, distribution boundaries, fault traces
gaps in the Flensburg region.
and geological profiles. Within a past research project
In the Danish project area GEUS investigated about
a statewide 3d model of the Tectonic Atlas of NW-
260 seismic profiles to map lithostratigraphic hori-
Germany has been constructed (Rosenbaum et al.
zons, faults and salt structures. On the German side 65
2012, Hese et al. 2012), which can be seen as a starting
relevant seismic profiles have been analyzed and used
model for further enhancements and developments.
to establish a connection between the horizons of the
Tectonic Atlas and Danish reflectors. In the next step
3 GEOLOGICAL MODELLING a seismic velocity model has been derived using well
logs and partly also stacking velocities to fill big data
In past and recent modelling studies the datasets of gaps. After time to depth conversion of all reflectors
the Tectonic Atlas have been used as primary input the comparison of depth contour maps on the German
constraints. Well data is used for depth corrections side with seismic reflectors on the Danish side showed
within the modelling workflow. Various relevant seis- only minor depth differences. A geological 3d model
mic profiles have been analyzed to map the under- has been built using the TectonicAtlas and Danish seis-
ground structure, to gain insights into the content mic reflectors. It is showing the overall structure of the
and uncertainties of the Tectonic Atlas and to estab- northern Glückstadt Graben and the northern rim of
lish a seismostratigraphic connection to Denmark. The the North German Basin near the Ringköbing-Fünen-
direct use of seismic reflectors together with datasets High. Moreover, it provides insights into the spatial
of the Tectonic Atlas is hampered by the fact, that the distribution of the main geothermal reservoir com-
seismic velocity model of the Tectonic Atlas (Jaritz plexes and salt diapirs. The project also studied fault
et al. 1991) is required for time to depth conversion in zones. Due to data gaps and difficulties to correlate
order to avoid artificial depth differences. It might be faults between the seismic profiles they have not been
reconstructed in future. integrated into the 3d model but can be shown on maps.
The recent models are triangulated surfaces, form- The model has been used to produce map sets show-
ing the boundaries of the geological objects. The aim ing the thickness and depth of reservoir complexes.
is a consistent model topology and a good agreement For 3d temperature modelling by researchers of the
150
Uni Aarhus the model geometry has been transferred the base surfaces have been cut by distribution bound-
from GOCAD to FEFLOW and expanded by Pre- aries to adapt certain boundaries to unconformities.
Permian horizons. The model provides information of After the first interpolation process they have been
the temperature field, especially within the range of corrected and manipulated in smaller regions to obtain
hydrothermal reservoir complexes (Fuchs & Balling consistency, to model the fault throws and in gen-
2016). The effect of the Zechstein salt diapirs is clearly eral to optimize the surface boundary contacts. The
visible causing temperature anomalies due to the high final model is set up by fault planes, hull surfaces of
thermal conductivity of rock salt. Temperature map salt diapirs and 12 lithostratigraphic horizons, which
sets have been produced for constant depths of 1, 2 are: base Tertiary, upper Cretaceous, lower Creta-
and 3 km and also for the depths of the top boundaries ceous, upper Jurassic, middle Jurassic, lower Jurassic,
of hydrothermal reservoir complexes. In general the upper Keuper, lower Keuper, upper Buntsandstein,
temperatures are sufficient for heat supply. Despite middle Buntsandstein, lower Buntsandstein and Zech-
the gradual thinning and decreasing depths of the stein (Fig. 2). The model will be integrated into
reservoirs towards north, single sandstone layers are the Geothermal-Information-System GeotIS (LIAG,
exceeding a minimum thickness of 15 m at most well Hannover). Moreover it provides a structural frame for
locations (GeoPower working group 2015). Moreover, possible future studies or applications (e.g. tempera-
collected porosity and permeability values of Dan- ture modelling or numerical simulation of processes).
ish and German sandstones are revealing a chance The investigation of the deep and shallow under-
for intermediate or good hydraulic conditions (e.g. ground is continued within the framework of the
Mathiesen et al. 2009, 2013). The project provides TUNB Project (Deep Underground North German
information on general trends of the underground Basin), which is coordinated by the Federal Insti-
structure and conditions to rule out suitable loca- tute for Geosciences and Natural Resources Germany
tions. Nevertheless there is a need for more detailed (BGR, Hannover). Project partners are the Geologi-
investigations to minimize uncertainties. cal Surveys of the federal states of northern Germany.
The aim is the development of a geological 3d model
3.2 GeotIS-StörTief project: 3D modelling of the of the North German Basin and the parameterization
Glückstadt Graben of formations and structures, which are relevant for
utilization.
The joint project GeotIS-StörTief is investigating the
role of deep reaching fault zones in the geothermal
energy use (LIAG Hannover, LLUR SH Flintbek,
KIT Karlsruhe). The Geological Survey of Schleswig- 4 GEOLOGICAL MODELS FOR
Holstein is participating with a subproject, which aims GEOTHERMAL PLANNING TOOLS
to study deep reaching fault zones, hydrothermal sand-
stone formations and faulted sandstone formations in Hydrothermal reservoir formations are sandstone lay-
the Glückstadt Graben area. One aim is the developers which must fulfill lithological and petrophysical
ment of a structural 3d model of the Glückstadt Graben requirements to enable a successful hydrothermal
to visualize the overall structure and especially fault project. It is necessary to estimate relevant param-
zones and reservoir complexes. The second aim is eters like depth, thickness, composition, porosity,
the characterization of sandstone formations and fault permeability and temperature. Geothermal 3d Geo-
zones by investigation of well reports and logs. These Information-Systems (e.g. GeotIS, Project GeoRG)
studies are complemented by interpreted seismic pro- aim to provide data to rule out areas, which might
files. Finally the subproject aims to merge the results be relevant for hydrothermal utilization. A geologi-
from the different disciplines. We are working on the cal 3d model provides a structural frame for a 3d
question whether it is possible to enhance and expand Information-System and is used to visualize the dis-
the structural model by interpolation and extrapolation tribution, depth and thickness of reservoir complexes.
of parameters like lithology or sandstone thickness, When a volume model of reservoir complexes is com-
which are known at well locations. bined with a 3d temperature model it is possible to
The modelling workflow started with the construc- do spatial requests. For example one can extract vol-
tion of fault surfaces and hull surfaces of the salt umes which are fulfilling criterias like a depth range
diapirs. Main constraints are depth projected fault of 800–3000 m and minimum temperatures of 50◦ C.
traces and salt diapir boundaries of the Tectonic Atlas. A reservoir complex is a generalized description of a
In some areas seismic interpretations and detailed geo- formation which is characterized by an intercalation
logical profiles have been used to aid the construction of shale and sandstone layers. It is important to get
of fault planes. The salt model has been corrected information on the thickness of a single contiguous
and partly rebuild after the completion of covering sandstone layer. Therefore, a geological model should
horizons. Triangulated surfaces of the base of the be complemented by additional more detailed param-
lithostratigraphic units have been interpolated using eters of single sandstone layers within the reservoir
various constraints, most importantly depth and thick- complexes, unfortunately these are commonly only
ness contour lines of the Tectonic Atlas. Within the known at well locations. The question is, whether a 3d
iterative fitting process the surfaces have been inter- mapping of the thickness variation of a single sand-
sected with fault and salt diapir surfaces. In addition stone layer can and should be done in a statewide
151
Figure 2. Geological 3d model of the Glückstadt Graben and adjacent areas. The pre-Cretaceous horizons are shown here
together with fault planes and the model of the salt diapirs (V.E.: x3).
scale. Increasing uncertainties arise due to the poor salt diapirs. In addition, more detailed information of
well data coverage in domains, which are character- lithological and petrophysical properties of sandstone
ized by complex facies changes or synsedimentary formations have been determined at well locations
tectonic activity and subsidence. At least in specific using reports and logs.
domains it appears to be acceptable to include single The modelling concept is an enhancement and
sandstone layers into a geological model. Important expansion of the Tectonic Atlas of NW-Germany. The
parameters are also the porosity and permeability of Tectonic Atlas is a very valuable dataset and a great
sandstones. Due to the sparseness of the porosity and achievement, which provides a seamless interpretation
permeability data and the facies controlled variability of thousands of seismic sections and wells. Never-
3d mapping is not possible. Momentarily, the visu- theless it is necessary to step forward in order to
alization of summarized trends (porosity vs. depth, increase the level of detail and geometrical certainty.
porosity vs. permeability) aims to give insights into Today, the 3d modelling software allows a synoptic
the parameter range. visualization and interpretation of different datasets
A geological model shows distinct boundaries of and offers a variety of tools and workflows to build
rock formations, although seismic sections are often models. The spatial consistency check during geolog-
disturbed e.g. in faulted areas or in the vicinity of salt ical modelling was an important correction step to
domes. A quantitative visualization of the geometrical decrease uncertainties. The recently finished model of
uncertainty is desirable and necessary to be able to the deep underground of Schleswig-Holstein is almost
assess the significance of the model. The estimation of consistent in a topological and geological sense. The
the uncertainty remains a difficult task due to the huge geometrical accuracy of the model depends on the
amount of causes and influences (e.g. data distribution given scale of the input data (1:300.000) and is vary-
and resolution, seismic processing and interpretation, ing locally, mostly influenced by seismic and well
velocity model). In a first approach, an incomplete data coverage and resolution, structural complexity
uncertainty estimate would be a useful contribution. and depth. The visualization of the model accuracy
(or uncertainty) remains a future aim and matter of
research. The model will be enhanced in future. One
5 CONCLUSIONS aim is to reach a higher level of detail, especially in the
shallow Tertiary underground area. It is also planned
The Geological Survey of Schleswig-Holstein is work- to enhance and expand the fault model.
ing on the evaluation of the geothermal potential of the For the use of new seismic interpretations together
underground. Past and recent research projects lead to with data of the Tectonic Atlas, we aim to reconstruct
the development of a new geological 3d model of the the seismic velocity model which was used for the
onshore area of SH, which provides information of the Tectonic Atlas (Jaritz et al. 1991). Finally, the velocity
overall structure, distribution, thickness and depth of model might be enhanced to optimize the time to depth
reservoir complexes, extent of main fault systems and conversion. We are investigating also, how to integrate
152
single sandstone units (e.g. Volpriehausen-Quickborn in Nordwestdeutschland. – Geologisches Jahrbuch, E 45:
sandstone) into the geological model and whether a 3d 23–57, Hannover.
mapping of other relevant parameters is possible. Kaufhold, H., Hable, R., Liebsch-Dörschner, T., Thom-
sen, C. 2011. Verbreitung und Eigenschaften mesozois-
cher Sandsteine sowie überlagernder Barrieregesteine in
Schleswig-Holstein und Hamburg – Basisinformationen
REFERENCES zur energiewirtschaftlichen Nutzung des tieferen Unter-
grundes. – SDGG, 74, Stuttgart.
Baldschuhn, R., Frisch, U., Kockel, F. 2001. Geotek- GeoPower working group 2015. GeoPower Erdwärme für
tonischer Atlas von Nordwest-Deutschland und dem die Region Südjütland – Schleswig. Schriftenreihe LLUR
deutschen Nordsee-Sektor – Strukturen, Strukturentwick- SH – Geologie und Boden: 21, www.llur.schleswig-
lung, Paläogeographie. – Geologisches Jahrbuch, A 153, holstein.de, Flintbek.
mit 3 CD ROMs, Hannover (BGR). Littke, R., Bayer, U., Gajewski, D., Nelskamp, S. 2008.
Bayer, U., Scheck, M., Rabbel, W., Krawczyk, C.M., Götze, Dynamics of Complex Intracontinental Basins: The Cen-
H.-J., Stiller, M., Beilecke, T., Marotta, A.M., Barrio- tral European Basin System: 590 S., Berlin/Heidelberg:
Alvers, L. & Kuder J. 2002. An integrated study of the Springer.
NE German Basin. – Tectonophysics, 314 (1–3): 285–307, Mathiesen, A., Kristensen, L., Bidstrup, T., Nielsen, L.
Amsterdam. H. 2009. Vurdering af det geotermiske potentiale i
Evans, D., Graham, C., Armour, A., Bathurst, P. 2002. Danmark. – Danmarks og Grønlands Geologiske Under-
The Millenium Atlas: Petroleum Geology of Central & søgelse Rapport 2009/59, Kopenhagen.
Northern North Sea: 390 S., London: GSL. Mathiesen A., Kristensen L., Møller Nielsen C., Weibel R.,
Fuchs, S., Balling, N. 2016. Improving the temperature Leth Hjuler M., Røgen B., Mahler A., Nielsen L. H. 2013.
predictions of subsurface thermal models by using high- Assessment of sedimentary geothermal aquifer parame-
quality input data. Part 1: Uncertainty analysis of the ters in Denmark with focus on transmissivity – European
thermal-conductivity parameterization. Geothermics (IF Geothermal Congress, Pisa, Italy.
2014: 2.949), in review. Maystrenko, Y., Bayer, U., Scheck-Wenderoth, M. 2005.
Fuchs, S., Balling, N. 2016. Improving the temperature The Glueckstadt Graben, a sedimentary record between
predictions of subsurface thermal models by using high- the North and Baltic Sea in north Central Europe. –
quality input data. Part 2: A case study from the Danish- Tectonophysics, 397: 113–126, Amsterdam.
German border region. Geothermics (IF 2014: 2.949), in Rosenbaum, S., Hese, F., Liebsch-Dörschner, T., Offermann,
review. P., Rheinländer, J., Thomsen, C. 2012. Geologische Mod-
Hese, F. 2012. 3D Modellierungen und Visualisierung von elle der Deck- und Speichergesteine Schleswig-Holsteins.
Untergrundstrukturen für die Nutzung des unterirdischen Schlussbericht des Teilvorhabens M6 im Rahmen des Ver-
Raumes in Schleswig-Holstein. Dissertation, Christian- bundprojektes CO2 -MoPa, Vorhaben: Dimensionierung
Albrechts-Universität zu Kiel. und Risikoanalysen bei der CO2 -Speicherung – Sonder-
Jaritz, W., Best, G., Hildebrand, G, Jürgens, U. 1991. programm GEOTECHNOLOGIEN; 107 S., LLUR SH,
Regionale Analyse der seismischen Geschwindigkeiten Flintbek.
153
Effects of ground heat exchanger variable pipe separation in

ground-source heat pump systems
N. Makasis, G.A. Narsilio, A. Bidarmaghz & I.W. Johnston

The University of Melbourne, Parkville, Victoria, Australia
ABSTRACT: Closed loop ground-source heat pump (GSHP) systems can efficiently provide clean and renew-
able energy for heating and cooling purposes using direct geothermal energy. These systems use Ground Heat
Exchangers (GHE) to transfer heat to and from the ground. Vertical GHEs contain loops, pipes with circulating
fluid, which transfer energy between the ground and the fluid. One very common assumption made in design-
ing GSHP systems is that, when installed, the loops containing the circulating fluid remain straight and evenly
separated along the length of the GHE. However, this is rarely true, as the high-density polyethylene (HDPE)
pipes can flex within the GHE before being grouted into position. This can result in thermal interference not
accounted for accurately in the design, with the worst case scenario represented by direct contact between the
inlet and outlet pipes, leading to a negative impact on the performance of the system. This paper investigates the
effect of this interference and the implications of ignoring it in design.
1 INTRODUCTION AND BACKGROUND

INFORMATION
1.1 GSHP systems

Ground Source Heat Pump (GSHP) systems can
be used to efficiently provide renewable geothermal
energy for heating and cooling purposes. These shal-
low energy systems use heat transfer principles to
extract and reject heat from and to the ground a few
tens of metres below the surface. A heat pump is
connected to a circuit within the building which trans-
fers the heat to and from the building as well as to
a series of ground heat exchangers (GHEs), which
transfer the heat from and to the ground. A GHE con-
tains loops (usually high-density polyethylene (HDPE)
pipes) with a circulating fluid (usually water) that
transports the heat between the system components.
These systems are known to typically run at a coeffi-
cient of performance of about four, meaning 4 kW of
heating/cooling energy is produced for every 1 kW of
electricity consumed. Moreover, GSHP systems are
the most used amongst the different applications of
direct geothermal energy (Preene & Powrie, 2009,
Johnston et al, 2011, Lund et al, 2015).
Figure 1. Schematic of a GSHP system in heating mode

1.2 Separation of pipes along the GHE length with separators (Not to scale).
There are many different design methods for GSHP
systems, both analytical and numerical. A common spacers (e.g., geoclipTM , sureclipTM ) are sometimes
assumption for all these methodologies is that the used during construction, to keep the pipe separa-
pipes remain equally separated along the length of the tion fixed at specified intervals (for example every
GHE, after their placement and grouting. In order to be couple of metres down the length of the GHE), as
more confident in this assumption, pipe separators or schematically shown in Figure 1.
155
However, even with the use of separators, a fixed
separation along the length of the borehole cannot be
guaranteed. It is considered highly likely that between
the separators, the location of the pipes in any horizon-
tal plane is likely to vary, possibly randomly. Clearly,
without separators, the location of the pipes is likely
to be even more random.
A question then arises about what effect this uneven
pipe separation may have in influencing the thermal
performance of the system. If the downward (inlet) and
upward (outlet) pipes come closer or even touch each
other, then a thermal “short-circuiting” between these
pipes is expected. This paper investigates the effect of
this thermal interference and the potential implications
of ignoring it in the design of GSHP systems.
2 FINITE ELEMENT MODEL

Figure 2. Modelling Geometry (ground domain to the left
2.1 Brief model description and detail of the embedded 114.5 mm diameter GHE to the
right, containing 25 mm outer diameter HDPE pipes).
To conduct the research, a validated state of the art 3D
numerical model, developed at the University of Mel-
bourne, was used (Bidarmaghz, 2014, Narsilio et al, Table 1. Material Properties.
2016, Bidarmaghz et al, 2016). The model is built and
Parameter Value Unit Description
solved using the finite element software COMSOL
and is based on the coupling of the governing equa- κground 2.7 W/(mK) Thermal conductivity
tions of heat transfer, fluid flow and energy balance. of ground
The heat transfer is primarily modelled by conduc- ρground 2353 kg/m3 Density of ground
tion, occurring mainly in the soil, the grout filling the κgrout 0.7, 2.1 W/(mK) Thermal conductivity
borehole GHE and the pipe walls and partially in the of grout
fluid, and convection, occurring mainly in the carrier ρgrout 1600 kg/m3 Density of grout
◦
fluid. (Note here that groundwater flow is not taken Tfarfield 18 C Average annual ground
into account). For the fluid flow in the pipes, 1D ele- temperature
ments are used to solve the continuity and momentum qfluid 11 L/min Flow rate within pipes
equations for fluid flow and the energy equation for
heat transfer. The results are coupled to those of the
3D heat transfer elements representing the grout and ground boundary effects. In the outer boundary of the
ground surrounding the pipes. ground domain, a constant temperature is prescribed
as a boundary condition equal to the average annual
ground temperature Tfarfield , as shown in Figure 2-left.
2.2 Conditions and parameters Another boundary condition used is that there is no
The modelling used in this research is based on typical thermal flux at the surface or the base of the model
weather and soil conditions for Melbourne, Australia, (qz=0 = 0). Lastly, an initial temperature based on the
with the material properties used presented in Table 1 annual ground temperature for the geographical loca-
(Colls, 2013). While the length of the GHEs modelled tion is used, which for Melbourne is around 18◦ C.
in this work may be shorter (to save computational These are all boundary and initial conditions typically
time) than the 50 m commonly encountered in Mel- used in GHE modelling works. The fluid flow rate
bourne, the geometry and dimensions used for the prescribed in the pipes is 11 L/min, typical of GHE
modelled GHE are typical of practice in the region. operation. Other boundary conditions are specified in
These can be seen in Figure 2. the next section.
2.3 Boundary and initial conditions 2.4 Modelling variable pipe separation
While typical GSHP systems include a field with mul- In order to simulate the effect of the random separation
tiple GHEs, in this work each simulation focuses on between the pipes with the length of the GHE, the pipes
one single GHE to isolate the effects of pipe separa- are modelled as having a sinusoidal shape for simplic-
tion within the GHE from the thermal interference that ity. This is believed to be an average/representative
may arise from adjacent GHEs. case as the pipes come closer together and may touch
In this case, a radius of 10 metres of ground around at the midpoints between the separators. In reality the
the GHE is considered to be adequate to avoid any random nature of this variable separation could have
156
Table 2. Parametric analysis variables.
Parameter Values Used Unit
Configuration (αpipe ) Inner (10.7), Outer (25.2) mm

κgrout 0.7, 2.1 W/(mK)
Thermal Load Low, Medium, High kW
will be as in Figures 3-A and 4-A throughout the length

of the GHE. In the modelling, it is assumed that sep-
arators are placed every 3 metres along the length of
the GHE, thus resulting in a wavelength (λ) of 3 m.
It should be noted that the presented models in the
‘Inner’ configuration account only for the pipes bend-
ing towards the centre of the GHE and not towards the
Figure 3. Cross-section of GHE with variable pipe separa- edge of the GHE. For example, in Figure 3 the furthest
tion as it varies along depth (Inner configuration). away from centreline of the GHE the pipes will be is
as shown in Figure 3-A, which is their starting posi-
tion (where separators are placed). The reason for not
investigating this behaviour is to keep the paper at a
reasonable length; however bending towards the edge
of the GHE will be considered as a further study to the
one presented.
3 PARAMETRIC ANALYSIS
Using the model described above, various one year

simulations were run to investigate the effects of
variable pipe separation. For each set of parame-
ters discussed below, two simulations were run, one
modelling a fixed pipe separation (U-loops remain-
ing straight, Figs 3-A and 4-A) and one modelling
a variable pipe separation (U-loop pipes adopting a
sinusoidal shape along the GHE length, Figs. 3 and 4).
Figure 4. Cross-section of GHE with variable pipe separa-
tion as it varies along depth (Outer configuration).
Three of each model’s input parameters were var-
ied to study their responses. The parameters which
were varied and their respective values are presented in
the pipes potentially touching for much greater lengths Table 2. The values chosen are believed to be towards
between some separators and less between others. The the upper and lower bounds of the range of values
modelling geometry of variable pipe separation can be usually encountered in Melbourne.
better seen in Figures 3 and 4, which show GHE cross- Firstly, the two different initial configurations of the
sections, how the location of the pipes varies along pipes were modelled (Figs 3 and 4) one closer to the
the GHE length and the geometrical values used. The edge of the GHE (Outer) and one closer to the core
schematic to the right of these figures shows the cen- (Inner).
treline of one of the pipes. As shown, two different Secondly, two different values for the conductivity
configurations have been modelled, the Inner config- of the grout (κgrout ) were used, one relatively low and
uration, where the separators cause the pipes to be one relatively high, to better understand the interplay
located more towards the centre of the borehole GHE between pipe to pipe thermal interference and the rate
(Fig. 3), and the Outer configuration, where wider sep- of heat transfer between the grout and the ground.
arators cause the pipes to be located more towards the Finally, three different thermal load distributions
edge of the borehole (Fig. 4). For both of these con- were used, as seen in Figure 5, to represent cases where
figurations, the pipes are assumed to bend and touch the total thermal load for an average building case
at the centre length of the pipe between the separators in Melbourne (about 20 kW peak-cooling and 20 kW
(Figs. 3 and 4 – D) and at the centre line of the GHE. peak-heating) is distributed over a different number of
The different configurations also imply a different GHEs, i.e. the GHE field is smaller/bigger.The smaller
α, which is the value of the maximum sinusoidal wave the field, the more energy one single GHE must pro-
amplitude, indicating the distance the pipes can move vide. For the High case a total of 7 GHEs are used,
from their initial position. It should also be noted that 9 GHEs for the Medium case and 11 GHEs for the
for the fixed pipe separation cases, the configuration Low case.
157
Figure 7. Difference in Tave,fluid between fixed and variable
Figure 5. Annual thermal load per GHE for three different pipe separation models over one year.
field sizes.
could significantly reduce the performance of the sys-
tem. The minimum value in particular shows that the
use of an antifreeze solution is required.This important
requirement would not be obvious when considering
fixed pipe separation. It should also be noted that for
all cases, i.e. any combination of parameters shown in
Table 2, the variable pipe separation model produced
a more extreme response than the fixed pipe separa-
tion model using the same parameters. A more extreme
response is defined by lower fluid temperatures being
reached in heating mode and higher fluid temperatures
in cooling mode, as shown in Figure 6 for one specific
set of parameters from Table 2.
Figure 7 shows the difference between the mean
fluid temperature of the fixed and variable pipe sepa-
ration models for the conditions of Figure 6. Looking
closely, it can also be observed that a relatively constant
difference of about 3 to 5◦ C exists during the heating
Figure 6. Comparing Tave,fluid for fixed and variable pipe period in winter, where the weather is less variable
separation models over one year.
in Melbourne. During the summer, which has highly
variable climate conditions, the difference has short
duration peaks, the highest being 9.11◦ C on day 249,
which is 4th of January. These differences are obvi-
ously very significant and can not only influence the
4.1 General effect of variable pipe separation
viability of a design but also result in inefficient and
The numerical results show that the effect of variable non-economical system designs.
pipe separation, that is usually ignored in design, can Figure 8 shows the average of the absolute value of
be significant to the operation of the GSHP system. the differences in Tave,fluid between each case modelled
The response of the system is presented in terms of with a variable pipe separation and the respective case
the average temperature of the fluid circulating in the modelled with a fixed pipe separation, over the one
pipes in the GHE, Tave,fluid . year simulations. For example, by taking the absolute
Figures 6 and 7 show a comparison between the value of each point in Figure 7 and calculating the
fixed pipe separation model and the variable pipe average over one year, the value obtained is then shown
separation model. The parameters that produced the in Figure 8 (which, for example, for the conditions of
most extreme difference between the two models Fig. 7, is 2.9◦ C)
corresponded to the High thermal load imposed on From the results in Figure 8, various observations
the Outer configuration with κgrout = 0.7 W/(mK). As can be made. Firstly, it is evident that the Outer con-
can be seen in Figure 6, the fixed pipe separation figuration is more influenced by the effect of variable
model results in a minimum and maximum Tave,fluid of pipe separation, as it shows larger fluid temperature
−0.6◦ C and 32.5◦ C respectively, which are within the difference between the variable and fixed separation
working range typically required by most heat pumps. cases in comparison to the inner configuration case
However, when the variable pipe separation is consid- for a given grout thermal conductivity. Secondly, it is
ered, these values change to −6.2◦ C and 41.7◦ C which clear that the thermal conductivity of the grout is also
158
edge of the GHE. Therefore, comparing the fixed and
variable pipe separation models, in the Outer config-
uration the difference in mean fluid temperature is
higher because the pipes in the variable pipe separa-
tion model approach the centre, forcing heat to travel a
longer distance through the grout. In the correspond-
ing fixed pipe separation case, the pipes are closer to
the edge, heat is more easily transferred to/from the
ground. In other words, the overall thermal resistance
of the Outer configuration varies more with depth with
respect to its fixed pipe separation case than that of the
Inner configuration. This means that the performance
of the Outer configuration is more affected by vari-
able pipe separation. A more thorough explanation of
the heat transfer through the solid materials is given in
Section 4.3.
Furthermore, even though data is not presented, it
Figure 8. Average annual Tave,fluid difference between each should also be noted that models with the Inner config-
case modelled with variable pipe separation and the equiva- uration, in general produced a more extreme (higher
lent one modelled with fixed pipe separation.
maximum and lower minimum) Tave,fluid values than
the models with the Outer configuration, most prob-
an influencing factor, with a lower value resulting in ably as the pipe concentration is mainly towards the
more extreme differences between fixed and variable centre of the GHE and heat needs to travel a longer dis-
pipe separation models, since the values that have a low tance through the grout. Further discussion is provided
κgrout are higher than the equivalent ones having a high in Section 4.3.
κgrout . Lastly, it can be seen that the higher the thermal
load applied to the GHE, the more extreme the differ-
ence between the fixed and variable pipe separation 4.3 The effect of thermal conductivity of the grout
models.
As mentioned earlier, a lower thermal conductivity of
the grout amplifies the extreme response of variable
pipe separation. Figures 9 and 10 show an example
4.2 The effect of pipe configuration
comparison of the two models (fixed and variable pipe
The two different configurations (Inner and Outer) rep- separation), in heating mode (Fig. 9) and cooling mode
resent the position the pipes are placed within the GHE (Fig. 10), using κgrout = 0.7 and κgrout = 2.1 W/(mK).
in the x-y direction, as well as the distance the pipes The figures show the temperature distribution at a
can move in that direction to reach the centre along the depth of 15 m, where the light colours denote high
GHE length. This distance forms the amplitude of the temperatures and the dark colours low temperatures.
wave-like shape the pipes take (αpipe ), as shown earlier In heating mode (Fig. 9), when heat is transferred from
in Figures 3 and 4. the ground, the temperature within the GHE is lower
The Outer configuration cases result in a higher than its surroundings, while in cooling mode (Fig. 10),
difference between the variable and fixed pipe sep- it is higher.
aration models as shown in Figure 8, meaning the As it can be seen, with a lower κgrout a wider temper-
effect of variable pipe separation is dominant. This can ature range within the GHE results from the thermal
be explained by considering that when αpipe is larger, loads, compared to the same case using a high κgrout
the pipes move more in the x-y direction compared (Comparing Fig. 9-a to 9-b, 9-c to 9-d, 10-a to 10-b and
to when αpipe is low. Considering Figures 3 and 4, 10-c to 10-d). This effect can be further understood
the difference between Figure 3-A and 3-D is much by considering how heat flows through the different
less significant than the difference between Figure 4-A materials, from the pipes, to the grout and the ground
and 4-D. For each configuration, when comparing the and vice versa. When κgrout is low, the heat trans-
variable pipe separation model with the fixed variable fer between the ground and the grout is less efficient
separation model, the key factor to consider is how heat leading to a larger temperature drop in the fluid and
is transferred in each case, noting that κgrout < κground the GHE and consequently a wider temperature range
in this study. In the Inner configuration, both variable within the GHE. Taking the heating mode as an exam-
and fixed pipe separation models have the pipes rel- ple (Fig. 9), as the fluid flows in the pipes, heat from the
atively closer to the centre of the GHE and therefore surrounding grout is transferred to the fluid, lowering
for both cases the heat needs to be transferred a com- the temperature of the grout. Heat is then transferred
parable distance through the grout towards the ground from the ground to the grout due to the differential
and vice versa. temperature. However, because of the low thermal con-
In the Outer configuration, the distance the heat ductivity of the grout, the heat transfer becomes less
needs to travel through the grout varies more signif- efficient and this cumulative effect over time causes
icantly, as the pipes are initially placed closer to the the fluid temperatures to reach more extreme values. In
159
Figure 11. Average annual Tave,fluid difference between a
Figure 9. Temperature distribution at a cross-section of the lower and higher value of κgrout , for each case.
GHE, at: z = −15 m, t = 7 Jul (Heating Mode), for Outer
configuration and High load.
Figure 12. Average annual Tave,fluid difference between High

and Low thermal load (High – Low) for all cases.
Figure 10. Temperature distribution at a cross-section of conductivity of the grout are compared instead here in
GHE, at: z = −15 m, t = 4 Jan (Cooling Mode), for Outer Figure 11.
configuration and High load. It can be seen that the Inner configuration per-
formance is more sensitive to the change of κgrout ,
showing a higher temperature difference. It can also be
the case of cooling, the concept is the same but with the seen that the variable pipe separation model is more
heat being transferred in the opposite direction. Fig- sensitive to this change compared to the fixed pipe
ures 9-b and10-b represent the variable pipe separation separation model, especially in the Outer configura-
models, where the temperature concentration effect is tion case, due to the pipes approaching closer to the
more pronounced around the pipes and Figures 9-d centre of the GHE, as discussed in the previous section.
and10-d show a smother, lower and more spread tem-
perature concentration in a larger central area between
4.4 The effect of thermal load
the straight pipe loops (fixed pipe separation).
To summarise, Figure 11 shows the average fluid The effect of thermal load peaks is investigated numer-
temperature difference between having a high and low ically in this study. It is observed that the response
κgrout for different cases. For clarity, each of the val- of the system is more extreme when a higher thermal
ues in Figure 11 are obtained by comparing the results load is applied on the system, meaning Tave,fluid reaches
of two simulations, one using a high value for κgrout higher temperatures in summer and lower in winter
(2.1 W/(mK)) and the other using a low value for κgrout for both the fixed pipe separation and the variable
(0.7 W/(mK)), keeping all other parameters are the pipe separation models. More importantly, the effect of
same. The absolute difference of Tave,fluid is calculated variable pipe separation tends to be more pronounced
for each time step and the average of that differ- as the thermal load intensity increases, showing a more
ence over the year is the resulting value displayed in significant Tave,fluid difference between the model with
Figure 11. This process is similar to how Figure 8 was fixed pipe separation and the model with variable pipe
created, but instead of comparing the fixed and vari- separation. Figure 12 shows the average annual fluid
able pipe separation models the different values for the temperature difference between different cases, when
160
using High and Low thermal load distributions. Each the effect of pipe separation, can be significant. This
value is obtained by first calculating the absolute dif- suggests that if the effect is ignored when designing
ference between a model using a High thermal load a GSHP system, the system may be under-designed.
and a model using a Low thermal load (and keeping From the parameters investigated, the thermal con-
all other parameters the same). Following, that dif- ductivity of the grout as well as the initial placement
ference is averaged over the one year simulation to location of the pipes have been found to have a sig-
provide the annual average displayed in Figure 12. This nificant influence on the magnitude of the effects
is again similar to how Figure 8 was created, but instead of variable pipe separation. The lower the thermal
of comparing the fixed and variable pipe separation conductivity of the grout compared to the thermal
models, the thermal load value is being compared conductivity of the ground, and the further apart the
instead. pipes are, the more significant the effect of variable
There are a few conclusions that can be drawn from pipe separation will be. The magnitude of the thermal
the results of Figure 12. Firstly, it is clear that when load distribution is also found to affect variable pipe
κgrout = 0.7 W/(mK), the effect that increasing the ther- separation, but at a relatively lesser degree.
mal load has is more dominant. This is in accordance Overall, even though this paper has investigated
to the discussion in Section 4.3, since the higher the only a limited range variables, geometries and con-
thermal load for the GHE the more heat needs to be figurations and different combinations will produce
transferred and thus a larger temperature difference different results, it is clear that the variability of the
between the fluid and ground is required. Having a pipe location within the GHE can have a significant
grout with a low rate of heat transfer, will lead to effect on the system performance and is therefore
requiring a greater thermal gradient for the heat to worthy of further investigation.
flow, meaning more extreme temperatures in the fluid.
Secondly, the Inner configuration is more sensi-
tive to the increase of the system’s thermal demand, ACKNOWLEDGEMENTS
since all values of the temperature difference in Fig-
ure 12 for an Inner configuration are higher than the Funding from the Australian Research Council (ARC)
respective Outer configuration. This is logical, since FT140100227, The University of Melbourne and the
the pipes are in general positioned closer together in Victorian Government is much appreciated.
the Inner configuration and therefore thermal inter-
ference is more dominant and amplified more by the
increased load. It should be noted, however, that when REFERENCES
κgrout = 2.1 W/(mK) the Inner configuration is only
slightly more sensitive than the Outer, which again Bidarmaghz, A. (2014). 3D Numerical Modelling of Vertical
Ground Heat Exchangers. Ph.D Thesis, The University of
suggests that the thermal conductivity of the grout
Melbourne.
is a critical factor. Lastly, but importantly, the vari- Bidarmaghz, A. Narsilio A. G, Johnston I & Colls S (2016).
able pipe separation model is generally more sensitive The importance of surface air temperature fluctuations on
to a change in thermal load. As shown in Figure 12, long-term performance of vertical ground heat exchang-
all values for the variable pipe separation model are ers, Geomechanics for Energy and the Environment (in
higher than the equivalent ones for the fixed pipe sep- press)
aration model with the difference becoming largest Colls, S. (2013). Ground Heat Exchanger Design For Direct
in the most extreme case (Outer configuration with Goethermal Energy Systems. Ph.D Thesis, The University
κgrout = 0.7 W/(mK)). This can be explained by the fact of Melbourne.
Johnston, I. W., Narsilio, G. A., & Colls, S. (2011).
that in this model, the pipes come closer towards the Emerging Geothermal Energy Technologies, 15, 643–653.
centre of the GHE and the higher the thermal load the http://doi.org/10.1007/s12205-011-0005-7
more heat is transferred via the grout, as discussed Lund, J. W., & Boyd, T. L. (2015). Direct Utilization of
earlier. Geothermal Energy 2015 Worldwide Review, (April),
19–25.
Narsilio, G. A., Bidarmaghz, A., Colls, S., & Johnston, I.
W. (2016). Geothermal Energy, Detailed Modelling of
5 CONCLUSION Ground Heat Exchangers, Computers and Geotechnics
(conditionally accepted 17 Dec 2015).
An investigation of variable pipe separation, using a Preene, M., & Powrie, W. (2009). Ground energy systems?:
parametric analysis approach has been presented. The from analysis to geotechnical design. Géotechnique,
analysis shows that, using a sinusoidal pipe model, 59(3), 261–271. http://doi.org/10.1680/geot.2009.59.3.261
161
Numerical study of the performance of energy diaphragm walls
D. Rammal & H. Mroueh

Laboratory of civil Engineering and geo-Environment (LGCgE)-Polytech’Lille,
University Lille1 Sciences and Technologies, Villeneuve d’Ascq, Lille, France
S. Burlon
Institut Français des Sciences et Technologies des Transports, de l’Aménagement,
et des Réseaux (IFSTTAR), Paris, France
M.E. Suryatriyastuti
ITECH, Saint-Maurice, France
ABSTRACT: By combining their mechanical and thermal roles, energy diaphragm walls are being more used in
European countries. The performance of energy walls depends on the thermal exchange with the surrounding soil.
Nevertheless, thermal exchanges could be largely affected by ground water flow, and by the thermal properties
of the ground. Especially, conductive and advective contributions of these exchanges may vary. In this paper,
numerical simulations are performed to study the influence of groundwater flow and soil thermal properties on
the heat transfer between energy diaphragm walls and saturated sandy soil. The effect of the active length of an
energy diaphragm wall on the total heat exchange phenomenon is studied. The efficiency of the system affected
by several heating-cooling cycles is also analyzed. The obtained results permit a better understanding for the
heat exchange between the thermo-active diaphragm walls and the surrounding soil.
1 INTRODUCTION where Qin and Qout (W) are the inlet and outlet heat
respectively, mf and cf are the mass flow rate (kg/s) and
Energy foundations are being increasingly used nowa- the specific capacity (J/kg/K) of the heat carrier fluid
days in several countries especially in Europe such as respectively, and Tin and Tout (K) are the temperatures
in Germany, Switzerland, Sweden, and Austria (Link of the inlet and outlet tubes.
et al. 2015; Lund et al. 2010; Sigfusson & Uihlein Through this relationship, it is difficult to assess
2015; Weber et al. 2015), due to their ability to trans- precisely the influence of ground conditions on the
fer the loads applied by the upper structures to the thermal exchanges: the presence of groundwater flow
surrounding ground and to supply these supported and the soil permeability, the soil thermal conduc-
structures with their heating and cooling needs. More- tivity and its specific heat capacity. It is therefore
over, they represent an appropriate solution for the necessary to identify the appropriate geological and
production of a clean and renewable energy. hydro-geological conditions to ensure the best thermal
Despite of the many applications on energy struc- exchange of geothermal foundations.
tures, there is no common approach for the assessment Several studies have been conducted on the heat
of their thermal performance which depends largely transfer for borehole heat exchangers (Sigfusson &
on the thermal transfer between them and the sur- Uihlein, 2015), energy piles (Bouazza et al. 2013;
rounding soil. It shall be kept in mind that the thermal Cecinato & Loveridge 2015; Cervera 2013; Ma &
needs of the foundation for heating or cooling play an Grabe 2010), and for energy tunnels (Di Donna &
important role in the overall estimation of the thermal Barla, 2016). Energy diaphragm walls are less stud-
efficiency of a geothermal foundation. The thermal ied but they present interesting benefits. Compared to
transfer between the geo-structures and the ground is energy piles, they possess a bigger exchange surface
equivalent to the thermal power between both. Most of with the soil especially in the case of metro stations.
the existing studies estimate the total thermal transfer It is therefore of major interest to study this type of
through Equation 1, by taking only into account the structures. Few studies exist dealing with the design
thermal and hydraulic properties of the heat carrier of the exchanger tubes (Xia et al. 2012), and with the
fluid (Gao et al. 2008; Bouazza et al. 2013). heat transfer in energy diaphragm walls (Sterpi et al.
2014). However, no study deals directly with the effect
of groundwater flow on the thermal transfer through
the assessment of the conductive and advective energy
163
2.2 Power transferred by advection
In the case of water flow with a significant Darcy
velocity, there may be advection in addition to con-
duction. Advection refers to the transfer of a certain
quantity by the fluid motion; such as the transfer of
Figure 1. Resistive model for the energy foundation and the
soil. heat through water flowing between the soil particles.
In a given direction, the thermal power transferred by
advection between the foundation and the soil is given
by the following relation:
exchanged between the energy diaphragm walls and
the surrounding soil. The aim of this paper is to
focus on this topic by considering the geometry of the
diaphragm walls and its influence on the groundwater where Pv (x, t) = average thermal power transferred by
flow. advection between the energy foundation and the soil
This work includes a numerical estimation of the volume situated at a distance x from the foundation
soil volume affected by temperature variation, and and at time t, ρw (kg/m3 ) = density of groundwater,
—–
the analysis of the evolution of the allowable ther- cw (J/kg/K) = specific heat capacity of groundwater,
mal power exchanged by energy walls with time. Then and v (m/s) = Darcy velocity in the given direction, Ts
the thermal conductive and advective powers are esti- and Tes are the same as those used in Equation 2.
mated by Péclet number. Following, the influence of In order to analyze precisely the effect of ground-
the active length of an energy diaphragm wall on the water flow, the dimensionless parameter Péclet num-
heat exchange phenomenon is considered. Thereafter, ber Pe is introduced, defined as the ratio of heat
the effect of the soil thermal properties is studied to transferred by convection to the heat transferred by
assess its influence regarding the performance of the conduction:
system. Finally, the cyclic thermal loading is taken
into account to study its impact on the variation of the
exchanged heat.
where λs (W/m/K) = thermal conductivity of the soil

and Lc (m) = characteristic length. The characteristic
2 ASSESSMENT OF ALLOWABLE THERMAL
length depends on the type of the problem but can be
POWER
considered as the ratio of the soil volume to the surface
area exchanging heat.
2.1 Power transferred by conduction
In all ground conditions, conductive heat transfer
exists when there is a temperature difference between 3 NUMERICAL MODELING
two mediums independently of the water flow condi-
tions. By an analogy to a resistive model, the power 3.1 Case study of a metro station
transferred by conduction can be expressed as follows
Diaphragm walls are usually used as bearing elements
(Fromentin & Pahud, 1997):
for towers, metro stations, and underground parkings.
They represent an interest for geothermal structures
because they possess a large exchange area with the
surrounding soil. The objectives in this study are to
study the effect of groundwater flow, the active length
of the diaphragm wall, and the soil thermal proper-
where the exchanged power Pc (x, t) = average ther- ties on the heat exchange phenomena between energy
mal power transferred by conduction between the diaphragm walls and the soil.
energy foundation and the soil volume situated at The active length of the diaphragm wall is the part
a distance x from the foundation and at time t, into which the exchanger elements are installed and
Tes (K) = temperature of the energy foundation which it is not always equal to the total length of the wall.
is considered as the average heat carrier fluid tempera- The choice of the length and the position of the heat
ture (Tes = (Tin + Tout )/2) for each season, and Ts (x, t) exchanger elements in the energy walls plays an impor-
(K) = temperature of the soil located at a distance x tant role in the heat exchange. These tubes may be put
from the foundation and affected by the heat exchange all along the length of the diaphragm wall or only along
phenomenon, and at time t. its embedment depth. Moreover, some energy walls
In this model, the soil and the energy diaphragm may be equipped with heat exchanger elements along
wall are considered as two resistors connected in series the far field face which is the face near the soil and the
(Fig. 1). The equivalent thermal resistance is the sum near field face which is the face near the excavation,
of the thermal resistances of the energy foundation Res while others may be installed only along the far field
and the soil Rs (m.K/W). face (Amis et al. 2009).
164
Table 1. Hydraulic boundary conditions and flow velocity.
Depth of the water

table (m) Average flow velocity
(m/s)
Left side Right side
(x = 0 m) (x = 60 m) kh = 10−5 m/s kh = 10−6 m/s
Case 1 50 30 3.34 × 10−6 3.34 × 10−7

Case 2 50 35 2.50 × 10−6 2.50 × 10−7
Case 3 50 45 8.34 × 10−7 8.34 × 10−8
Table 2. Hydraulic and thermal properties of soil and

concrete.
Soil Concrete
Hydraulic conductivity (m/s) 10−6 –10−5 –

Porosity 0.4 0.15
Specific heat capacity (J/kg/K) 1000 880
Figure 2. Geometry of the considered metro station. Thermal resistance (m.K/W) 0.5 0.06
In the following study, the wall is supposed to be

equipped with heat exchanger tubes from the both boundaries, and at the metro diaphragm walls in con-
faces (near field and far field faces). Two cases are tact with air at the exterior side. The top of the layer
considered depending on the active length of the is subjected to surface temperature variations, while
exchanger elements: the first case is the wall W1 with an internal heat flux of 0.0544 W/m2 is imposed at the
the exchanger elements along its whole length, and base. The initial ground temperature is set as 14◦ C. The
the second case is the wall W2 with the exchanger temperature of the energy diaphragm walls is fixed
elements only along its embedment length. to 5◦ C in winter and 25◦ C in summer, correspond-
ing to average values usually noted in these structures
(Suryatriyastuti, 2013). The assumption of imposing
a constant temperature in the energy walls seems suf-
3.2 Presentation of the model
ficient to study the thermal transfer with the ground
The numerical modeling is conducted with the finite during a short period where transient effects can be
difference software FLAC3D (ITASCA, 2005). The neglected as a first attempt.
soil volume is modeled in three dimensions over one
meter thickness and presents a vertical length of 50 m
3.3 Ground and material properties
and a width of 60 m. These values are chosen on the
base of a preliminary parametric study in a way to In the model there are two different materials: con-
be able to assess the effect of the water flow on the crete for the energy wall and the surrounding saturated
temperature profile near the wall. The dimensions of sandy soil. Table 2 presents the thermal and hydraulic
the diaphragm wall and the raft foundation constitut- properties used in this study for the two materials.
ing the modeled metro station are inspired from the The thermal resistances used for the calculation
project “Grand Paris” (Fig. 2). Each diaphragm wall of the allowable power transmitted by conduction
has 32.5 m length and 1.2 m width. The embedded are determined according to two approaches. For the
parts are of 10.5 m length. The raft has a surface of soil, the thermal resistance can be considered as the
21 m × 1.5 m. inverse of its thermal conductivity: Rs = 1/λs . Numer-
Concerning the hydraulic boundary conditions, the ous theoretical and experimental studies have shown
top and the base of the domain and the walls are consid- the relation between the soil thermal conductivity and
ered impermeable, whereas hydraulic flux is imposed its nature, granulometry, compacity, and water content.
at the left and right boundaries to generate ground- In this study, the following correlation is chosen to
water flow. The soil is considered homogeneous with describe the thermal conductivity of the studied sandy
a hydraulic conductivity value kh (m/s). Six cases are soil (Kersten, 1949):
considered to take into account three types of boundary
conditions and two values of the hydraulic conduc- 1
= λs = 0.1442(0.7 log w + 0.4).100.6243ρd (5)
tivity coefficient. For each case, the average Darcy Rs
velocities are presented in Table 1.
Concerning the thermal boundary conditions, adi- where w = water content of the soil and ρd
abatic boundary conditions are imposed at the lateral (kg/m3 ) = its dry density.
165
Table 3. Thermal resistance of piles SIA (2005). Table 4. Péclet number for each case.
Pile Total thermal Pe (kh = 10−5 m/s) Pe (kh = 10−6 m/s)

diameter resistance
Pile type (m) (m.K/W) Case 1 7.0 0.7
Case 2 5.2 0.52
Driven tube with 0.3–1.5 0.15 Case 3 1.75 0.18
double U-tube
Double 0.3–1.5 0.10–0.11
U-tube∗
Precast or cast Triple 0.3–1.5 0.07–0.08
in situ U-tube∗
In order to analyse the heat exchange in details, the
Quadruple 0.3–1.5 0.06 total average allowable thermal power Ptotal between
U-tube∗ the wall and the soil is defined as follows:
∗
U-tube attached to reinforcement.
where P̄ is the average of the allowable thermal

exchanged power P(x, t), and P(x, t) corresponds to
the sum of the power exchanged by conduction and by
advection between the wall and the soil situated at dis-
tance x from the wall and at time t during the loading
season. The characteristic length Lc , in this case i–t is
equal to 1 m; the thickness of the studied model.
4.2 Variation of the allowable exchanged power

with distance and time
Figure 4 presents the variation of the allowable thermal
power along the x-direction for case 1 (kh = 10−5 m/s)
Figure 3. Temperature profile around the diaphragm
for the right and the left wall, along the downstream
wall-kh = 10−5 m/s – a) case 1 b) case 2 c) case 3.
side. The allowable power increases till a certain dis-
tance. For the first 20 days of the considered heating
period, the allowable power increases till a distance
4.5 m from the edge of the right wall, and then the
For the concrete, the values indicated by the
power takes a constant value, whereas for the left wall,
Swiss Society of Architects and Engineers are taken
the increase in the allowable power is up to a distance
(Table 3). Considering the geometry of the diaphragm
6 m from the edge of this wall.
wall, it can be considered that the wall is equivalent to
Figure 5 presents the variation of the total allow-
a bored pile equipped with 4 U-tubes whose diameter
able thermal power exchanged by conduction Pc and
ranges between 0.3 and 1.5 m.
by advection Pv for the whole system along the heat-
ing season. As time increases the power exchanged by
advection decreases, and thus the efficiency of the sys-
tem. On the other hand, the conductive power can be
4 THERMAL PERFORMANCE IN WINTER considered to have a constant value during the whole
season.
4.1 Influence of ground water flow Figure 6 represents the variation of the total allow-
The simulations presented in this section are carried able thermal power Ptotal with Péclet number in loga-
for one winter season. The influence of water flow on rithmic scale. Ptotal increases with Pe : for Pe > 1.7,
the temperature profile at the end of the loading season power of the order of 35 W/m2 can be exchanged.
is presented in Figure 3, which shows that the effect These values are greater than those recommended by
of advection becomes significant as the water flow Brandl (2006) that are 30 W/m2 .
increases. Péclet number for each case is presented in The variation of the total allowable thermal power
Table 4 for two values of hydraulic conductivity kh : exchanged by conduction Pc and by advection Pv
10−5 and 10−6 m/s. is presented in Figure 7. It can be noted that the
166
Figure 6. Variation of Ptotal as a function of Péclet number.
Figure 4. Variation of the allowable thermal exchanged

power along x direction for the right and left wall.
Figure 7. Influence of conduction and advection on the

Figure 5. Variation of the total allowable thermal exchanged
exchanged powers Pc and Pv .
power during the heating season (case 1, kh = 10−5 m/s).
thermal power exchanged by conduction is approxi- configurations: the wall totally equipped (W1) and the
mately constant and tends to decrease slightly as Pe wall with only the embedment part being equipped
increases, also the effect of conduction becomes more (W2) with the heat exchanger elements. Figure 8
important as Pe decreases (for low hydraulic conduc- shows the difference in the exchanged power between
tivity, kh = 10−6 m/s). On the other hand, the power these two configurations. The equipment of the
exchanged by advection Pv increases linearly with Pe . whole wall induces an enhancement in the allowable
exchanged power which appears to be more impor-
4.3 Influence of the active length of the wall tant for high water flow. For a water velocity of
334 × 10−6 m/s the power exchanged increases by 32
The effect of the total length of the wall is ana- %, and this increase only reaches 10 % for lower water
lyzed through the comparison of two different velocities of the order 8.34 × 10−7 m/s.
167
Figure 8. Ratio of the total thermal power between the
configurations W1 and W2 (kh = 10−5 m/s). Figure 9. Variation of the total allowable thermal power with
the thermal conductivity of the soil.
4.4 Effect of soil thermal properties on the heat
exchange phenomena with time during each period. Figure 11 presents the
variation of the exchanged power for the three heating-
The thermal properties as the thermal conductivity and cooling cycles in the zone near the structure which is
specific heat capacity may have an impact on the heat being affected by the heat exchange. Starting with the
exchange between the soil and the energy foundation. heating period, the exchanged power is approximately
The thermal conductivity of the soil is increased to constant for small water flow velocities, and then it
3 W/m/K and the specific heat capacity to 1500 J/kg/K. decreases slightly at the end of this period. This can
It was noticed that for the same hydraulic conditions be interpreted by the fact that the temperature in the
imposed before (kh = 10−5 m/s), the heat exchange soil surrounding the wall starts to decrease along this
increases slightly as these parameters increase. From period due to continuous heat extraction from the soil.
Figure 9, it is clear that the increase in the exchanged Then for the cooling period, negative values indicate
heat is more evident for case 3 with the smallest water that heat is being injected into the soil, thus increasing
velocity than the other two cases. This may be related the soil temperature near the foundation. The allow-
to the relative effect of the conductive and advective able exchanged power decreases during the loading
terms. As the conductive term presented by the ther- period. This slight decrease of the exchanged power
mal conductivity becomes more important, the effect in the heating or in the cooling period is obvious for
of the advective term presented by the water velocity important water flow conditions, whereas for small
becomes less significant. This result is in accordance water flow, the variation has a constant profile for each
with Di Donna & Barla (2016). period.
The power exchanged is greater in the cooling
period than in the heating one, this is due to the fact
5 EFFECT OF CYCLIC LOADING ON THE that at the beginning of the cooling period, the ground
ALLOWABLE EXCHANGED THERMAL temperature is the lowest and thus the difference in
POWER temperature between the soil and the wall is the great-
est, leading to the greatest exchanged power. Then the
The temperature variation near an energy diaphragm soil temperature rises up causing a decrease in the heat
wall affected by heating-cooling cycles is presented in exchange.
Figure 10. In this figure, the temperature variation is During the second thermal loading cycle, the heat
measured from the edge of the right wall and for three transferred between the soil and the wall increases
heating-cooling cycles where each cycle consists of a significantly in both periods, since the heat is stored
heating period followed by a cooling period (Hi and Ci during the cooling period in the soil surrounding
are the heating and cooling periods of cycle i respec- the wall, thus increasing the efficiency of the heat
tively for i varying from 1 to 3). Along the horizontal exchange.
distance, as the distance from the edge of the wall During the third heating-cooling cycle, the
increases, the temperature increases during the heat- exchanged power during the two periods drops down
ing period and decreases during the cooling period till compared to the second cycle; this is due to the fact
reaching a constant value for both periods equal to the that part of the heat stored in the soil near the wall
initial ground temperature at a distance 6 m away from in cycle 2 disseminates to points far away from the
the edge of the wall. Katzenbach et al.(2008) obtained wall edge, this lowers the geothermal potential of the
similar results for the variation of temperature near a ground around the wall and thus decreases the total
bored pile wall for a section parallel to the water flow power. In addition, with time the efficiency of the
direction. system starts decreasing.
Concerning the allowable power exchange during According to our results, the power exchange is
the thermal cycles, for the first loading cycle, the greater in the cooling period than in the heating period.
power exchanged between the soil and the wall varies This may be related to the thermal properties of the soil
168
with heat exchanger tubes compared to wall equipped
with the tubes only along its embedment length. On
the other hand, this enhancement is considered to be
negligible for small groundwater velocities.
Soil thermal properties are also proved to have an
influence on the allowable power exchanged between
the soil and the wall, where this influence appears to
be significant for low water flow velocities.
The variation of the allowable exchanged power
affected by thermal cyclic loading was evaluated
through considering the loading cycle as a heating
period followed by a cooling period. It is obtained that
thermal cyclic loading leads to the depletion of the
power transferred by the soil volume.
Finally, adequate understanding of the local geol-
Figure 10. Variation of temperature during the loading ogy and groundwater flow is of great importance to
cycles. estimate the impact of groundwater flow, soil ther-
mal properties, the active length of the wall, and the
cyclic thermal loading on the performance of energy
diaphragm walls, and to adopt the best solution from an
economical and thermal point of view. Further study
should be carried regarding the behavior of energy
diaphragm walls under coupled thermo-mechanical
loading.
REFERENCES
Adam, D. & Markiewicz, R. 2009. Energy from earth-coupled
structures, foundation, tunnels and sewers. Géotechnique
59(3): 229–236.
Amis, T., Robinson, C.A., & Wong, S. 2009. Integrat-
ing geothermal loops into the diaphragm walls of the
Knightbridge palace hotel project.
Bouazza, A., Laong, B., & Sigh, R. 2013. Soil effective ther-
mal conductivity from energy pile thermal tests. Coupled
Figure 11. Variation of the exchanged power under cyclic phenomena in environmental geotechnics. London: CRC
loading (kh = 10−5 m/s). Press.
Brandl, H. 2006. Energy foundations and other thermo-active
that seems to exhibit a higher potential to gain heat in ground structures. Géotechnique 56(2): 81–122.
Cecinato, F. & Loveridge, F.A. 2015. Influences on thermal
the cooling period than to lose heat in heating period.
efficiency of energy piles. Energy 82: 1021–1033.
These results are consistent with those obtained by Cervera, C.P. 2013. Ground thermal modeling and analysis
Hassani Nezhad Gashti et al. (2015). of energy pile foundations.
Di Donna, A. & Barla, M. 2016. The role of ground con-
ditions on energy tunnels’ heat exchange. Environmental
6 CONCLUSION geotechnics.
Fromentin, A. & Pahud, D. 1997. Recommendations pour
la réalisation d’installation avec pieux échangeurs. Laus-
Numerical analyses conducted in this paper aim to
sane: 247 pages.
focus on the impact of the presence of water flow Gao, J., Zhang, X., Liu, J., Shan, Li K., & Yang, J. 2008.
on the transfer phenomena between energy diaphragm Thermal performance and ground temperature of vertical
walls and soil. Significantly, for high groundwa- pile foundation heat exchangers: A case study. Applied
ter flow presented by high hydraulic conductivity thermal engineering 28: 2295–2304.
and Darcy velocity, the allowable exchanged thermal Hassani Nezhad Gashti, E., Malaska, M., & Kujala, K. 2015.
power increases, due to the effect of coupled conduc- Analysis of thermo-active pile structures and their perfor-
tive and advective phenomena. For this, the presence mance under groundwater flow conditions. Energy and
of water flow should be taken into account for the Buildings 105: 1–8.
ITASCA. 2005. FLAC3D Version 3.0 Theory and Back-
evaluation of the performance of the system.
ground. Itasca consulting group, Inc. Minnesota, USA.
It is found that the heat transfer process is a func- Katzenbach, R., Clauss, F., Waberseck, T., & Wagner, I. 2008.
tion of the active length, thus it should be carefully Coupled numerical simulation of geothermal energy sys-
considered for the dimensioning of energy walls. In tems. International association for computer methods and
the presence of high groundwater flow, the transfer advances in geomechanics; Proc. intern. conf., Goa, 1–6
is enhanced in the case of a wall totally equipped October 2008. India: Curran associates, Inc.
169
Kersten, M. 1949. Thermal properties of soils. Minnesota, Sigfusson, B. & Uihlein, A. 2015. 2014 JRC Geothermal
LII(21): 233 pages. energy status report: 64 pages.
Link, K., Rybach, L., Inhasly, S., & Wyss, R. 2015. Geother- Sterpi, D., Angelotti, A., Corti, D., & Ramus, M. 2014.
mal energy in Switzerland. World geothermal congress; Numerical analysis of heat transfer in thermo-active
Proc. intern. conf., Melbourne, 19–25 April. Australia. diaphragm walls. Numerical methods in geotechnical
Lund, J.W., Freeston, D.H., & Boyd, T.L. 2010. Direct utiliza- engineering.
tion of geothermal energy 2010 worldwide review. World Suryatriyastuti, M. 2013. Numerical study of the thermo-
geothermal congress; Proc. intern. conf., Bali, 25–29 active piles behavior in cohesionless soils, Lille: 143
April. Indonesia. pages.
Ma, X. & Grabe, J. 2010. Efficiency increase of soil heat Weber, J., Ganz, B., Schellschmidt, R., Sanner, B., & Schulz,
exchangers due to groundwater flow and air injection R. 2015. Geothermal energy use in Germany. World
wells. World geothermal congress; Proc. intern. conf., geothermal congress; Proc. intern. conf., Melbourne,
Bali, 25–29 April. Indonesia. 19–25 April. Australia.
SIA. 2005. Utilisation de la chaleur du sol par des ouvrages Xia, C., Sun, M., Zhang, G., Xiao, S., & Zou, Y. 2012. Exper-
de fondation et de soutènement en bèton, guide pour la imental study on geothermal heat exchangers buried in
conception, la réalisation et la maintenance. Documen- diaphragm walls. Energy and buildings 52: 50–55.
tation D0190-Swiss society of engineers and architects,
Geneva, 100 pages.
170
Analysis of thermally induced mechanical interactions in energy pile groups
Alessandro F. Rotta Loria & Lyesse Laloui

Swiss Federal Institute of Technology in Lausanne, Laboratory of Soil Mechanics, Lausanne, Switzerland
ABSTRACT: This study investigates the thermally induced mechanical interactions among closely spaced
energy piles that partially operate as geothermal heat exchangers over a time-scale that is typical of practical
applications. The analysis is based on the results of a full-scale in-situ test of a group of energy piles and a
coupled 3-D thermo-mechanical finite element analysis. The work highlights two types of thermally induced
mechanical interactions in energy pile groups, i.e., first- and second-kind interactions. The former interactions
develop during early stages of geothermal operations of energy piles. The latter interactions develop during
successive stages of geothermal operations of energy piles. The impact of these interactions on the variation of
the mechanical behaviour of energy pile groups varies with time. Attention must be devoted to these interactions
throughout the design process (e.g., geotechnical and structural) of energy piles because they play an important
role in the serviceability performance of these foundations.
1 INTRODUCTION other (e.g., widely spaced pile groups), their individual

responses can be considered independent and compa-
Over the last decade, energy piles have shown potential rable to the case of an isolated pile. However, when
to couple the structural support role of conventional piles in a group are close enough to each other (e.g.,
pile foundations with the role of geothermal heat closely spaced pile groups), their individual responses
exchangers to satisfy the energy needs of the build- are influenced by the neighbouring piles and differ
ing environment. Because of this twofold operation, from that of an isolated pile. In the latter case, the
energy piles are subjected to multi-source actions: influences between the individual pile responses of the
thermal and mechanical loads. group represent interactions (e.g., mechanical). These
The thermal loads that are applied to energy piles interactions occur between the piles, the connecting
represent an innovative challenge for geotechnical slab and the surrounding soil. They have been shown to
and structural engineers because they induce unprece- manifest through so-called group effects, to control the
dented thermally induced effects on the mechanical response of pile groups to loading and to deserve to be
response of these foundations. considered for an optimal design of pile foundations.
For over ten years, a large amount of research has To date, despite the increasing research on energy
investigated the impact of thermally induced effects pile groups, knowledge of the development and impact
on the thermo-mechanical behaviour of single iso- of thermally induced group effects and interactions
lated energy piles. Because it has been proved that among closely spaced energy piles on their thermo-
the impact of thermally induced effects on the stress mechanical behaviour has been preliminary due to
and displacement variations in single energy piles can the lack of field data about the exploitation of energy
be comparable to the impact of the superstructure piles that either partially or entirely operate as geother-
mechanical loads that are applied to these founda- mal heat exchangers for time-scales that are typical of
tions, increasing efforts are devoted to propose codes practical applications.
of practice and standards for an optimal design of these To address this challenge, Rotta Loria & Laloui
ground structures. (2016) recently carried out a full-scale in-situ test
In recent years, an increasing amount of research of a group of closely spaced energy piles that par-
is also being devoted to investigate the impact of tially operate as geothermal heat exchangers over a
thermally induced effects on the thermo-mechanical time-scale that is typical of practical applications, and
behaviour of energy pile groups (Salciarini et al., performed a coupled 3-D thermo-mechanical finite
2013; Jeong et al., 2014; Di Donna & Laloui, 2014; element analysis.
Mimouni & Laloui, 2015; Suryatriyastuti et al., 2015; This paper expands on the analyses and results of
Di Donna et al., 2016). the aforementioned experimental and numerical study
Extensive amounts of research (e.g., Poulos & and proposes an analysis of the thermally induced
Davis, 1980; Fleming et al., 2008) have proved that mechanical interactions among closely spaced energy
when piles in a group are located far enough from each piles.
171
Figure 1. (a) The EPFL Swiss Tech Convention Centre (http://www.tstcc.ch/, author: Frédéric Rauss); (b) plan view of the
foundation including the four energy piles; (c) schematic of the soil stratigraphy.
2 EXPERIMENTAL TESTING
each of the conventional piles. The energy piles were
equipped with four 24-m-long high-density polyethy-
2.1 The foundation and site
lene U-loops that are connected in series. The inlets
The pile foundation that was considered for the exper- and outlets of the absorber pipes were thermally insu-
imental test is located under the recently built Swiss lated. The top of the U-loops were installed 4 m below
Tech Convention Centre, Lausanne, Switzerland (cf., the pile heads to limit the influence of the climatic
Fig. 1 (a)). The foundation supports a 9 × 25 m2 water conditions on the heat exchange process.
retention tank and comprises a group of four end- All of the energy piles were instrumented with
bearing energy piles (labelled EP1, EP2, EP3 and EP4 strain gauges, optical fibres and thermocouples along
in Fig. 1 (b)) and sixteen semi-floating conventional their lengths as well as with pressure cells at their
piles (labelled P1-16 in Fig. 1 (b)) below a heavily rein- toes. Piezometers and thermistors were installed in
forced 0.9 m-thick slab. In plain view, the energy piles two boreholes in the soil. More detailed information
form a triangle within a 4.21 m square in which the on these instruments, which allow the thermo-hydro-
central pile is located 3 m from the others. The energy mechanical response of the foundation to be monitored
piles are 28 m long and 0.9 m in diameter, and the con- during the simulation of different operations of the
ventional piles are 16 m long and 0.6 m in diameter. energy piles via the use of a dedicated heating mod-
All of the piles were bored, cast onsite and are made ule, are reported by Mimouni & Laloui (2015), and
of reinforced concrete. Vertical loads of 0, 800, 2200 Rotta Loria & Laloui (2016).
and 2100 kN are applied to energy piles EP1, 2, 3 and The soil stratigraphy of the site (cf., Fig. 1 (c)) was
4, respectively. Vertical loads of 300 kN are applied to extrapolated based on information that was obtained
172
Figure 2. Geometry and boundary conditions of the finite element model of the entire foundation.
during the construction of the foundation and data piles through linear entities in which a heat carrier
from Laloui et al. (2003; 2006) for another exper- fluid is assumed to flow, which allows the problem
imental site that is located 200 m away. During the of the heat exchange that occurs in the pipes-pile-soil
construction of the piles, the groundwater table was system to be considered.
located at the top of the deposit, which is estimated
to be in an overconsolidated state condition (over-
consolidation ratio of OCR ≈3–4). Layers of alluvial 3.2 Modelling choices
soil and sandy-gravelly moraine were encountered at The numerical analysis of the response of the rein-
shallow depths. The upper soil profile of the alluvial forced concrete foundation in the soil under mechan-
soil was inferred to reach a depth from the uppermost ical and thermal loads is based on the following
surface of the successively built slab of z = −8.6 m. assumptions: (i) the displacements and deformations
The lower sandy-gravelly moraine layer was located of all of the materials can be representatively described
between depths of z = −8.6 and −16.6 m (Laloui through a linear kinematic approach under quasi-static
et al., 2003; 2006). A thin layer of bottom moraine conditions (i.e., negligible inertial effects); (ii) the
was present below the sandy-gravelly moraine layer materials that constitute the pile foundation are consid-
between depths of z = −16.6 and −20.1 m and laid on ered to be isotropic with pores that are fully filled by air
a molasse layer. The energy piles were socketed 8.8 m and are assumed to be purely conductive domains with
into this bottom molasse layer. equivalent thermo-physical properties that are given
by the fluid and the solid phases; (iii) the materials
that make up the soil layers are assumed to be isotropic,
2.2 Features of the experimental test fully saturated by water and purely conductive domains
with equivalent thermo-physical properties that are
The experimental test involved the application of a
given by the fluid and the solid phases; (iv) the loads
heating-passive cooling cycle to energy pile EP1 (for
that are associated with this problem have a negligible
approximately 5 and 10 months, respectively), which
impact on the variation of the hydraulic field in the soil;
was the only energy pile of the group that operated as a
and (v) the materials that compose the foundation and
geothermal heat exchanger (cf., Fig. 1 (b)). This paper
surrounding soil are considered to be representatively
devotes particular attention to the heating phase of the
described by linear thermo-elastic behaviours.
test. Throughout the test, the mechanisms and phe-
Under these conditions, a thermo-mechanical math-
nomena occurring in the operating energy pile EP1, in
ematical formulation is employed. Interested read-
the three surrounding non-operating energy piles EP2,
ers can find more detailed information about the
3 and 4, and in the soil were recorded.
mathematical formulation exploited for the numerical
analysis in the work of Batini et al. (2015).
3 NUMERICAL MODELLING
3.3 Boundary and initial conditions
3.1 Finite element model
Restrictions are applied to both the vertical and hor-
A 3-D finite element model of the site was developed izontal displacements on the base of the model (i.e.,
using the software COMSOL Multiphysics (COM- pinned boundary) and to the horizontal displacements
SOL, 2014) (cf., Fig. 2). The model reproduces the on the sides (i.e., roller boundaries). The initial stress
foundation supporting the water retention tank. It also state due to gravity in the foundation and the soil is
accounts for the presence of the pipes in the energy considered to be geostatic and assumes a coefficient
173
of Earth pressure at rest of K0 = 1. No residual stresses The experimental and numerical data include vari-
from the installation of the piles are considered in ations of the parameters from the beginning of the
these elements and in the adjacent region of soil. This test over time. Therefore, they reflect the impact of
hypothesis may not be completely representative of the geothermal operation of energy pile EP1 on the
reality but can be applied successfully in almost all thermo-mechanical behaviour of the foundation.
methods of pile groups deformation analysis by choos-
ing appropriate values of the soil moduli (Poulos &
Davis, 1980). The temperature is fixed on each of the 4.1 Temperature variations along the
external boundaries of the model (T = 13.3◦ C). The energy piles
initial temperatures in the pipes, energy and conven-
tional piles, slab and soil are set to T0 = 13.3◦ C, which Figure 3 presents the temperature variations that were
is the average temperature that was recorded at the observed experimentally and numerically along the
beginning of the experimental test between depths of lengths of the operating energy pile EP1 and of the
z = −4.9 and −28.9 m from the surface of the site (this non-operating energy piles EP2, 3 and 4.
temperature corresponds to the portions of the energy The geothermal operation of energy pile EP1
piles that are not thermally insulated). The fluid that involved average temperature changes along its unin-
circulates inside the pipes is water. The inner diam- sulated portion of T = 5, 10, 15 and 20◦ C after
eter of the pipes is φ = 26.2 mm (the outer diameter t = 2, 8, 35, and 156 days, respectively. These changes
is 32 mm, and the wall thickness is 2.9 mm). A ther- were observed in both the experimental and numerical
mal conductivity of λp = 0 W/(m K) is imposed in the results (cf., Fig. 3 (a)).
shallowest 4 meters of the inlet and outlet of the pipes After t = 2 and 8 days (i.e., during the early stages of
to simulate the thermal insulation near the ground sur- the heating phase of energy pile EP1), the correspond-
face. The trends of the inlet temperature and velocity of ing portions of the non-operating energy piles EP2,
the fluid in the pipes that were experimentally recorded 3 and 4 were characterised by no changes in temper-
throughout the test are considered as input parameters ature. However, temperature changes were observed
for the numerical simulation. over time because heat diffused through the soil from
EP1 and indirectly heated them. After t = 35 and 156
days (i.e., during the late stages of the heating phase
3.4 Classification of the numerical simulation and of EP1), heat diffusion resulted in average experimen-
material properties tal temperature variations of T = 1.6, 0.7 and 1.1◦ C
and T = 5.3, 3.6 and 4.5◦ C, respectively. The numer-
The numerical analysis that is described in this paper ical results showed slightly higher average temperature
is considered a Class C1 prediction (Lambe, 1973) changes than the experimental results (cf., Fig. 3(b-d)).
because it was carried out after the modelled in-situ This difference was attributed to differences between
test was performed and the associated results were the actual and modelled heat diffusion processes. The
available. differences between the actual and modelled heat dif-
The material properties that were used for the fusion processes in the foundation were inferred to be
numerical analysis (Rotta Loria & Laloui, 2016) are caused by (i) potential inhomogeneity (spatial and of
those that were recently proposed by Di Donna et al. material properties) of the soil layers of the site that
(2016) for the characterisation of the site with two were not accounted for in the numerical model and (ii)
main changes. These changes were made to accu- different positions of the pipes inside EP1 than those
rately represent the thermo-mechanical behaviour of that were considered in the simulation.
the foundation during the late stages of geothermal In addition to the observed temperature changes
operation and included the linear thermal expansion with time that corresponded to the uninsulated por-
coefficient of layers B, C and D, as well as the ther- tion of energy pile EP1, temperature variations also
mal conductivity of the solid particles of all of the occurred in the shallowest 4 m of EP2, 3 and 4 even
soil layers. The variation of the former parameter was though the pipes of EP1 were thermally insulated at
based on the ranges of variability that typically char- these depths. This behaviour was observed in both the
acterise the thermal expansion coefficient of moraine experimental and numerical results and was attributed
and molasse deposits in the geographical area of Lau- to the impact of the heat exchange operation of energy
sanne. The modification of the latter parameter was pile EP1 on the shallower portions of the other piles.
based on the determination (without accounting for The experimental results further indicated changes in
capacitive effects) of an effective thermal conductiv- temperature at the surface of the foundation, which
ity for the soil deposit of λeff = 2.78 W/(m K) based were attributed to the variation of the surface thermal
on the experimental data. conditions during the experimental test. The numerical
results showed slightly smaller temperature variations
in the shallowest 4 m of all of the piles than the exper-
4 COMPARISON BETWEEN EXPERIMENTAL imental results. They also indicated that no changes in
AND NUMERICAL RESULTS temperature occurred at the surface of the foundation.
These results were consistent with the fixed tempera-
This section presents a comparison between the exper- ture boundary condition that was imposed on the top
imental and numerical results. surface of the numerical model.
174
Figure 3. Comparison between the experimental and numerical temperature variations observed along (a) the operating
energy pile EP1 and (b-d) the non-operating energy piles EP2, 3, and 4, after t = 2, 8, 35 and 156 days of testing.
4.2 Vertical strain variations along the

(i.e., after t = 2, 8, 35, and 156 days, respectively).
energy piles
Maximum positive (contractive) vertical strains varia-
Figure 4 shows the variations in vertical strain that tions of εν = 31, 56, 68 and 79 µε were recorded
were observed experimentally and numerically along along the insulated portion during the same stages
energy piles EP1, 2, 3 and 4. of the test. Marked negative vertical strains were
The heating due to the geothermal operation of observed with time in the bottom portion of this
energy pile EP1 during the first phase of the test pile. Similar results were obtained by the numerical
resulted in an expansion of the portion of EP1 in which analysis.
the pipes were not thermally insulated and a com- The heating of the operating energy pile EP1
pression of the thermally insulated portion because also induced an expansion of the surrounding non-
of the entrapment with the slab (cf., Fig. 4 (a)). operating energy piles EP2, 3 and 4 (cf., Fig. 4 (b–d)).
Maximum negative (expansive) vertical strains of After t = 2 and 8 days (i.e., during the early stages of
εν = −22, −56, −109 and −167 µ ε were recorded the heating phase of EP1), the expansions of EP2, 3
along the uninsulated portion of energy pile EP1 and 4 were caused by (i) the negative strain of pile
during the experiment when it was subjected to temper- EP1 as a result of its direct heating and (ii) the associ-
ature changes of T = 5, 10, 15 and 20◦ C, respectively ated upward deformation of the slab. This deformation
175
Figure 4. Comparison between the experimental and numerical variations in vertical strain observed along (a) the operating
energy pile EP1 and (b–d) the non-operating energy piles EP2, 3, and 4, after t = 2, 8, 35 and 156 days of testing.
176
was thus purely mechanical. However, the evolution are evidenced through thermal and thermally induced
of deformation along the piles (decreasing from top to mechanical interactions between the operating and
bottom) indicates that the deformation was compara- non-operating energy piles.
ble to that caused by an upward force applied at their The thermal interactions between the piles appear
heads. After t = 35 and 156 days (i.e., during the late during successive stages of geothermal operations.
stages of the heating phase of EP1), the expansions The thermally induced mechanical interactions
of piles EP2, 3 and 4 were caused by (i) the negative between the piles are always present throughout the
strain of EP1 as a result of its direct heating, (ii) the geothermal operations.Two types of thermally induced
associated upward deformation of the slab, (iii) the mechanical interactions can be distinguished: first-
negative strains of these elements as a result of their and second-kind interactions. First-kind interactions
indirect heating and (iv) the negative strain in the soil develop during early stages of geothermal operations
as a result of its heating. In contrast to the deforma- of energy piles and are primarily caused by the direct
tion of the non-operating energy piles EP2, 3 and 4 heating and associated thermally induced deformation
during the early stages of the geothermal operation of of the operating energy piles. Second-kind interactions
energy pile EP1, this deformation was characterised by develop during successive stages of geothermal oper-
both mechanical and thermal contributions. Marked ations of energy piles and are caused by (i) the direct
negative vertical strains of up to εν = −106 µε were heating and related thermally induced deformation of
observed during these stages in the lower portions of the operating energy piles, and (ii) the indirect heating
EP2, 3 and 4 in both the experimental and numer- and related thermally induced deformation of the soil
ical results. These negative strains were significantly surrounding the operating energy piles as well as of the
greater than those that developed in the top portions of non-operating energy piles. The presence of the slab
these elements (i.e., between εν = −10 and −30 µε). is key for the development of all of these interactions.
They were also greater than those under free ther- The rate of the heat exchange process characteris-
mal expansion conditions, which can be calculated ing energy pile groups also appears to be crucial for
according to a one-dimensional scheme as the development and magnitude of these interactions.
The soil-pile thermal expansion coefficient ratio, the
pile-soil stiffness ratio and the slab-soil stiffness ratio
(Rotta Loria & Laloui, 2016) further appear to be very
where αEP is the linear thermal expansion coefficient important characteristics. The soil-pile thermal expan-
of the energy pile and T is the observed temperature sion coefficient ratio has the greatest influence on the
variation. thermally induced mechanical behaviour of the piles
The marked expansive vertical strains that were during successive stages of geothermal operation. The
observed in the bottom portions of all of the piles experimental and numerical results expand upon the
during the late stages of the heating phase of EP1 key role of the soil-pile thermal expansion coefficient
occurred because as heat diffused through the sys- ratio that was recently described by Bourne-Webb et al.
tem, the mechanical response of the foundation was (2015) for single isolated energy piles.
governed by the thermally induced deformation of the
molasse layer. These strain variations were caused not
only by the interplay between the thermally induced
deformations (direct and indirect) of the piles and the 6 CONCLUDING REMARKS
slab but also and primarily by the thermally induced
deformation of the soil mass (e.g., molasse layer) sur- This study investigated the development, magnitude
rounding the piles. The value of the thermal expansion and impact of the thermally induced mechanical inter-
coefficient of the molasse layer, which was found to actions that develop in energy pile groups as a conse-
be greater than that of the piles based on the results quence of the partial or entire operation of the piles as
of the numerical analysis, was the key factor of this geothermal heat exchangers on the thermo-mechanical
phenomenon. Heating the very stiff molasse layer over behaviour of such foundations. The main conclusions
time caused a marked expansion of this layer.This field that can be drawn from this work are:
was superimposed on the expansion field of the bot- • Thermally induced mechanical interactions among
tom portions of the piles. Remarkably high expansive energy piles are always present throughout geother-
vertical strains therefore developed in these settings. mal operations of these foundations.
• Two types of thermally induced mechanical inter-
actions can be distinguished: first- and second-kind
5 DISCUSSION interactions. First-kind interactions develop during
early stages of geothermal operations of energy
The results that were described in the previous sections piles and are primarily caused by the direct heating
demonstrate that the behaviour of groups of closely and associated thermally induced deformation of
spaced energy piles that operate partially as geother- the operating energy piles. Second-kind interactions
mal heat exchangers over time-scales that are typical develop during successive stages of geothermal
of practical applications is characterised by significant operations of energy piles and are caused by (i)
thermally induced group effects. These group effects the direct heating and related thermally induced
177
deformation of the operating energy piles, and (ii) energy piles for different design solutions. Computers and
the indirect heating and related thermally induced Geotechnics, 86, 199–213.
deformation of the soil surrounding the operating Bourne-Webb, P., Bodas Freitas, T. & Freitas Assunção, R.
energy piles as well as of the non-operating energy (2015) Soil–pile thermal interactions in energy founda-
tions. Géotechnique, 1–5.
piles. The magnitude and development of these Di Donna, A. & Laloui, L. (2014) Numerical analysis
interactions are governed by the interplay between of the geotechnical behaviour of energy piles. Interna-
the thermally induced responses of the operating tional Journal for Numerical and Analytical Methods in
and non-operating energy piles and soil to temper- Geomechanics, 39, 861–888.
ature changes. The presence of the slab represents Di Donna, A., Rotta Loria, A. F. & Laloui, L. (2016) Numer-
a key contribution for the development of all of the ical study on the response of a group of energy piles
aforementioned interactions. under different combinations of thermo-mechanical loads.
• The soil-pile thermal expansion coefficient ratio, Computers and Geotechnics, 72, 126–142.
the pile-soil stiffness ratio and the slab-soil stiff- Fleming, K., Weltman, A., Randolph, M. & Elson, K. (2008)
Piling engineering, CRC press.
ness ratio appear to be crucial characteristics for the Jeong, S., Min, H. & Lee, J. K. (2014) Thermally induced
development and magnitude of all of these inter- mechanical response of energy piles in axially loaded pile
actions. The former characteristic was shown to groups. Applied Thermal Engineering, 71, 608–615.
have the most influence on the thermally induced Laloui, L., Moreni, M. & Vulliet, L. (2003) Comportement
mechanical behaviour of both the operating and d’un pieu bi-fonction, fondation et échangeur de chaleur.
non-operating energy piles during successive stages Canadian Geotechnical Journal, 40, 388–402.
of geothermal operation. Significant attention must Laloui, L., Nuth, M. & Vulliet, L. (2006) Experimental
be paid to this parameter because it characterises and numerical investigations of the behaviour of a heat
the analysis and design (e.g., geotechnical and exchanger pile. International Journal for Numerical and
Analytical Methods in Geomechanics, 30, 763–781.
structural) of energy pile groups. Lambe, T. (1973) Predictions in soil engineering. Géotech-
• The rate of the heat exchange process that charac- nique, 23, 151–202.
terises energy pile groups controls the development Mimouni, T. & Laloui, L. (2015) Behaviour of a group
and magnitude of all of the thermally induced of energy piles. Canadian Geotechnical Journal, 52,
interactions in these foundations. 1913–1929.
Poulos, H. G. & Davis, E. H. (1980) Pile foundation analysis
and design, New York, Wiley.
Rotta Loria, A. F. & Laloui, L. (2016) Thermally
ACKNOWLEDGEMENTS induced group effects among energy piles. Géotechnique,
Submitted.
The financial support of N. 160117 (Division I- Salciarini, D., Ronchi, F., Cattoni, E. & Tamagnini, C.
III) from the Swiss National Science Foundation is (2013) Some remarks on the thermomechanical effects
acknowledged. induced by energy piles operation in a small piled
raft. International Journal of Geomechanics, 10.1061/
(ASCE)GM.1943-5622.0000375.
Suryatriyastuti, M., Burlon, S. & Mroueh, H. (2015)
REFERENCES On the understanding of cyclic interaction mechanisms
in an energy pile group. International Journal for
Batini, N., Rotta Loria, A. F., Conti, P., Testi, D., Grassi, W. & Numerical and Analytical Methods in Geomechanics,
Laloui, L. (2015) Energy and geotechnical behaviour of 10.1002/nag.2382.
178
Ground thermal response to borehole ground heat exchangers
O. Mikhaylova, I.W. Johnston & G.A. Narsilio

The University of Melbourne, Melbourne, Australia
ABSTRACT: Ground heat exchangers (GHEs) are the elements of ground source heat pump (GSHP) systems
that provide thermal interactions with the ground. Closed-loop borehole GHEs are commonly used for GSHP
systems in urban areas where land availability can be limited. The installation costs of borehole GHEs are
usually the largest component of the capital costs of GSHP systems. More research, including experimental
studies, into GHEs can improve GHE design and reduce their costs. This paper presents the first findings
of a full-scale experimental study of the ground thermal response to a 120 kW commercial GSHP system in
Melbourne, Australia. The system uses twenty-eight 50 m deep borehole GHEs to interact with the ground.
Several temperature monitoring boreholes were installed close to some of the GHEs to monitor their thermal
impact on the ground. The first sets of ground monitoring data are summarised to show trends of ground thermal
disturbance by real-life thermal loads.
1 INTRODUCTION The Elizabeth Blackburn School of Sciences

(EBSS) full-scale shallow geothermal installation in
Ground heat exchangers (GHEs) are the key elements Melbourne, Australia was designed to study ground
of ground source heat pump (GSHP) systems which thermal reactions to borehole GHEs. The GSHP sys-
provide thermal interactions between the ground and tem is heavily instrumented to monitor power and
the fluid circulating within the systems. The installa- energy input and output, GHE fluid temperatures
tion costs of GSHP systems largely depend on the costs throughout the system, ground temperatures around
of GHEs, so an optimum design of GHEs is impor- GHEs at different depths, undisturbed ground temper-
tant for financial feasibility of such systems. More atures, building thermal loads and on-site weather. The
research, including experimental studies, should be system was commissioned in March, 2014 and has
undertaken into GHE-ground thermal interactions, so been in continuous operation since.
that more informative design decisions can be made This paper presents some of the first observations
in sizing of GHEs. of the EBSS instrumental installation. Based on this
Yavuzturk & Spitler (2001) summarised criteria data, a number of performance characteristics have
for the experimental data sets for GHE model val- been observed. It is suggested that these characteristics
idations. Such observations should be continuously should be taken into account when developing more
collected from the beginning of the system operation comprehensive design methodologies.
and include, at least, measurements of inlet and out-
let circulation fluid temperatures and fluid flow rates
over time. Also, such data sets should provide accu-
rate information about the geometrical parameters of 2 EXPERIMENTAL SET-UP
the GHEs and thermal properties of the ground, grout
and circulating fluid. If obtained, such data sets can The 1,500 m2 two-storey school building (Fig.1) was
be used to extend understanding of thermal processes fitted with a 120 kW GSHP system which provides
during GHE-ground interactions and to validate GHE heating and cooling energy. The heat pumps are cou-
analytical and numerical models. pled with twenty-eight 50 m deep double U-loop bore-
Even though quite a few experimental studies into hole GHEs (Fig. 2). Sixteen of the twenty-eight GHEs
GSHP systems have been published, quality moni- are located under the thermally insulated concrete
toring data sets involving relatively long-term perfor- floor slab of the building and the remaining twelve
mance of GHEs under real-life building loads are rare. GHEs are located outside the building footprint. A
In addition, there have been only a few experimental view of a GHE being installed at the site is shown
studies of the thermal response of the ground around in Figure 3.
GHEs. Recently, Cullin et al. (2015) commented that This paper focuses on the observations of grout
the observational data of the quality that is suitable and ground temperatures around a single GHE located
for the validation of GHE design methodologies is under the building as indicated in Figure 2. Two ground
difficult to obtain. temperature monitoring boreholes were installed to
179
Figure 1. The Elizabeth Blackburn School of Sciences.
Figure 3. A GHE being installed.
Figure 2. Location of GHEs and monitoring boreholes.
record ground temperatures around this GHE at

R = 1.3 m and R = 3.2 m from its centre. A plan
view section of this GHE and nearby temperature
monitoring boreholes is shown in Figure 4.
The vertical location of the temperature sensors
attached to the external wall of the downward (or inlet)
U-loop leg of the GHE and installed in the two mon-
itoring boreholes is shown in Figure 5. The GHE has
eleven temperature sensors installed along its down-
ward U-loop leg. The monitoring boreholes are fitted Figure 4. A plan view section of the GHE under con-
with eleven and ten temperature sensors installed along sideration showing location of temperature sensors (not to
their lengths. In addition to the grout and ground tem- scale).
perature sensors, inlet and outlet in-water temperature
ports as well as a water flow meter were fitted into
the header pipes leading to the GHE to record thermal
loads applied to it. of the ground and grout were measured in the labo-
Continuous core samples were collected from the ratory using a TCi scanner (www.ctherm.com). The
site to study the thermal conductivity of the ground. average thermal conductivity of the ground and grout
The ground at the site from around 1.5 m and down to materials are around 2.7 W/(m · K) and 2.2 W/(m · K)
50 m is Silurian mudstone (Johnston 1992). In addi- respectively. The ground water level measured in the
tion, several samples of the grout were collected during open borehole on the site is about 18 m below the
the grouting of the GHEs. The thermal conductivities ground surface.
180
Figure 7. Farfield ground temperatures at selected dates
over a year to a depth of about 50 m.
Figure 5. A vertical location of the temperature sensors
attached to the outside wall of the downward U-loop leg of
the GHE and installed in two monitoring boreholes. Dots
indicate sensors.
are shown in Figure 7. The ground temperatures at
depths of up to 5 m were subjected to ambient tem-
perature fluctuations whereas the temperatures below
5 m were almost constant through the year. The undis-
turbed ground temperature below about 5 m at the site
is around 19.4◦ C.
3.2 Thermal loads

The thermal loads applied to GHEs were determined
by the building heating and cooling demands. Typical
thermal power applied to the single monitored GHE
in winter (heating) and summer (cooling) is shown in
Figure 8. As seen in the figure, the GSHP system is
scheduled to work from 7 am until 6 pm during week-
Figure 6. Annual ambient air temperatures recorded on site. days and typically switched off for the rest of the time.
The heating power from the GHE normally reached
around 3 kW whereas the applied cooling power to the
GHE was typically a little more.
The cumulative geothermal energy to the GHE from
3 EXPERIMENTAL OBSERVATIONS
the start of the system operation is shown in Figure 9.
As observed, from approximately April to November
3.1 Air and farfield ground temperatures
(including a winter in Melbourne), the system mainly
Figure 6 presents the outside shade air temperatures works in heating. From approximately November to
recorded on the roof of the building during a one year April (including a summer in Melbourne) the system
period starting August 2014. The maximum tempera- predominantly works in cooling.
ture recorded during this period was 38.6◦ C in January, When the system started operation in March, 2014,
2015 and the minimum temperature was 2.4◦ C in July, the cumulative ground energy was zero. By March,
2015. The average air temperature during this period 2015, after a full year of operation, this one GHE
was 15.8◦ C. extracted 1,336 kWh of geothermal energy for heating
The ground temperatures at around 8 m from the and injected 1,732 kWh for cooling. In total, the GHE
installed GHEs (Fig. 2) were monitored to observe the supplied around 3,070 kWh of geothermal energy
values of the temperatures not affected by the installa- for air-conditioning the building. In March, 2015,
tion or the farfield ground temperatures. Some of these the cumulative heating applied to the GHE reached
observations at selected dates to a depth of about 50 m 397 kWh (cooling). This demonstrates that the overall
181
Figure 10. Annual observations of the temperatures of the
outside wall of the downward U-loop leg of the GHE at a)
0.5 m and 30 m and b) 0.5 m and 50 m below the underside
of the building slab.
3.3 Grout and ground thermal reactions to GHE

Figure 8. Typical geothermal power to the GHE during a)
heating in winter and b) cooling in summer. Grout and ground thermal properties, including ther-
mal conductivity, diffusivity and temperatures, deter-
mine the performance of GHEs. Temperatures around
GHEs might substantially change during operation of
a system and can, therefore, affect its performance,
especially in the long-term. It is important to antici-
pate this aspect of system operation and consider it in
the design of GHEs.
In the EBSS facility, grout and ground thermal reac-
tions to GHEs have been being studied in detail. Some
observations on the grout and ground temperatures at
different stages of the system operation are presented
in this section of the paper.
3.3.1 Grout and ground temperatures along and

around GHE
To demonstrate grout thermal response along the GHE,
the observations of three sensors – at 0.5 m, 30 m and
Figure 9. Cumulative ground load applied through the 50 m below the underside of the building slab – were
GHE. selected for presentation and discussion. These sen-
sors are attached to the outside wall of the downward
U-loop leg of the GHE. The annual temperatures
recorded by these sensors are shown in Figure 10.
annual load applied to the ground throughout the first Generally, as shown in the figure, the 0.5 m sen-
year of operation (from March, 2014 to March, 2015) sor recorded lower temperature peaks in heating and
was cooling dominant with a net injection of thermal higher temperature peaks in cooling compared to the
load to the ground of nearly 400 kWh per GHE. 30 m and 50 m sensors. Also, the observed tempera-
The overall maximum cumulative heating energy ture peaks at the depth of 30 m are lower in heating
extracted from the GHE reached around −800 kWh. and higher in cooling than at a depth of 50 m. As
The overall maximum cumulative cooling energy an example, the minimum temperatures in heating
applied to the GHE was slightly more than 600 kWh. at these sensors were observed on 13 August, 2014
182
Figure 12. Minimum and maximum average temperatures
along the length of the GHE for the grout and the ground at
R = 1.3 m and R = 3.2 m radially from the GHE.
Figure 11. Annual observations of the temperatures of the
ground at a) 0.5 m and b) 30 m below the underside of the
building slab at the outside wall of the downward U-loop
leg of the GHE at R = 1.3 m and R = 3.2 m radially from the
GHE.
temperature at the U-loop wall was 30.5◦ C whereas
at 1.3 m and 3.2 m, the maximum temperatures were
21.8◦ C and 21.3◦ C respectively. At the depth of 30 m,
with 12.1◦ C recorded at 0.5 m, 13.1◦ C – at 30 m and the maximum temperature at the U-loop wall was
14.2◦ C – at 50 m. A similar trend is observed in cool- 29.7◦ C whereas at 1.3 m and 3.2 m, the maximum
ing where the maximum temperatures at these sensors temperatures were 20.5◦ C and 19.8◦ C respectively.
were on 11 February, 2015 with 30.5◦ C recorded at A similar trend is observed in heating. Overall, the
0.5 m, 29.7◦ C – at 30 m and 27.8◦ C – at 50 m. ground temperatures follow a similar trend as the grout
These observations demonstrate that the grout is temperatures with higher cooling and lower heating
affected by heat extraction or injection unevenly along peaks recorded at the 0.5 m depth compared to the
the length of the GHE. In particular, the grout thermal 30 m depth at the same radial distances.
disturbances (or its temperature deviations from the Figure 12 presents a vertical profile of the ground
undisturbed ground temperature of 19.4◦ C) in both temperatures at and around the GHE at depths of up
heating and cooling were higher at the top part of the to 50 m from the underside of the building slab for the
GHE and decrease with the depth of the GHE. In addi- same radial locations as considered previously. For this
tion, these thermal disturbances are nonlinear along plot, the temperatures at U-loop walls were considered
the length of the GHE. Such thermal reactions can be at the end of night recovery (just before the GSHP sys-
explained by higher heat exchange rates between the tem switched on at 7 am each day, see Figures 8). This
grout and circulating fluid in the top section of the excludes immediate temperature rises and drops at the
GHE. Such disturbed grout temperature profile deter- U-loop wall due to the temperature of the water cir-
mines the pattern of the temperatures of the ground culating inside the GHE. Also, after a night recovery,
adjacent to the GHE. the temperatures at the sensors attached to the down-
To consider ground thermal disturbance around the ward and upward U-legs at the same depths normally
GHE, Figure 11 shows annual changes in temperatures equalise. Hence, the grout temperatures at these times
at 0.5 m and 30 m below the building floor slab at three show general temperature trends of the grout.
radial distances from the GHE: at the downward U- At different depths, the grout and ground tem-
loop wall, at R = 1.3 m and at R = 3.2 m. These plots peratures reached peaks at different times (see, for
demonstrate that, at both depths, for heating and cool- example, Fig. 11). For Figure 12, for each radial loca-
ing, the thermal disturbance is significantly higher at tion, the times were found when the maximum and
the U-loop wall compared to the thermal disturbance minimum temperatures, averaged along the length of
at 1.3 m and 3.2 m distances. Indeed, over the mon- the GHE, occurred. Grout or ground temperatures have
itoring period, at the depth of 0.5 m, the maximum been plotted at these times to show their most thermally
183
disturbed states at heating and cooling. The dates of
these times are shown in the figure.
From the figure, the first 5 m of the grout and
ground were considerably more thermally influenced
by the GHE, both in heating and cooling. The temper-
ature changes due to energy extractions and injections
are nonlinear along the depth. Since the GHE is located
under the insulated floor slab, there is only little tem-
perature exchange with the surface. Such boundary
conditions might contribute to the observed temper-
ature profiles. In general, the deviation of the grout
and ground temperature from the undisturbed ground
temperature of 19.4◦ C is larger in cooling compared
to heating. This follows from the fact that the annual
cumulative thermal load applied to the ground was
cooling dominant (Fig. 9).
Figure 12 also illustrates the propagation of ther-
mal disturbance through the ground over time. The
maximum average temperature along the depth of the
GHE was recorded on 21 February at the U-loop wall,
on 21 March at 1.3 m and at 5 April at 3.2 m from
the GHE centre. Hence, the maximum thermal dis-
turbance in cooling occurred 28 days later at 1.3 m
and 43 days later at 3.2 m from the GHE compared
to the grout thermal disturbance at the U-loop wall.
Similarly, the minimum values of the average temper-
atures were observed at 15 August, 20 September and Figure 13. Typical grout temperatures at the outside wall of
28 October at the U-loop wall, 1.3 m and 3.2 m from the downward U-loop leg of the GHE over a week of heating
the GHE respectively. at a) 0.5 m and b) 30 m depths.
Two points can be made based on the monitoring
data presented in this section. Firstly, current analyti-
cal models of GHEs assume different grout and ground
thermal reactions to those observed to be operating At these depths, the difference between the disturbed
in the EBSS installation. For example, infinite and and undisturbed temperatures was no more than about
cylindrical line source models ignore both top and 0.5◦ C. Although measurements cover only 1 year of
bottom boundary conditions of a GHE. This results the design period, this small difference may be desir-
in constant ground temperatures along the length of able in an efficient design. In addition, the balancing of
the GHE at any radial distance from it and at any ground cooling with ground heating loads and ground
particular point of time of GHE operation (Marcotte thermal recovery appear to contribute to such a low
et al. 2010). Another popular analytical solution, the thermal disturbance. Further observations and analysis
finite line source model, assumes that the surface is should be undertaken to evaluate the ground thermal
at a constant temperature equal to the undisturbed disturbance trends in the long-term.
ground temperature at the site (Marcotte et al. 2010).
Such an assumption leads to the ground vertical dis-
turbed temperature profile different from the observed 3.3.2 Grout thermal recovery during nights and
in the EBSS installation. A further study should be per- weekends
formed to understand how observed disturbed ground Since the building heating and cooling schedule
temperature patterns might affect GHE performance in includes significant periods when the GSHP system
a long-term. Also, further research is required to eval- is turned off (Fig. 8), the EBSS facility provides an
uate the accuracy of the GHE performance predictions opportunity to observe grout thermal recovery during
made by current GHE analytical models. standby periods of the system. Figures 13 and 14 show
Secondly, even though the GHE delivered a sig- typical temperatures of the grout at the outside wall of
nificant amount of geothermal energy (3,070 kWh the downward U-loop leg of the GHE over a typical
during the first year, see Section 3.2), the resultant week of heating and a typical week of cooling respec-
thermal disturbance of the ground around the GHE is tively. The grout temperatures are presented for three
not significant. Indeed, at R = 3.2 m, the maximum depths, 0.5 m, 30 m and 50 m below the underside of
average ground temperature along the 50-m moni- the building slab. As shown, the temperatures followed
toring borehole was recorded on 5 April (Fig. 12). the building operation schedule, so the grout was being
However, by this time, the ground temperatures just thermally disturbed due to the applied thermal loads at
slightly exceeded the undisturbed ground temperature 7 am to 6 pm during weekdays and recovering during
of 19.4◦ C at depths below 5 m at that radial distance. weeknights and weekends.
184
can influence the design of GHEs, and particularly the
required lengths of GHEs. Hence, anticipated breaks
in building thermal loads, which might help to keep
ground and grout temperatures within targeted oper-
ational values, should be considered in the design of
GHEs.
4 CONCLUSIONS
The paper presents an overview of the first observa-

tions of the Elizabeth Blackburn School of Sciences
(EBSS) full-scale shallow geothermal installation in
Melbourne, Australia. The GSHP system is heav-
ily instrumented to monitor power and energy input
and output, GHE fluid temperatures throughout the
system, ground temperatures around GHEs at differ-
ent depths, undisturbed ground temperatures, building
thermal loads and on-site weather.
The average annual ambient temperature at the site
over the monitoring period was 15.8◦ C. The farfield
ground temperatures over the top 5 m fluctuated fol-
lowing the seasonal fluctuations of the ambient air
temperature. The undisturbed ground temperatures at
Figure 14. Typical grout temperatures at the outside wall of depths of 5 to 50 m were nearly constant at 19.4◦ C.
the downward U-loop leg of the GHE over a week of cooling During the first year of operation, a single GHE
at a) 0.5 m and b) 50 m depths. provided around 3,070 kWh of geothermal energy to
heat and cool the building. Over this time, the building
cooling demand was higher than its heating demand, so
During the week of heating (Fig. 13), the U-loop the annual thermal energy applied to the ground was
wall temperatures decreased each weekday starting cooling dominant with about 400 kWh more energy
from 7 am due to heat extraction from the ground. injected into the ground than extracted for this one
At nights between weekday operations, the tempera- GHE.
tures increased because of the ground thermal recovery The experimental results suggest that the grout and
when the GHSP system was switched off. However, the ground thermal disturbance is nonlinear along the
recovery over weeknights was not complete and the U- length of a GHE in both heating and cooling. Over
loop wall temperatures were lower at 6 pm on Friday the upper 5 m of the GHE, the grout and ground tem-
compared to the beginning of the week, at 7 am on peratures around the GHE were significantly lower
Monday. During the weekend, the system was not in during heating and significantly higher during cooling
operation, so the grout recovery continued. The week- of the building compared to the sections below. In gen-
end recovery was significant and the wall temperatures eral, the grout and ground thermal disturbance around
at 7 am on Monday after the weekend were almost the the GHE decreases with its depth. Existing analyti-
same as at 7 am of the previous Monday. A similar cal models of GHEs do not consider such disturbed
thermal recovery trend can be observed in Figure 14 ground temperature profiles around GHEs. Further
where the U-loop wall temperatures during a week of investigations are required to evaluate whether such
cooling are presented. a disturbed ground temperature pattern can signifi-
Also, Figures 13 and 14 illustrate the same point cantly influence the performance of GHEs, especially
made before that the grout and ground around the in the long term, and should be considered in sizing
GHE were more thermally influenced at the top sec- GHEs.
tions, closer to the slab, and the thermal disturbance When the GSHP system was in a standby mode,
decreases with depth. This can be observed by com- the grout showed significant thermal recovery. Such
paring the grout temperatures at three different depths recovery periods seem to significantly reduce ground
plotted. thermal disturbances. Hence, scheduled pauses in the
The observed trends suggest that, when the GSHP building thermal loads should be anticipated and con-
system is switched off, grout recovery occurs. The sidered in the sizing of GHEs. Potentially, this can
recovery over nights and especially over weekends reduce design lengths of GHEs. Further research
was significant, so it can considerably influence the should be undertaken to investigate whether the exist-
grout temperature around GHEs. Since the grout and ing GHE modelling methodologies accurately simu-
ground temperatures around GHEs determine the per- late ground thermal recovery when GSHP systems are
formance of GSHP systems, grout thermal recovery switched off.
185
ACKNOWLEDGEMENT Validation of vertical ground heat exchanger design
methodologies. Science and Technology for the Built
The authors would like to acknowledge the support Environment, Taylor & Francis, 21(2), 137–149.
provided by the Sustainable Energy Pilot Demon- Johnston IW (1992) Silurian and Lower Devonian engineer-
ing properties. Engineering Geology of Melbourne, Peck,
stration (SEPD) Program funded by the Depart-
W.A. et al (Eds), Balkema, Rotterdam.
ment of Economic Development, Jobs, Transport and Marcotte D, Pasquier P, Sheriff F and Bernier M (2010)
Resources of the Government of Victoria. The importance of axial effects for borehole design
of geothermal heat-pump systems. Renewable Energy,
35(4), 763–770.
REFERENCES Yavuzturk C and Spitler JD (2001) Field Validation of a
Short Time Step Model for Vertical Ground-Loop Heat
Cullin JR, Spitler JD, Montagud C, Ruiz-Calvo F, Rees Exchangers. ASHRAE Transactions, 107(1), 1–9.
SJ, Naicker SS, Koneèný P and Southard LE (2015)
186
Effect of moisture migration on the thermal conductivity of a

geothermal well seal
C. Walton-Macaulay & J.C. Evans

Department of Civil and Environmental Engineering, Bucknell University, Lewisburg, PA, USA
L. Daher
Geotechnical Specialist, Gannett Fleming, Camp Hill, PA, USA
ABSTRACT: Bentonite is typically used as the seal between the fluid circulating pipes and the adjacent
soil/rock formation for geothermal well systems. Over the design life of a geothermal well, the seal is subjected
to many cycles of heating and cooling. After many cycles of heating and cooling throughout the design life of
the geothermal system, the bentonite is expected to still function as a competent seal, by maintaining contact
with the fluid circulating pipes and the adjacent soil/rock formation surrounding the circulating pipes. Moisture
migration can occur from the adjacent soil/rock formation to the seal or from the seal to the adjacent soil/rock
formation. If moisture migration out of the bentonite seal is excessive, defects in the seal such as separation of
the seal from either the soil formation or the circulating pipes can occur. Such defects could therefore potentially
affect the thermal conductivity. This research presents the results of a study of the effect of moisture migration
on thermal conductivity. A model well, seal and sand formation were created in the laboratory and the system
was subjected to 30 cycles of heating and cooling. The average thermal conductivity for the heating cycles was
0.69 W/k.m and that for the cooling cycles was 0.49 W/k.m. In previous tests using a closed system without
allowing moisture migration, the average thermal conductivity for the heating cycles was at 0.79 W/k.m, and
for the cooling cycles was 0.19 W/k.m. The study results are encouraging in that after thirty cycles of heating
and cooling with the bentonite seal in an open system free to gain or lose moisture, no degradation in thermal
conductivity was observed. However, moisture content analysis of bentonite seal samples from the model showed
moisture migration had indeed occurred illustrating the need for a long-term field study of the performance of
bentonite geothermal well seals.
1 INTRODUCTION 1.1 How does the geothermal system work?

The circulating pipes (typically HDPE plastic pipes)
The use of geothermal (ground source) heat pumps
placed in the borehole are connected to a heat pump
in both residential and commercial construction is
unit above ground.A seal material (typically bentonite)
increasing. Geothermal heat pumps are installed in the
used around the pipes allows for efficient heat transfer
earth with heating/cooling loops that circulate fluids
between the pipes and the earth. In an active heat-
over the system design life. This includes one or more
ing system, a fluid (typically a mixture of water and
pre-drilled boreholes in which the circulating pipes are
antifreeze) is circulated in the plastic pipes (loops)
placed, and the surrounding space around the pipes
underground to allow heat transfer from the earth to
filled with a thermal conductive material.
the fluid. The fluid absorbs the heat and with cir-
Geothermal heat pumps have the potential to recy-
culation, the heat is returned to the heat pump. The
cle energy from the earth and reduce primary energy
heat pump extracts the heat, and re-circulates the
consumption and emissions of greenhouse gases. The
cooler fluid for another cycle of heat transfer. The heat
U.S. Environmental Protection Agency estimated that
pump distributes the extracted heat as warm air. In an
geothermal heat pumps could reduce energy con-
active cooling system, the roles are reversed and the
sumption by as much as 44% when compared to
heat pump extracts heat from the surrounding hot air
air-source heat pumps, and reductions in energy con-
and distributes the cool air as air-conditioning. The
sumption can be as high as 72% when compared to
extracted heat is transferred into the pipe loops with
conventional electrical heating and air conditioning
the fluids and circulated underground to allow heat
(USGAO, 1994).
transfer from the fluids to the earth.
187
Figure 1. Section of geothermal borehole with gap forma-
tions (Allan and Philappacopoulus, 1998).
1.2 Why a bentonite seal?

Bentonite is typically used as the grouting material to
seal the geothermal well. It is placed in slurry form
consisting of bentonite (sodium montmorillonite) a
viscosifying polymer, and water with a solids con-
centration of not less than 20%. Bentonite grout is
employed due to its 1) good thermal conductivity,
which is critical for heat transfer, 2) its low perme-
Figure 2. Side view of geothermal well model.
ability which is essential to protect the groundwater,
and 3) its relatively low viscosity allows for its place-
ment with no voids, which is critical for efficient heat
transfer, and groundwater protection. performance requirements necessary for an efficient
However, the bentonite grout slurry has a very high function of the well system. An efficiently functioning
water content and there is a risk of free water migrat- system includes the control of ground water migration
ing within the slurry, and between the slurry and vertically between formation strata and the conduc-
the surrounding formations. It has be postulated that tion of heat horizontally through the grout from the
excessive moisture migration could cause shrinking circulating pipes or from the surrounding formation
and cracking, or desiccation of the well seal, which material.
would compromise its integrity (Evans & Sicwebu, BENSEAL® and EZ-MUD® , used in a closed sys-
2014). A compromised seal manifests itself as the for- tem study by Evans & Sicwebu, (2014), was also
mation of gaps between the grout and piping, and/or used for this experimental study. BENSEAL® and EZ-
the grout and surrounding soil interfaces, and result in MUD® are manufactured/sold by Baroid Industrial
a loss of efficiency in thermal conductivity. Figure 1 Drilling Products. BENSEAL® is a bluish/gray granu-
shows the types of gap formation that compromises lar natural Wyoming sodium bentonite. EZ-MUD® is
the grout seal. a liquid polymer emulsion and is used to stabilize the
BENSEAL® from excessive swelling and sloughing
during the slurry preparation. The EZ-MUD® makes
1.3 Research objective the slurry more viscous at a given bentonite content,
thereby increasing the workability of the grout.
The research reported in this paper extends the closed Figure 2 shows a photograph of the labora-
system studies of the thermal and hydraulic properties tory geothermal well model used to contain the
of geothermal well seals (Evans & Sicwebu 2014). BENSEAL® /EZ-MUD® grout. The apparatus (cell)
Using an open system, that allows for moisture migra- used for grout placement was constructed of a 152 mm
tion (not used in Evans & Sicwebu, 2014), this research inside diameter polyvinyl chloride (PVC) pipe, which
studies the thermal conductivity over numerous cycles is a similar radial dimension to that of an actual
of heating and cooling and how moisture migrates geothermal system. The height of the apparatus is
within the bentonite seal and between the seal and 30 cm, representing only a portion of the vertical
formation. height of a typical geothermal system. The cell had
an inner High Density Polyethylene (HDPE) pipe of
25 mm diameter to circulate the heating and cooling
2 MATERIALS AND METHODS fluid. The piping material used is a material that is used
in field geothermal well systems.
To study the effects of moisture migration on thermal In a field installed geothermal well system, mois-
conductivity on the geothermal well bentonite grout ture migration can occur into or from the formation.
seals, the grout material was selected based on the Moisture migration depends on the water retention
188
Figure 3. Cross-sectional view of the open system geother-
mal well model.
Figure 4. Open system geothermal cell test setup model.
capacities of the bentonite slurry and of the forma-

tion. Though surface water tensions may effect water
retention capacities, it is the mechanisms related to
the materials physico-chemical interactions at micro-
scopic levels, in particular, the transfer of interlayer
water to the macropores (that can be triggered by tem-
perature increases) that has the greater effect on its
water-retention (Villar and Gomez-Espina, 2008). In
this experimental system, the cell piping material was
pre-drilled with holes of 3.175 mm diameter spaced
30.48 cm on center vertically and horizontally to allow
for moisture migration between the bentonite slurry
and the formation. Figure 3 shows a schematic dia-
gram of a slice through the geothermal well model. A
non-woven geotextile fabric was placed on the inside
of the outer cell wall, and then the cell was placed in a
larger system of approximately 50 cm in diameter and
40 cm deep. Fine sand was placed within the larger sys-
tem and around the cell, which acted as the formation
soil around the geothermal system. The fine sand is
well graded with a maximum particle size of 4.75 mm
with just over two percent larger than the 0.075 mm
size.
To maintain moisture equilibrium within the sand
formation, bowls of water were placed on top of the
sand and the entire surface of the larger system was
covered with a plastic cling wrap. This approach effec-
tively maintained 100% humidity in the air above the
top of the sand. The geothermal system setup is shown Figure 5. Thermocouple placement in the geothermal sys-
in Figure 4. tem model.
To measure the thermal conductivity of the ben-
tonite slurry seal, an instrumentation setup identical
to that of Evans & Sicwebu, (2014) was used, and The geothermal test system included a flow sys-
therefore only a summary is presented here. A total of tem as shown in Figure 6. The flow system included a
sixteen thermocouples were used in this study. Four- peristaltic pump for a continuous flow through the sys-
teen thermocouples were placed within the annulus of tem’s continuous loop for heating and cooling fluids.
the cell as shown in Figure 5, and used to calculate The flow rate of the pump was 0.1 mL/hr.
the system’s thermal conductivity using Fourier’s law Based on Fourier’s law and the geothermal system
for heat flux density in soils. Two thermocouples were test setup, the thermal conductivity, kt of the grout
used to measure the inflow and outflow temperatures slurry is determined from the heat flux q (Equation 1)
of the circulating fluids Knowing the temperature dif- and by the formulated Equation 2:
ference and flow rate allowed for a calculation of heat
loss/gain as the fluid circulated.
189
Figure 6. Schematic of thermal and hydraulic conductivity
testing apparatus (Evans & Sicwebu, 2014).
In Equation 1 and 2, ṁ is the mass flow rate, cp is Figure 7. Schematic of bentonite slurry sampling locations
the specific heat capacity of water (1 g/◦ C), T is the water content determination.
change in temperature of the grout, R1 is the inside
radius of the cylindrical cell, R2 is the outer radius,
and x is the height of the geothermal model.
Table 1. Final moisture content results of 18 bentonite
sample after being subjected to heating and cooling cycles.
3 RESULTS Relative Relative

Water water Water water
3.1 Moisture Migration Sample content content Sample content content
# (%) (%) # (%) (%)
A series of gravimetric water content tests were per-
formed at the end of the heating and cooling cycles in 1 327 −34.6 10 441 −11.8
the open system geothermal well model, to determine 2 260 −47.9 11 490 −2.1
the moisture content migration trends within the grout 3 281 −43.9 12 451 −9.8
mixture. Figure 7 presents a visual aid of the location 4 352 −29.6 13 498 −0.5
of grout samples used to determine the moisture con- 5 384 −23.3 14 527 +5.5
tents. A total of eighteen water content samples were 6 430 −14.0 15 567 +13.4
7 331 −33.8 16 336 −32.8
chosen at the locations shown in Figure 7 to establish a 8 465 −7.0 17 569 13.7
diametric and vertical patterns along which moisture 9 469 −6.2 18 550 10.0
migration, if any, can be determined. The diametric
pattern consists of sampling from three depths with
each depth having six diametric sampling locations.
For each depth, the samples were obtained in series
starting from an outside edge of the specimen going
towards the center core (e.g. 1, 2, 3) and continued percent (Sample #1) to the maximum of 569 percent
from the other side of the center core towards the dia- (Sample #17).
metrically opposing outer edge (e.g. 4, 5, 6). Similarly
the vertical patterns consist of six vertical lines of sam-
pling, with each line of sampling having three samples 3.1.1 Lateral migrations
at varying depths (e.g. 1, 7, 13). To determine the relative water contents shown in
The gravimetric water content data of all the eigh- Table 1, the difference in the measured water content
teen samples obtained from the bentonite grout are from the initial water content of 500 percent of the ben-
shown in Table 1. The mean water content for the grout tonite grout at the start of testing were compared to the
samples is 429 percent. There appears to be a large initial water content. Given the following variables:
variation in the water content data. The table shows w (measured water content); wini (initial water con-
the range of water content from a minimum of 327 tent), the relative water content, Rw is determined by
190
Figure 8. Water content test results with respect to distance
from center of the geothermal system. Figure 9. Water content test results for a closed geothermal
system. (Evans & Sicwebu, 2014).
Equation 3. Rw can also be expressed as a percentage

by multiplying Rw by 100. shown from one end to the other (diametrically), it is
noted that moisture loss (relative to the initial water
content of 500 percent) occurred at all test locations
for the top and center portions of the grout (Samples
1–6, 7–12), but with a higher rate of loss at the top
The relative water content data in Table 1 is plot- portion. However, moisture increases occurred at most
ted in Figure 8 with respect to the distance sampled test locations for the bottom portion of the grout (Sam-
from the center of the test setup. This provides a ples 13–18). The changes in moisture content at each
visual representation of relative water contents diamet- of these top, center and bottom sections of the grout
rically across the bentonite grout. Samples 1 through 6 indicates a trend of high moisture loss at the top, mod-
were taken at approximately the same elevation across erate moisture loss at the center and moisture increase
the grout and within the top section, with Samples 7 at the bottom, which indicates a moisture migration
through 12 taken within the center section, and Sam- with depth.
ples 13 through 18 within the bottom section. An Evans & Sicwebu, (2014) performed a series of
anomaly of relative water content reduction is noted gravimetric water content tests using a closed sys-
for Samples 13 through 18 that were sampled dia- tem, whereby net moisture migration from the grout
metrically within the top section of the grout. The mixture was not possible. A closed polyvinyl chloride
noted drop from Samples 15 to 16, then the immediate (PVC) pipe was used to contain the bentonite seal. Fig-
increase in Sample 17, does not appear to be consis- ure 9 shows the relative water contents of samples from
tent with the general trends noted. In general, there the closed system at distances from the center. Though
appears to be a slight decrease in water content dia- net moisture migration was prohibited out of the grout
metrically across the grout (from right to left) at each mixture, moisture migration redistribution within the
sampling elevation. The average relative water content grout mixture occurred. Figure 9 indicates a general
at the top, center, and bottom sampling elevations are consistent moisture contents radially but with loss of
−32.2, −11.8, and 8.4 percent respectively. A negative moisture content only at the top section of the grout
relative water content implies a loss of moisture from mixture, and very close to the center (Samples 1–6).
the initial condition, and a positive change implies a Due to the increased variation of relative water con-
gain of moisture from the initial condition. The aver- tents in the open system geothermal well model used in
age diametric relative water content for the grout is this study, than that of the closed system used by Evans
11.9 percent. & Sicwebu, (2014), it appears that the open system did
Decreases in water content diametrically across the allow for moisture migration.
grout may be due to the idea that moisture can migrate
from areas with high temperature towards cooler areas 3.1.2 Vertical migrations
(Chen, 1988). This process of moisture migration The water content data from Table 1 for the open
due to thermal gradients is called thermo-osmosis. system is also shown in Figure 10, but with respect
Thermal gradients can occur from non-heterogeneity to sampling locations relative to depth. This aids as
within the grout such as variations in consolidation. a visual representation of water contents vertically
Differences in the water contents across the grout through the bentonite grout. For example, Samples 1,
system may also be due to the effects of boundary 7, 3 represent a vertical layout of samples at differ-
conditions created by the off-centered placement of ent depths that were obtained for water content tests,
the grout system within the surrounding soil (see Fig- and were the furthest on the left side from the cen-
ure 4). Though a relative change in water content is terline of the grout system (see Figure 7 for complete
191
Figure 11. Water content test results for a closed geothermal
Figure 10. Water content test results with depth. system. (Evans & Sicwebu, 2014).
layout). Showing the water content for Sample 16 as

an anomaly, as previously discussed in the diametric
distribution of water contents, Figure 10 shows in gen-
eral, increases in the relative water content with depth.
Relative water contents were much lower at the top
(approximate depth of 7.5 cm), slightly higher at the
center section (approximate depth of 15 cm), and about
or above the initial water content at the bottom section
(approximate depth of 22.5 cm).
The average vertical relative percent change in
water content with depth were determined at each ver-
tical location. The average relative water content with
depth for all vertical locations vary from −23.0 per-
cent (Samples 1, 7, 13) to 0.1 percent (Samples 6, 12,
18), with an average of −12.7 percent for all vertical
locations. The increases of relative water content of
the bentonite grout with depth are most likely due to
gravitational migration.
Similarly, the relative water content data from the Figure 12. Water content test results of the natural sand soil
closed system by Evans & Sicwebu, (2014) were ana- formation around the geothermal well.
lyzed with respect to depth at each vertical layout and
graphically shown in Figure 11. The figure shows very
minimal variation of water content with depth from the
3.2 Steady state thermal conductivity
initial water grout slurry water content of 500 percent
for the closed system. The exception is noted in only A steady state thermal conductivity value was com-
one of the vertical layouts (Samples 4, 10, 16), where puted for each heating and each cooling cycles from
the relative water content vary by as much as −27 the data obtained from the thermocouples. The com-
percent (Sample 4) and only at the top section of the puted steady state thermal conductivity values for 30
grout, and closest to the center of the grout with the cycles of heating and cooling cycles are presented in
pipes circulating heating and cooling fluids. Figure 13 as a function of cycle number. The first
To assess moisture migration through the grout noted observation from Figure 13 is that the thermal
slurry, into and from the formation, a series of water conductivity for cycles of heating is higher (average
content were performed on the formation material value 0.688 W/m-K) than that of the cycles of cooling
(natural fine sand) around the geothermal well annu- (average of 0.491 W/m-K). These findings are consis-
lus and presented in Figure 12. Due to the apparent tent with Evans & Sicwebu, (2014) for higher thermal
relative increases in some of the water contents at the conductivity values for the heating cycles than those of
top section (e.g. Samples 2, 8, 14) above that of the the cooling cycles. Thermal conductivity changes with
center section, it appears that lateral moisture migra- temperature because heat conduction in non-metallic
tion may have occurred. However, the apparent relative solids is mainly due to lattice vibrations (phonons)
increases in water content at depths, indicate that mois- (Evans & Sicwebu, 2014). As atoms vibrate more ener-
ture migration with depth is more pronounced than getically at one part of a solid due to high temperatures,
lateral moisture migration. heat is transferred to less energetic neighboring atoms.
192
Figure 14. Thermal conductivity for cooling cycles of open
Figure 13. Thermal conductivity test results for open sys- and closed systems.
tem.
Due to the material temperature increase, the internal

particle velocity increases as well, thereby increas-
ing the thermal conductivity. The increased velocity
transfers heat with less resistance (Ziman, 1967).
The grout manufacturer specifies the thermal con-
ductivity of the grout to be 0.74 Watts/m. ◦ C. This
specified value of 0.74 Watts/m. ◦ C is similar to the
average value for cycles of heating of 0.688 Watts/m.
◦
C. Figure 13 shows that there was no significant
degradation in either heating or cooling thermal con-
ductivity values for 30 cycles. Furthermore, the ther-
mal conductivity for the 30 cycles for both heating Figure 15. Depiction of moisture migration in the grout
and cooling, appear consistent throughout the cycles seal.
with some anomalies. A noted anomaly was of two
separate reductions in thermal conductivity during the
cooling cycles, but with subsequent increases in the
cycles immediately following. Another anomaly was migration in and out of the system as moisture can sub-
an increase in the thermal conductivity for the heating stantially affect thermal conductivity (Salomone et al.
cycles closer to the end of the cycles, but also with sub- 1984; Salomone and Kovacs 1984; Salomone and Mar-
sequent decrease of thermal conductivity thereafter, lowe 1989). A bentonite grout condition of moisture
that approaches the more consistent values. migration was a limitation in the study by Evans &
Though there were no change in thermal conductiv- Sicwebu, (2014). The testing apparatus in this study
ity within the 30 cycles for the open system, a change allowed for moisture migration within and between
in the constant thermal conductivity between the open the grout and soil formations. Moisture migration
and closed systems for the cooling cycle occurred as occurred laterally as a result of a drying thermal gra-
shown in Figure 14. No change was noted in the ther- dient, and vertically (downwards) within the grout
mal conductivity between the open and closed systems seal as a result of a wetting gravity gradient. The
for the heating cycles. lateral migration of moisture indicates that moisture
migration between the soil formation and the grout
seal occurred. Figure 15 shows the direction of mois-
ture migration and also shows the results of moisture
4 CONCLUSIONS AND LIMITATIONS migration as a negative change at the top of the grout
and a positive change at the bottom for a net negative
The integrity and performance of a geothermal well change in the relative water content of the grout seal.
system is dependent on several components of the Though moisture migration occurred, moisture
system that includes the formation soil and bentonite migration did not have an effect on the thermal
grout conditions, and the thermal conductivity of the conductivity of the grout seal, as the seal maintained
bentonite grout seal.An improved understanding of the its initial thermal conductivity of heating and that of
impact of cyclic heating and cooling would include the cooling in 30 cycles. However, a change in the thermal
study of an open system that would permit moisture conductivity between the open and closed systems for
193
the cooling cycle occurred. These results illustrate the Armitage, D.M., Bacon, D.J., Massey-Norton, J.T., and Miller
need for a long-term field study of the performance of J.D. (1980). Ground-water Heat Pumps: an Examination
bentonite geothermal well seals. of Hydrogeologic, Environmental, Legal, and Economic
Factors Affecting Their Use. U.S. Department of Energy.
Washington, D.C. DOE/CS/20060-5120(V.1).
Chen, F.H. (1988). Foundations on Expansive Soils, Devel-
opment in Geotechnical Engineering, Vol. 54. Elsevier
ACKNOWLEDGEMENTS Science Publishing Company, New York.
Evans, J. and Sicwebu, A. (2014). Geothermal Well Seals
The authors acknowledges the financial support Subjected to Cyclic Heating and Cooling. New Frontiers
provided by the Bucknell University Program for in Geotechnical Engineering: pp. 88–97.
Undergraduate Research, the Chiloro Fund of the Salomone, L. and Kovacs, W. (1984). Thermal Resistivity of
Department of Civil and Environmental Engineer- Soils. Journal of Geotechnical Engineering. 110(3), pp.
ing, and the Michael and Laureen Costa endowment 375–389
of the Jeffrey C. Evans Geotechnical Engineering Salomone, L. and Kovacs, W, and Kusuda, T. (1984). Ther-
Laboratory. The idea of moisture migrations in mal Performance of Fined-Grained Soils. Journal of
Geotechnical Engineering. 110(3), pg 359–374.
geothermal systems were in part based on dis- Salomone, L. and Marlowe, J. (1989). Soil and Rock Clas-
cussions with Mr. William Reichart, President of sification According to Thermal Conductivity: Design of
the Pennsylvania Ground Water Association (www. Ground-Coupled Heat Pump Systems. Report to Electric
pwgwa.org) and President of William W. Reichart, Inc. Power Research Inistitute, Report No. EPRI CU-6482.
a geothermal system installation company. Chantilly, VA.
USGAO, (1994). Geothermal energy: outlook limited for
some uses but promising for geothermal heat pumps, U.S.
General Accounting Office RECD-94-84
REFERENCES Villar, M.V., Gomez-Espina, R. (2008). Effect of temperature
on the water retention capacity of FEBEX and MX-80 ben-
Allan, M.L., Philippacopoulos, A.J. (1998). Thermally Con- tonites, Unsaturated Soils: Advances in Geo-Engineering,
ductive Cementitious Grouts for Geothermal Heat Pumps. London, UK
FY 1998 Progress Report, BNL 66103, Brookhaven Ziman, J. (1967). The thermal properties of materials.
National Laboratory In Materials (pp. 111–126). WH Freeman
194
Heat content in soil-borehole thermal energy systems in the vadose zone
T. Başer
Department of Structural Engineering, University of California San Diego, Gilman Dr. La Jolla, CA, USA
Y. Dong
Colorado School of Mines, Golden, CO, USA
J.S. McCartney
Department of Structural Engineering, University of California San Diego, Gilman Dr. La Jolla, CA, USA
ABSTRACT: This study focuses on understanding the heat content within soil-borehole thermal energy stor-
age (SBTES) systems installed in different types of soils in the vadose zone. Temperature fluctuations in the
atmosphere can create a variety of heat flux conditions resulting in different temperature gradients in the subsur-
face, even when a surficial insulation layer is incorporated. A three-dimensional (3D), transient finite element
model was built in COMSOL to consider the representative field conditions as well as coupled heat transfer and
water flow processes in the unsaturated soil within the SBTES system. The heat content is used to quantify the
heat gain above the temperature profile expected above the ambient ground temperature fluctuations. The heat
content changes with different type of soils as the hydraulic and thermal properties are specific to soil types.
Results indicate that presence of an insulation layer leads to a significant heat gain in the shallow subsurface in
all types of soils.
1 INTRODUCTION temperature distributions in heat exchanger arrays

installed in different types of soils in unsaturated con-
Soil-Borehole Thermal Energy Storage (SBTES) sys- ditions, considering coupled heat flow and thermally
tems are used to store heat collected from renewable induced water flow. Thus a series of numerical analy-
sources so that it can be used later for heating of build- ses were performed on unsaturated silt and clay soils to
ings (Sibbitt et al. 2012; Zhang et al. 2012, McCartney evaluate the role of the surficial insulation layer along
et al. 2013, Başer & McCartney 2015). They func- with the atmospheric boundary conditions. These are
tion in a similar way to conventional geothermal heat compared with preliminary data from a field SBTES
exchange (GHE) systems, where heat is transferred site in San Diego, CA.
from a source to a sink via circulation of fluid through
a series of closed-loop heat exchangers. However, they
differ from GHE systems in that the heat is injected or 2 BACKGROUND AND FIELD STUDY
extracted continuously over the course of a season into
the borehole heat exchanger array. Further, the bore- Several field and numerical studies have established
hole array in a SBTES system is overlain by a hydraulic that SBTES are proven to be efficient at storing heat
barrier to retain pore water within the subsurface and in the subsurface (Sibbitt et al. 2012, Zhang et al. 2012,
a thermal insulation layer to minimize heat losses to Başer et al. 2015b). However, a better understanding of
the atmosphere (Başer et al. 2015a, 2015b, 2016). the heat transfer processes in these systems is required
In the shallow subsurface, below the soil- as the temperature increase in the SBTES arrays is
atmosphere interface, heat transport plays a critical highly dependent on the thermal properties of the soils
role in determining the energy fluxes between the soil and these properties change with the type of soil and
surface and the atmosphere. Atmospheric conditions the degree of saturation of the soil. To prevent heat loss
may be very complex due to the climatic changes and from the upper surface of an SBTES system, layers of
can create a variety of heat flux conditions at the soil expanded polystyrene (EPS) are placed atop the array
surface. In arid and semiarid regions, temperature gra- beneath a vegetative soil layer. There have only been a
dients in the shallow subsurface can be very large and few studies justifying the role of the insulation layer.
may have a significant effect on temporal temperature Başer et al. (2016) performed a numerical study on this
distributions. topic, but only considered 2D flow processes and did
The main goal of this paper is to understand the not evaluate the change in temperature above ambient
impact of the insulation layer on spatial and temporal ground temperature fluctuations.
195
Figure 1. UCSD SBTES system: (a) Plan; (b) Elevation.
Recently a full-scale SBTES system was installed

at the Englekirk Research Center on the University
of California San Diego Campus. The SBTES system
consists of a vertical borehole array and two horizontal
arrays. Borehole array includes 15 m deep 15 borehole
heat exchangers in a hexagonal array with a spacing
of 1.5 m, as shown in Figure 1. The heat exchangers
consist of high density polyethylene tubing with a “U”-
shape coupling at the base. Three additional boreholes
were installed that included thermistor strings, which
have six thermistors along the length of a single cable.
Their purpose is to measure the temperature distribu-
tion with depth to infer heat transfer processes within Figure 2. Initial temperatures of UCSD site with and with-
the array. One of these thermistor strings was installed out insulation layer (a) Time series for different depths; (b)
in an isolated borehole to observe undisturbed ground Temperature profiles on February 15, 2016; (c) Ambient air
temperature fluctuations during operation, which is a temperature.
useful variable to assess the heat storage in an SBTES
array.
After drilling the boreholes, the soil excavated to ambient air temperature fluctuations on the tempera-
a depth of 1 m around the array. After placing a layer ture penetration into the soil under the array and in
of site soil to cover the heat exchangers, a 30 mil- the undisturbed ground. The ground temperature data
thick hydraulic barrier was placed on the soil surface. collected inside of the array where the array is covered
A layer of expanded polystyrene (EPS) insulation was with an insulation layer and outside of the array from
then placed on top of the hydraulic barrier, and the site September 13, 2015 to February 15, 2016 are plotted
soil was then backfilled up to grade. Before operation in Figure 2. Placing an insulation layer decreases the
of the system, the initial ground temperature fluctua- effect of the ambient temperature fluctuation on the
tions are being collected to determine the effect of the soil temperature.
196
3 NUMERICAL MODEL
3.1 Model formulation

A transient three-dimensional finite element model
was built in COMSOL to predict temperature distribu-
tions inside and outside of the thermal heat exchanger
arrays. The model was developed considering both
heat transfer and water flow since when the unsatu-
rated soils is heated water flows due to the decrease in
density. The other driving mechanism for water flow
is the alteration of surface tension with temperature.
Thermally-induced vapor flow is not considered in this
study.
Water flow in unsaturated soils can be expressed
by Richards’ equation assuming that air pressure in
the pores equal to atmospheric pressure. Thus mass
balance in unsaturated soils is expressed as follows
(Bear 1972):
Figure 3. Model geometry.
For heat transfer, a constant heat flux of 30 W/m was

where n = porosity of the soil; Sw = wetting (water) applied at the borehole boundaries for a period of 90
degree of saturation (dim.); ρw = density of water days. A sinusoidal temperature function was applied at
(kg/m3 ); µw = dynamic viscosity of water (Pa · s); the top assuming that maximum and minimum daily air
Pc = capillary pressure (Pc = Pnw − Pw ) (kPa); t = time temperatures are 25◦ C and 10◦ C. The bottom temper-
(s); kint = intrinsic permeability of soil (m2 ); krw = ature was fixed to 12◦ C because of the groundwater.
relative permeability of water (dim.); g = gravitational For water flow, zero flux was assumed for all bound-
acceleration (m2 /s); and Qm = mass source (kg/(m3 s)). aries except the bottom boundary, where a constant
Since phase change between liquid water and water total head of 14 m was applied.
vapor was neglected in this study, Qm was assumed to The initial temperature of the domain was assumed
be to zero. to be uniform and equal to 12◦ C. This is equal to the
Heat transfer in unsaturated soils is governed by mean annual air temperature in San Diego, CA, and
the combined Fourier’s and Newton’s law. The govern- represents the transition profile between the hot and
ing equation for heat transfer is porous media can be cold seasons of the year. The water table was assumed
expressed as follows: to be at a depth of 16 m, coinciding with the bottom
of the heat exchangers. The initial conditions for the
profiles of degree of saturation and suction with depth
correspond to hydrostatic conditions.
The entire domain is assumed to be a uniform and
where ρ = total density of the soil (kg/m3 ); Cp = isotropic soil layer. The properties of two soil types
specific heat capacity of the soil at constant pressure were considered for the soil layer, those of Hopi silt
(J/(kgK)); uw = Darcy velocity; T = absolute tempera- and Denver claystone. The thermal and hydraulic prop-
ture (K); λ = apparent thermal conductivity of the soil erties needed for coupled heat transfer and water flow
(W/(mK)); and Q = heat source (W/m3 ). analyses of these soil are given in Figures 4(a) to 4(d)
(Lu and Dong 2015).These properties differ for the dif-
ferent type of soils as they are dependent on the grain
size, pore size distribution, and degree of saturation
3.2 Model geometry and boundary conditions
of the soil. Saturated thermal conductivity of Hopi silt
The model geometry consists of an array of 15 m-deep and Denver claystone are 5.3 × 10−7 and 2.2 × 10−7
vertical borehole geothermal heat exchangers installed (m/s), respectively.
in a deep homogeneous soil layer in an array having a After the implementation of the initial and boundary
width of 30 m and a depth of 30 m. A 0.1 m-thick insu- conditions, the system of partial differential equations
lation layer was placed on top of the heat exchangers, (1) and (2) in three-dimensional domain was simul-
which was covered by a layer of site soil. The details taneously solved using the COMSOL Multiphysics
of the model geometry is given in Figure 3. Symme- software that is based on the finite element method.
try was used in configuring the model geometry, with The simulated domain has a volume about 10 times
one of the geothermal heat exchangers at the corner of that of the heat exchanger domain in order to minimize
the array, and the other two are spaced at a distance of boundary effects. Also initial simulations verified that
2.5 m from the center in orthogonal directions. there was no boundary effects.
197
determined. In heat transfer modeling efforts surface
temperature boundary conditions are important as the
ambient air temperatures have an effect up to depth of
10 m from the surface (Brandl 2006). The first analysis
was performed to observe the penetration of surface
temperature without any heat input into the soil layer.
Then, heat input is initiated until the last time step
reaches 90 days for two cases; without the insulation
layer and with the insulation layer. Temperature pro-
files inside of the array for Hopi silt were plotted for
different time steps at x = 1.25 m, y = 1.25 m, and are
given in Figures 5(a), 5(b), and 5(c).
Ambient temperature fluctuations has an effect on
the soil temperature distribution in the subsurface.
However, the amplitudes decrease depth with due to
the thermal inertia of the soil (Brandl 2006). The pene-
tration depth from the field data is 7 m while it is 11 m
from the numerical results. This is mainly because
of the different thermal properties of the soils. The
baseline temperature distribution trend from numeri-
cal analysis is compatible with those of measured in the
field as shown in Figure 5a. Also insulation layer in the
numerical analysis led to a relatively lower temperature
values inside of the array.
A maximum temperature of 43◦ C without the insu-
lation layer is observed at 7.5 m at the end of the
heating while the temperature reaches a value of 44◦ C
at a depth of 6 m in the analysis with the insulation
as shown in Figures 5b and 5c. Yesiller et al. (2005)
defined a new parameter for evaluating exothermic
reactions in municipal solid waste landfills called
the “heat content”. This parameter can be used to
account for the amount of heat in the landfill above
that expected for seasonal ground temperature fluctu-
ations at a given depth. In a similar way heat content,
HC (◦ C × day/day) of the heat exchanger array was
determined by first calculating the area between the
time series curves for the temperature increase by
the constant heat input and the baseline temperatures.
Then it was divided by the duration of the analysis
period to normalize HC with respect to time. HC can
be expressed by the following equation:
To calculate the area, first the temperature time series

at different locations are plotted in Figures 6(a) to 6(c).
The maximum temperature inside of the array where
is very close to the center borehole was 55◦ C while
it decreased a value of 23◦ C outside of the array. The
insulation layer had a negligible effect on temperature
Figure 4. Hydraulic and thermal properties of soils used in at a depth of 6 m. This effect is the same inside and
the analyses (a) SWRC; (b) HCF; (c) TCF; (d) Cp vs VWC. outside of the array.
Heat contents for two different depths was calcu-
lated and plotted at the different locations (inside and
4 ANALYSIS outside of the array) for Hopi silt as shown in Figures
7(a) and 7(b). It was observed that the insulation layer
Spatial and temporal distributions of the temperatures has a greater effect on the heat content of the soil closer
inside and outside of the heat exchanger array were to the surface.
198
Figure 6. Temperature time series at a depth of z = 6.0 m:
Figure 5. Temperature profiles inside of the heat exchanger (a) 0.5 m from the center; (b) 1.25 m from the center; (c) 3.5 m
array for Hopi silt (a) Baseline; (b) Heat input without from the center.
insulation; (c) Heat input with insulation.
Denver claystone are plotted in Figure 8(a) at 1.5 m.
The heat content must be dependent of soil type as The maximum heat content value of 32 (◦ C × day/day)
the thermal and hydraulic properties change with dif- was observed at 1.5 m. This value is 31% greater than
ferent soils. To understand the effect of soil type on the the value of 24.4 which was observed for Hopi silt in
heat content a series of analyses were performed for Figure 8(b).
Denver claystone applying the same boundary con- The contribution of the insulation layer on heat con-
ditions. Although not given here, the temperature in tent was also an interesting subject to investigate. Thus,
the middle of the array reached a temperature value heat contents were determined for the temperature
of 51.7◦ C at the depth of 6.2 m due to the relatively gain by the insulation layer having different thick-
low thermal conductivity. As the insulation layer has nesses (h = 0.1 m and h = 0.2 m). This time the area
its greater effect very close to surface heat contents for for the temperature increase with insulation layer and
199
Figure 7. Heat content changing with distance from the cen- Figure 8. Heat content versus distance from the center for
ter of the array (a) at a depth of 1.5 m; (b) at a depth of Denver claystone (a) Denver claystone; (b) Comparison of
6.0 m. different soil types.
the temperature increase without the insulation layer

were calculated for different soil types. The contribu- by Lu and Dong (2015) has shown that there is a sig-
tion of the insulation layer ranges between 60–70% nificant increase in thermal conductivity as well as
for Hopi sit while this value is 24% for Denver clay- specific heat capacity with increasing degree of satu-
stone inside of the array. It was also observed that ration from dry to unsaturated conditions. Thus in the
when the thickness of the insulation layer is doubled, modeling efforts thermally induced water flow is con-
the heat content increases in the array while it slightly sidered as the water in the pores decrease in density
decreases outside of the array at very close to the sur- and moves away from heat source to cold regions. This
face. This is because with the thicker insulation layer movement is mainly dependent on the hydraulic con-
more heat can be preserved inside the array result- ductivity of the different soils (Catolico et al. 2016)
ing in a lower temperature increase outside of the heat and the temperature gradients.
exchanger array. This increase also differs with differ- To understand the unsaturated water flow in the heat
ent type of soils as shown in Figures 9(a) to 9(d). The exchanger arrays due to the temperature gradients the
heat content increase with increasing insulation layer initial and final volumetric water content profiles were
is quantitatively greater in soils having higher ther- plotted for Hopi silt in Figure 10. It was seen that even
mal conductivity than in soils with a comparatively though this movement is small, its contribution to the
lower thermal conductivity. There was no significant specific heat capacity is not negligible. The biggest
increase in heat content at a depth of 6 m due to the increase in volumetric water content was observed
effect of the insulation layer. This is because the heat very close to the heat source due to the higher temper-
at this depth is not affected greatly by the surface ature increase while the thermally induced water flux
temperature fluctuations. magnitude decreases with increasing distance from
Thermally-induced water flow has been studied by the center. It should be noted that this analysis does
many researchers as the thermal properties of soils not consider thermally-induced vapor flow, which may
change with degree of saturation (Philip & deVries have an additional effect on the thermally induced
1957, Smits et al. 2012, Başer et al. 2014). The study water flow. However, this analysis is quite complex and
200
Figure 10. Volumetric water content profiles inside and
outside of the array.
5 CONCLUSIONS
This study focuses on numerical simulations of the

spatial and temporal temperature distributions in soil-
borehole thermal energy storage (SBTES) systems
with and without a surficial insulation layer. A 3D
transient finite element model was built to consider
representative field conditions in an unsaturated soil
layer in and around the SBTES system. Experimen-
tal and numerical results indicate that installing an
insulation layer on top of the heat exchanger array
leads an increase in the temperature in the shallow
subsurface by reducing the heat exchange between the
subsurface and the thermal storage array. The “heat
content” of a thermal heat exchanger array defined as
the area between the time series curves for the tem-
perature increase by the constant heat input and the
baseline temperatures normalized by the time was used
to quantify the heat storage in the array. The magni-
tude of heat content (HC) was observed to increase in
the array with increasing depth. Soils having a higher
thermal conductivity has a lower HC values as the heat
can dissipate faster under high temperature gradients.
Increase in temperature with an insulation layer is also
quantified with the increase in HC. This increase is
bigger in the soils having a relatively higher thermal
conductivities. When the insulation layer is doubled
the preserved amount of heat in the array increases.
This increase is significant right below the surface
while there is no significant increase with increasing
depth.
ACKNOWLEDGEMENTS
Figure 9. The effect of insulation layer on the heat content
(a) Hopi silt h = 0.1 m; (b) Hopi silt h = 0.2 m; (c) Denver Funding from National Science Foundation (NSF
claystone h = 0.1 m; (d) Denver claystone h = 0.2 m. 1230237) is much appreciated. The opinions are those
of the authors alone and do not reflect those of the
sponsor.
will be left for a future study. The conclusions regard- REFERENCES
ing the role of the insulation layer are still expected to
be valid regardless of not considering this additional Başer, T., Linkowski, D. & McCartney, J.S. 2014. Charg-
coupled flow process. ing and discharging of soil-borehole thermal energy
201
storage systems in the vadose zone. In: Bouazza, Abdul- Lu, N. & Dong, Y. 2015. A closed form equation for
malek (Editor); Yuen, Samuel T S (Editor); Brown, Bruce thermal conductivity of unsaturated soils at room tem-
(Editor). 7th International Congress on Environmen- perature. Journal of Geotechnical and Geoenvironmental
tal Geotechnics: ICEG2014: Engineers Australia, 2014: Eng. 141(6), 04015016.
362–369. Philip, J.R. & de Vries, D.A. 1957. Moisture movement
Başer, T. & McCartney, J.S. 2015a. Development of a full- in porous materials under temperature gradients. Trans.
scale soil-borehole thermal energy storage system. Proc. Amer. Geophys. Union 38:222–232.
Int. Foundations Conference and Equipment Exposition Sibbitt, B., McClenahan, D., Djebbara, R., Thornton, J.,
(IFCEE 2015). ASCE. pp. 1608–1617. Wong, B., Carriere, J., & Kokko, J. 2012. The performance
Başer, T., Lu, N., & McCartney, J.S. 2015b. “Operational of a high solar fraction seasonal storage district heating
response of a soil-borehole thermal energy storage sys- system – Five years of operation. Energy Procedia, 30:
tem.” ASCE Journal of Geotechnical and Geoenviron- 856–865.
mental Engineering. 04015097-1-12. 10.1061/(ASCE) Smits, K.M., Sakaki, S.T., Howington, S.E., Peters,
GT.1943-5606.0001432. J.F., & Illangasekare, T.H. 2013. Temperature depen-
Başer, T., McCartney, J.S., Moradi, A., Smits, K., and Lu, N. dence of thermal properties of sands across a wide
2016. “Effect of a thermo-hydraulic insulating layer on the range of temperatures (30–70◦ C). VadoseZone Journal,
long-term response of soil-borehole thermal energy stor- doi:10.2136/vzj2012.0033.
age systems.” GeoChicago 2016: Sustainability, Energy Zhang, R., Lu, N. & Wu,Y. 2012. Efficiency of a community-
and the Geoenvironment. Chicago. Aug. 14–18. pp. 1–10. scale borehole thermal energy storage technique for solar
Bear, J. 1972. Dynamics of Fluids in Porous Media. Dover, thermal energy. Proc. GeoCongress 2012. ASCE. 4386–
Mineola, N. Y., 764 p. 4395.
Brandl, H., 2006. Energy foundations and other thermo- Yesiller, N., Hanson, J.L. & Liu, W-L. 2005. Heat Generation
active ground structures. Géotechnique 56(2): 81–122. in Municipal Solid Waste Landfills. Journal of Geotech-
Catolico, N., Ge, S., & McCartney, J.S. 2016. Numeri- nical and Geoenvironmental Eng. 131(11), 1330–1344.
cal modeling of a soilborehole thermal energy storage
system. Vadose Zone Hydrology. 1–17. doi:10.2136/
vzj2015.05.0078.
202
Advanced shallow geothermal systems: Temperature induced cracking of

backfill materials and system hydraulic conductivity
H. Anbergen
APS Antriebs-, Prüf- und Steuertechnik GmbH, Rosdorf, Germany
W. Rühaak & I. Sass

Technische Universität Darmstadt, Darmstadt, Germany
J. Frank
Frank GeoConsult GmbH, Hamburg, Germany
ABSTRACT: Shallow geothermal systems are one of the key technologies for a renewable and sustainable
energy supply. The most common system is the borehole heat exchanger (BHE). These systems are capable to
provide both, heating and cooling. A significant reduction of fossil thermal and electrical energy consumption
can be achieved by using this kind of renewable geothermal energy. Furthermore the underground can be used as
a thermal storage (UTES). For an efficient use of these advanced systems, fast thermal loading cycles are needed.
However, legal constraints for the groundwater protection must be fulfilled as well. The fast heat extraction from
the underground requires materials that are resistant to rapid temperature changes. These temperature changes
might include temperatures below zero degrees Celsius and consequently a potential freezing of the pore water
of the surrounding grouting material. The frost resistance of grouting materials is discussed controversy over the
past decades. Recently a testing procedure was developed that quantifies the influences of freeze-thaw-cycles on
the hydraulic conductivity of the system BHE. The main component is a testing device that simulates the in-situ
geomechanical boundary conditions and quantifies the sealing capability of the grout. Due to the consideration
of the in-situ direction of the freezing process, the results differ substantially from earlier investigations on frost
resistance. With this procedure standardized and repeatable evaluations become feasible. This paper presents the
testing device itself including numerical and experimental proofs of concept. Besides the numerical simulation
of the phase change processes inside a grout specimen, results from calibration and round robin test will be
discussed. The thermo-hydraulic influences and implications on the modelling of the heat flow in BHEs are
analysed. The experimental results enable a comparison of the frost resistance of commercial grouts used (e.g.
in central Europe). Finally the options and needs for further investigations and the implementation for the
construction and operation of advanced geothermal systems are discussed.
1 INTRODUCTION is potentially a region of increased hydraulic conduc-

tivity. In order to prevent an increased water transport
1.1 Borehole heat exchanger through the borehole, the sealing of the borehole is
mandatory in several countries, such as Germany. The
The use of geothermal energy is still one of the most
BHE needs to be back filled as soon as the probe
promising technologies for achieving the goals of car-
is installed. Besides the permanent embedment, the
bon dioxide reduction. The most popular and installed
backfill needs to meet two main features. On the one
type of shallow geothermal systems is the borehole
hand it needs to have adequate sealing properties,
heat exchanger (BHE). A heat pump is connected to a
which ensure the hydraulic integrity of the aquicludes.
polyethylene (PE) probe that is mounted into a bore-
On the other hand the temperature conductivity needs
hole. The space between probe and borehole wall is
to be high enough in order to provide sufficient thermal
filled with a backfill material that ensures a permanent
exchange with the surrounding soil.
embedment of the probe in the underground. Figure 1
The requirements must be fulfilled at any time
shows a schematic of a double-U-pipe BHE.
during any states of operation.
1.2 Requirements on the backfill material 1.3 Freeze-thaw-cycles

In central Europe it is likely that layered aquifers The most critical states of operation for a BHE are
and aquicludes are penetrated. The drilled borehole large amplitudes of the temperature of the carrier
203
Figure 3. Schematic of the developed FTC testing cell and
its infrastructure for measurement of hydraulic conductivity
and for tempering the specimen.
Figure 1. Schematic of a borehole heat exchanger and its
components (DGG & DGGT 2014). focusses on the described frost induced cracking and
the effects on the hydraulic sealing.
2 TEST PROCEDURE
2.1 Testing setup and procedure

The test procedure aims to assess the frost resistance
of geothermal grouts with a strong focus on the in-situ
boundary conditions. In order to ensure reliable results
with a direct link to the in-situ system, the following
requirements to the testing procedure were formulated
(Anbergen et al. 2011).
1) The crucial parameter for an assessment of backfill
materials for BHEs is the hydraulic conductivity
(of the system).
2) For an evaluation of the frost resistance it is neces-
sary to measure the hydraulic conductivity (of the
system) before and after a FTC-simulation.
Figure 2. Temperature profile of a BHE system in northern 3) The specimen must remain in the test cell dur-
Germany during winter time. ing the whole test procedure and tree-dimensional
mechanical boundary conditions (ε1,2,3 and/or
o’1,2,3 ) must be applied.
fluid. Depending on the type of operation, temper- 4) The freezing direction of the test procedure must
atures above 70◦ C (UTES) or below 0◦ C (extensive follow the in-situ freezing direction from inside to
heat extraction) need to be taken into account. Fig- the outside.
ure 2 shows monitored operation temperatures of a
BHE system in northern Germany during periods of Based on these requirements a testing device
extensive heat extraction. (Figure 3) was developed that allows an independent
It is obvious that the flow temperature decreases to assessment of the hydraulic properties of grouting
a level below 0◦ C. At this temperature phase change material under cyclic freeze-thaw-events.
events from pore water to ice become possible. This
phenomenon is commonly referred to as freeze-thaw-
2.2 Specimen preparation
cycles (FTC). The ice lens growth with its ice pres-
sure can induce cracking. It was uncertain how this In order to fulfill the forth requirement a special speci-
affects the hydraulic sealing of the system. This paper men design is needed. The specimens are a simplified
204
Figure 4. Specimen composed of an axial pipe and grout for
combined tests of the system’s hydraulic conductivity and the
frost resistance of the grout.
model of the system BHE. An axial PE pipe is sur-

rounded by grouting material (Figure 4). Hence the
pipe can be tempered by a fluid and operation cycles
of a BHE can be simulated. Figure 5. Quartz flour as a standard material for round robin
For specimen preparation special casting systems tests.
were built, as the pipe must be perfectly centered in
the grout body. The grout suspension was mixed and according to ASTM D-5084 (2010) and DIN 18130-1
filled into the casting system, embedding the pipe. (1998).
The specimen cures under in-situ temperature (10◦ C)
and humidity conditions (evaporation prevented) for
28 or 56 days. For the test the specimen is demolded 3 PROOF OF CONCEPT
and trimmed to defined axial dimensions with planar
surfaces. 3.1 Round robin tests
One critical issue about geotechnical testing is to
2.3 Hydraulic conductivity of the system ensure the reproducibility of a testing procedure and its
results. In order to test whether the results are indepen-
For the assessment of potential hydraulic processes in dent from the laboratory where the tests are performed
the underground it is necessary to have information or not, the testing equipment was built up in three labo-
about the hydraulic conductivity. In many cases the ratories.As a standard test material quartz flour (Figure
hydraulic conductivity of the observed material has 5) of a defined grain size distribution was chosen. The
no dependency on the direction of flow. However, in industrial production ensures a comparable quality and
the case of BHEs the hydraulic conductivity strongly availability.
depends on direction. Due to the contact surface of The results of the three laboratories with its test dif-
the PE pipe and the backfill material there is a gap fer only within a small range, as shown in Figure 6.
that is a preferred water connection area. The hydraulic With this basis it is proven, that the testing set up can
conductivity in direction of the BHE pipe is higher than be installed independently and produce comparable
the hydraulic conductivity of the bulk backfill material results.
or bulk PE. For groundwater flow analysis the higher
hydraulic conductivity needs to be taken into account.
Consequently, it is necessary to assess the system BHE 3.2 Visual observations
(or an accurate model) instead of the bulk material for
reliable observation of the groundwater flow. All backfill specimen tested were observed visually
regarding the cracking schemes. The schemes are of
a concentric shape. With increasing distance from PE
2.4 Freeze-thaw tests and change in hydraulic pipe, which is the frost entrance point, the number of
conductivity cracks increases. This is in agreement with the theory
The formation of ice lenses during the freezing proof cryostatic suction processes (Konrad & Lemieux
cess of the backfill material likely induces cracking. 2005, Unold 2006).
The cracks are of a concentric shape and are preferred
water path ways for vertical water flow. The number
3.3 Tracer tests
and size of frost induced cracks depends on the mate-
rial tested. Hence, a general assessment whether the Several specimens were tested with a tracer fluid in
backfill material is prone to cracking or not is feasible. order to visualize the preferred flow paths through the
In order to provide a unified test conditions, all tests system. It can be observed that most water transport
are executed with the same pore-pressure, freezing and occurs between the PE pipe and the backfill material
thawing temperatures and durations. For freezing the (Figure 7).This is in good agreement with observations
fluid temperature was set to −10◦ C for a minimum of grouted wells (Baumann et al. 2003).
of 20 hours, for thawing to +8◦ C. Before and after Besides the dominant water flow along the PE pipe,
each FTC the hydraulic conductivity was determined there is fluid transport through the matrix. Especially
205
Figure 8. Relative increase in hydraulic conductivity. Com-
parison of bulk material and system specimens.
Figure 6. Comparison of the measured hydraulic conductiv-

ities of the standard material in three different laboratories. model was set up in order to verify whether the
specimen freezes axisymmetric or not. The model
was implemented using the software code FEFLOW
(Diersch 2014). FEFLOW is a common code for
thermos-hydraulic coupled groundwater modelling
and is frequently applied for geothermal simulation.
In order to simulate the phase change of the freezing
process a plug-in was developed that manipulated the
thermo-hydraulic parameters of the fluid. It includes
an apparent heat capacity approach that respects the
latent heat effects during crystallization of water to
ice and vice versa (Mottaghy & Rath 2006, McKenzie
et al. 2007).
The numerical simulations prove the primary radial
freezing process inside the testing cells as well that
the chosen testing times are sufficient for a complete
phase-change of the pore water (Anbergen 2015b).
4 RESULTS
4.1 System hydraulic conductivity

The hydraulic conductivity of the system is signifi-
cantly higher than the hydraulic conductivity of the
Figure 7. Inner surface of a specimen, split up after tracer bulk backfill material. Due to the PE pipe and its
test. preferred flow path the axially directed water flow
increases about two orders of magnitude (Anbergen
the concentric, frost induced cracks are a preferred 2015). Figure 8 shows the relative increase in hydraulic
water path. Again the total amount of water streaming conductivity.
through the cracks depends on the material and its frost
resistance.
3.4 Numerical model

As the testing procedure follows the in-situ freez- With kMat the hydraulic conductivity of the bulk
ing direction radially from inside out, a numerical material, kSys the hydraulic conductivity of the system.
206
Figure 9. Relative increase in hydraulic conductivity of the
system caused by 6 (or more) FTC.
4.2 Frost resistance

In dependence on the number of FTC, there is an Figure 10. Comparison of the absolute range of increase
increase of the system hydraulic conductivity. The first in hydraulic conductivity. Comparison of bulk material and
FTCs have the gravest effect on the conductivity. system specimens.
Most materials tested show no further increase after
6 FTC, which is in agreement with observations on due to cyclic freeze-thaw-simulations. A comparison
clay-liners of landfill-covers (ASTM D-6035 2008). of the obtained results of the laboratory and the large
scale tests is shown in Figure 10.
It is obvious that results are alike and the measured
system hydraulic conductivities have a similar order
of magnitude. This proofs the equality of the testing
setups. Besides the general systems hydraulic conduc-
With kFTC the hydraulic conductivity of the system tivity the increase due to FTCs is comparable as well,
after 6 (or more) FTC. as jointly published (Anbergen et al. 2015b).
The increase of the system hydraulic conductiv-
ity caused by FTC depends on the properties of the
grouting material. There is a significant difference
between materials with or without swelling (clay) com- 5.2 Practicability of the test procedure
ponents. Backfill materials with a certain percentage The practicability of the test procedure was tested of
of swelling components have a higher frost resistance the past years. About 80 percent of the geothermal
than pure cement based backfills. Figure 9 shows the backfill materials available in Germany were tested.
relative increase in hydraulic conductivity due to FTC There is a broad acceptance of the procedure. The lat-
(a minimum of 6 FTC were executed). est guidelines for shallow geothermal systems suggest
a quality control according to the described testing
routine (e.g. DGG & DGGT 2014, VDI 4640-2 Draft
5 DISCUSSION 2015). With the procedure an independent and uni-
fied assessment of the frost-restistance of geothermal
5.1 Comparison to large testing setups backfill materials becomes feasible.
While the described laboratory testing setup was estab-
lished, a large scale test was carried out by Kuckelkorn
& Reuß (2013). Specimens of an axial length of 2.75 m 6 OUTLOOK
and a diameter of 150 mm were installed in a hydraulic
conductivity apparatus. The specimens were com- Based on the findings, there is a further development of
posed of a real size HDPE-probe and different backfill the testing cell. Due to its construction there are several
materials. The system’s axial hydraulic conductivity further fields of research such as the heat transport of
was determined as well as the change of conductivity buried power cables, in which the cell can be applied.
207
Therefore the cell will be equipped with further sensors Dalla Santa, G., Galgaro, A., Tateo, F. & Cola, S. (2016).
and a thermal conductivity measuring system. Modified compressibility of cohesive sediments induced
The FTC-behavior of geothermal backfill was ana- by thermal anomalies due to a borehole heat exchanger.
lyzed by several further researchers (e.g. Erol & Engineering Geology 202: 143–152
Diersch, H.-J.G. (2014). FEFLOW – Finite Element Modeling
François 2014). There thermos-mechanic models were of Flow, mass and heat transport in porous and fractured
developed in order to predict the cracking schemes. media. Springer, Berlin, Germany
The mechanical approaches will be combined with DIN 18130-1 (1998). Baugrund – Untersuchung von
the thermo-hydraulic phase change plug-in (Anbergen Bodenproben; Bestimmung des Wasserdurchlässigkeits-
et al. 2015a). The performance of the plug-in is beiwertes – Teil 1: Laborversuche. Deutsches Institut für
assessed in an international benchmark for phase- Normung e. V. (ed.), Beuth Verlag, Berlin.
change codes, called INTERFROST (Grenier et al. DGG & DGGT.(2013). Empfehlungen des Arbeitskreises
2016). Geothermie – Oberflächennahe Geothermie – Planung,
Besides the ongoing benchmark, there are fur- Bau, Betrieb, Qualitätssicherung, Version der Offenle-
gung, Deutsche Gesellschaft für Geowissenschaft e.V.
ther numerical investigations based on the developed (DGG) und Deutsche Gesellschaft für Geotechnik e.V.
FEFLOW code (Dalla Santa et al. 2016) in order to (DGGT) (ed), Ernst & Sohn, Berlin
increase the reliability of the prediction of thermal Erol, S. & François, B. (2014). Efficiency of various grouting
plumes of BHEs. materials for borehole heat exchangers. Applied Thermal
Engineering 70: 788–799
Grenier, C., Rühaak, W. & The Interfrost Team (2016). The
REFERENCES InterFrost benchmark of Thermo-Hydraulic codes for cold
regions hydrology – first inter-comparison phase results.
Anbergen, H., Frank, J., Albrecht, I. & Dittrich, H. (2011). EGU General Assembly 2016, Vienna
Prüfzelle zur Bestimmung des Frost-Tau-Wechsel- Konrad, J.-M. & Lemieux, N. (2005). Influence of fines on
Widerstands von Verpressmaterial für EWS, bbr – frost heave characteristics of a well-graded base-course
Fachmagazin für Brunnen- und Leitungsbau, 2011/10: material. Canadian Geotechnical Journal 42(2): 515–527
38–43. Kuckelkorn, J. M. & Reuß, M. (2013). Hydraulische Sys-
Anbergen, H., Frank, J., Müller, L. & Sass, I. (2014). Freeze- temdichtheit und Frostbeständigkeit von Erdwärmeson-
Thaw-Cycles on Borehole Heat Exchanger Grouts: Impact den, bbr – Fachmagazin für Brunnen- und Leitungsbau,
on the Hydraulic Properties, Geotechnical Testing Jour- Sonderheft Geothermie, (2013): 6–13
nal, 37 (4): 639–651. McKenzie, J.M., Voss, C.I., & Siegel, D.I. (2007). Groundwa-
Anbergen, H., Rühaak, W., Frank, J. & Sass, I. (2015a). ter flow with energy transport and water-ice phase change:
Numerical simulation of a freeze–thaw testing procedure Numerical simulations, benchmarks, and application to
forborehole heat exchanger grouts. Canadian Geotechni- freezing in peat bogs. Advances in Water Resources, 30:
cal Journal 52(8): 1005–1022 966–983
Anbergen, H., Frank, J., Reuß, M., Kuckelkorn, J., Müller, L. Mottaghy, D., & Rath, V. (2006). Latent heat effects in sub-
& Sass, I. (2015b). Hydraulische Integrität des Systems surface heat transport modelling and their impact on
Erdwärmesonde. bbr – Fachmagazin für Brunnen- und palaeotemperature reconstructions. Geophysical Journal
Leitungsbau, 2015/02: 34–41 International, 164: 236–245
ASTM Standard D-5084 (2010). Standard test methods Unold, F. (2006). Der Gefriersog bei der Bodenfros-
for measurement of hydraulic conductivity of saturated tung und das Kompressionsverhalten des wieder aufge-
porous materials using a flexible wall permeameter. tauten Bodens. Ph.D. thesis, Universität der Bundeswehr
Annual book of ASTM Standards, ASTM International, München, München, Germany
West Conshohocken, Pa., USA VDI 4640-2 Draft (2015) Thermische Nutzung des Unter-
Baumann, K., Niehues, B., Tholen, M. & Treskatis, C. (2003). grundes – Erdgekoppelte Wärmepumpenanlagen. Verein
Untersuchungen zur Bestimmung von Qualitätskriterien Deutscher Ingenieure (ed), Beuth Verlag, Berlin
für Abdichtungsmaterialien im Brunnenbau, Abschluss-
bericht, Deutsche Vereinigung des Gas- und Wasserfaches
e.V. – DVGW, Bonn
208
Studying cryogenic fracturing process using transparent specimens
M. Cha
N.B. Alqahtani
King Abdulaziz City for Science and Technology (KACST), Riyadh, Saudi Arabia
B. Yao, L. Wang, X. Yin & Y.S. Wu

Colorado School of Mines, Golden, Colorado, USA
T.J. Kneafsey
Lawrence Berkeley National Laboratory, Berkeley, California, USA
ABSTRACT: Cryogenic fracturing exploits thermal gradient and resulting local tensile stress to initiate frac-
tures on a surface exposed to cryogenic fluids. This study investigates the development and morphology of
cracks generated from cryogenic thermal shock in a borehole geometry under no external confining stress.
A borehole was drilled through transparent acrylic specimens. Liquid nitrogen was injected into the wellbore,
and the fractures were initiated by the thermal shock. The initiated fractures allowed further penetration of the
cryogen, which helped to propagate fractures throughout the specimen. Fracture growth was characterized by
abrupt starts and stops, which suggest that the tensile stress inside the borehole must reach a certain threshold for
fracture initiation and growth. Two distinctive patterns in crack development were observed: horizontal-planar-
radial pattern created by longitudinal thermal contraction, and vertical cracks by circumferential contraction.
The horizontal cracks appeared to be spaced by a certain length, known as the exclusion distance.
1 INTRODUCTION than two Olympic swimming pools or supply water for

65 families for one year. Finally, high pressure down-
Hydraulic fracturing has been demonstrated as one of hole injection of fracturing fluids containing chemical
the most effective technologies and are widely applied additives has led to a contentious community and polit-
in combination with horizontal drilling for developing ical climate over underground water contamination.
shale reservoirs (Steward, 2013). Due to the general In contrast to hydraulic fracturing, cryogenic frac-
availability and low cost of water, fracturing fluids of turing using liquid nitrogen (LN) offers potentially
hydraulic fracturing are mostly water-based, contain- greater fracturing capabilities without any of the issues
ing proppants and some chemical additives (Sharma associated with water-based fracturing fluids.
et al., 2004, Shaefer, 2005). Cryogenic fracturing is a relatively new stimula-
Several major shortcomings accompany the appli- tion technology that looks to expand and improve the
cation of water-based fracturing fluids. First, water can traditional hydraulic fracturing technology. Cryogenic
cause significant formation damage to shale forma- fracturing rests on the idea that a sharp thermal gra-
tion in the sense that clay-rich shale tends to absorb dient caused by contacting with and vaporization of
water and swell, narrowing the conductive fractures a cryogen, can induce fractures when brought into
and pores. Also, capillary retention of water would contact with a much warmer rock under downhole
partially or completely block the flowpath of hydrocar- conditions. Cryogen exists in the gaseous phase at
bon from matrix to fracture networks (Mazza, 1997). standard conditions but takes a liquid form at low tem-
Secondly, water usage in large quantities is concern- peratures, such as liquid nitrogen and liquid carbon
ing, placing demands upon local water supply and dioxide. Specifically, when liquid nitrogen is injected
environments, especially in areas where water sup- into a borehole, heat from the rock near the borehole
ply is at shortage. For instance, during the period will quickly transfer to the liquid nitrogen at boiling
of 2009-6/2011 in Texas, the median water usage point (−195.8◦ C or −320.4◦ F at atmospheric pres-
in hydraulic fracturing for each horizontal well in sure), resulting in rapid cooling of the near-borehole
Barnett, Eagle Ford, and Haynesville were 10,600, area, which will cause the surface of the rock or
16,100, and 21,500 m3 , respectively (Nicot and Scan- borehole wall to contract. Once the tension due to con-
lon, 2012). The lower volume, 10,600 m3 can fill more traction is sufficiently increased, fractures orthogonal
211
to the interface of cryogen and rock can be initiated. nearby offset well that had undergone traditional frac-
These newly induced fractures can be further extended turing with nitrogen gas. Although the increased initial
by high pressure gas from LN vaporization. Note that production rate in this research suggests the efficacy
nitrogen has a liquid-to-gas expansion ratio of 1:694 of cryogenic fracturing, there could be a number of
at 20◦ C (68◦ F) and atmospheric pressure. reasons why an offset well in a shale formation might
Although cryogenic fracturing has not been widely produce differently, including anisotropic stress con-
deployed for developing unconventional reservoirs, it ditions and heterogeneous reservoir conditions over
was tested in a few field cases during the 1980s and short distances.
1990s. Instead of water, Lillies and King (Lillies and Although several field cases have been imple-
King, 1982, King, 1983) pumped gelled liquid carbon mented, during the past 15 years, no further studies
dioxide at −28.9◦ C to −40◦ C (−20◦ F to −40◦ F) to were continued for better understanding and appli-
stimulate tight gas sand formations using standard tub- cation of this fracturing technology. The fracturing
ing and casing configurations. On average, 3–4 days processes, mechanisms, and controlling factors of
after the fracturing treatments, oil and gas wells were cryogenic fracturing are still poorly understood. We
cleaned up with complete flowback of vaporized liq- conducted preliminary cryogenic tests to understand
uid carbon dioxide, without producing any formation the cryogen and material behaviors by performing sub-
damage. In these cases, the gelled carbon dioxide was mersion tests and applying cryogen to boreholes in
capable of carrying proppants due to its higher viscos- unconfined concrete specimens (Cha et al., 2014). In
ity than liquid CO2 , hence the fractures were able to this study, we investigate the development and mor-
stay open. Accordingly, all the wells for which they phology of fractures generated from cryogenic thermal
published results experienced increased production shock in a borehole geometry in the laboratory. Liq-
rates (Lillies and King, 1982, King, 1983). uid nitrogen is injected into the center of a transparent
McDaniel et al. (1997) conducted simple laboratory acrylic block to visualize fracture initiation. It is to
studies in which coal samples were immersed in LN for understand the cryogenic fracturing mechanisms and
observation of their fracturing process. The coal sam- toward developing and improving the process for field
ples experienced significant shrinkage and broke into applications.
smaller cubic units, creating microfractures orthogo-
nal to the surface exposed to the liquid nitrogen. The
researchers found that repeated exposure cycles to the
cryogen caused the coal to break into smaller pieces, 2 EXPERIMENTAL SETUP
or become rubblized. After three cycles of submersion
into liquid nitrogen and warm-up to ambient tem- 2.1 Devices and procedure
peratures, the coal sample was reduced to grain-size
We consider fracturing cryogenic thermal shock,
particles. McDaniel et al. (1997) then continued field
which depends on pure thermal gradient and result-
tests with liquid nitrogen, and published before-and-
ing thermal tensile fracturing and subsequent cryogen
after production rates for five wells. The results were
transportation into fractures.
mixed: three CBM wells showed increased production,
In the test setup, we are mainly concerned about
one CBM well showed equivalent production, and one
cooling the borehole as rapidly as possible to maxi-
low permeability sandstone well initially completed
mize thermal gradient. This is done by flowing LN2
with slick water fracturing showed decreased produc-
continuously through the borehole. In this lab-scale
tion. By injecting liquid nitrogen, Grundmann et al.
experiment, LN2 was pumped from the Dewar by
(1998) treated a Devonian shale well and observed an
pressure difference using a liquid nitrogen withdrawal
initial production rate 8% higher than the rate in a
device (Figure 1). Liquid nitrogen was transported by
a vacuum-jacketed hose to the specimen, and injected
into the borehole and then directed to an outlet. A
pressure transducer was attached to monitor the bore-
hole pressure. In this thermal shock setup, pressure
inside the borehole was basically the same as the
pressure inside the Dewar. Cryogenic fracturing was
done purely by thermal gradient; little pressure existed
inside the boreholes (less than 70 kPa) throughout
the thermal shock. Pressure inside the borehole, LN2
consumption, and temperature were monitored and
logged. A pressure transducer was located at the top
of 20 cm-long stainless steel extension tube (3.2 mm
OD), which create vapor cushion and dissipate low
temperature to limit heat transfer (temperature above
0◦ C (32◦ F) observed at the top of the tubing throughout
testings). Thermocouples are used to measure cryo-
genic temperature, and their thin wires allow prompt
Figure 1. Setup for cryogenic stimulation experiments. response to temperature changes. Having data for both
212
temperature and pressure provide reliability about data with droplets of liquid, and finally flown in a more con-
interpretation. More complete information about the tinuous phase of liquid with still a significant amount
setup can be found in (Cha et al., 2014). of gas phase intermixed.
Taking advantage of the specimen being transpar-
ent, we observed the flow characteristics inside the 2.2 Specimen preparation: transparent acrylic
borehole. Upon the start of the experiment, nitrogen specimens
inside the borehole was flown initially as a gas (for
about 1 ∼ 2 minutes), and then flown as a gas mixed Two acrylic specimens are used as transparent speci-
mens. Acrylic specimens are chosen because they are
transparent, and relatively brittle, which is one of the
important characteristics of rocks.
The dimensions of the acrylic specimen 1 are illus-
trated in Figure 2a. The acrylic cylinder is 10 cm in
diameter and 23 cm in height and the borehole is drilled
from top, and 18 cm in depth and 1.3 cm in diameter.
A 1.3 cm O.D. stainless steel tube was inserted and
attached to the borehole wall using epoxy to the depth
of 6.4 cm. An LN2 inlet tubing was inserted to 5.7 cm
beyond the casing end.
The sample dimensions of Specimen 2 are the same
as those of the Specimen 1. However, unlike Specimen
1, both the steel casing and the inlet point were 3.8 cm
in depth (Figure 2b). The injection point was purposely
placed higher than Specimen 1 to study the effect of
the injection point location.
3 RESULTS
3.1 Temperature, pressure, and LN2 consumption

Temperature of the specimen dropped rapidly with the
introduction of LN2 , and reached LN2 boiling point
within five minutes in this laboratory setups. The tem-
perature distribution at the surface was also dependent
on the proximity to the cracks due to transportation of
LN2 through cracks. Temperatures dropped by non-
negligible amount shortly after the end of the test (e.g.
TC #2 of Specimen 1 and TC #2, TC #4, and TC #5
Figure 2. Acrylic specimens tested – Dimensions and of Specimen 2 in Figure 3), which probably caused by
locations of the stainless steel (SS) casing and the inlet tube. the pressure drop at the borehole.
Figure 3. Locations of thermocouple tips and temperature evolutions during the cryogenic thermal shock experiments.
213
Although a lot of LN2 (20 kg) was flown in the characterized by abrupt starts and stops. This suggests
experiment 1, most of the fractures occurred at an that the tensile stress generated inside the borehole
early stage (within 15 minutes). The Dewar lever was must reach a certain threshold for fracture initiation
opened fully during releasing LN2 without interme- and growth. The increased material’s brittleness at low
diate closure. The experiment for Specimen 2 was temperature may have also contributed to this behav-
terminated by depletion of the LN2 tank. The duration ior. During the experiment, clear audible sounds were
of the experiment 2 was 11 minutes and the amount of emitted, when the fractures were observed to grow. The
nitrogen consumption was 7.6 kg. Pressure measured magnitude/amount of instantaneous growth between
at the borehole was in the range of 20–35 kPa, which starts and stops tends to decrease as the fracture grew
was exerted by the Dewar tank. larger. Most of the cracks occurred within 20 min-
utes. Two distinctive patterns in crack development
were observed: horizontal, planar, radial fractures, and
3.2 Crack development vertical cracks joining the horizontal fractures.
Images of specimens were captured throughout the The horizontal, planar, radial fractures form the
testings using a digital camera. dominant pattern of crack morphology. This can be
explained by the fact that the specimen is cylindri-
3.2.1 Specimen 1 cal with a borehole height greater than the diameter,
Images of specimens were captured in a video through- which makes thermal contractions more pronounced
out the experiment (Figure 5). It was observed that in the longitudinal direction. The horizontal fractures
fracture growth was not continuous, but rather jumpy, were clearly spaced by a certain length, which can be
considered as an “exclusion distance”. An exclusion
distance exists because a set of crack cannot be created
closer than a certain length due to a limited amount of
thermal contraction (Figure 6). The behavior of exclu-
sion distance also exist in other phenomena, such as
Figure 6. Crack morphology and driving thermal tensile

Figure 4. Borehole pressure and LN2 released during the stresses – Specimen 1.
test – Specimen 1.
Figure 5. Crack development. The steps do not represent all the crack growth steps – Specimen 1.
214
mud crack, frost heaving area, and dissolution pipes of the experiment, and the temperature on surface near
etc. (Buijse, 2000, Toramaru and Matsumoto, 2004, the crack was readily affected by the proximity to the
Jenkins, 2005). cracks.
Fractures were generated in the vertical direction in
a less magnitude compared to the horizontal fractures,
caused by the circumferential thermal contraction. The 3.4 Effect of casing/inlet location
vertical tensile fractures tend to initiate from or form For both Specimen 1 and Specimen 2 tests, one major
between the horizontal fractures and bridge them. horizontal fractures were initiated from early stage at
It is energy-efficient to start from one pre-existing the steel cased parts of the acrylic cylinders (Figure 5
defect (i.e., a horizontal fracture) and propagate toward and Figure 7) although they are far from inlet ports.
another pre-existing defect (Figure 5 – Figure 6).At the This is perhaps due to the efficient heat transfer of
end of the experiment, the specimen showed a complex the casings, which has a high heat conductivity. The
fracture morphology created by the interplay between steel casing will also have shorter period during which
longitudinal and circumferential thermal contractions it is under Leidenfrost effect. These fractures, how-
(Figure 5). ever, did not propagate as far as those in the borehole.
This, perhaps, is due to the steel casing blocking the
3.2.2 Specimen 2 flow of LN2 into these fractures; this may be also due
Two horizontal fractures were observed: one in the to the steel casing, epoxied to the acrylic cylinder,
steel cased part of the acrylic sample and another right changing the stress condition and preventing further
next to the inlet port (Figure 7). Following the initiation growth of the fracture. Clearly, the steel casing and
and propagation of the horizontal fractures, vertical the epoxy have influenced heat transfer, flow of LN2 ,
cracks were created and they bridged the horizontal stress distribution and eventually affected the fracture
fractures. distribution.
Similar to the Specimen 1 test, the fracture in the We notice that the distribution of cryogenic temper-
uncased part was located close to the inlet port. In this ature inside the borehole was affected by the location
test, however, one big radial wing was created (com- of the injection point (Figure 2). Fractures were mainly
pared to three in the previous test), which means that generated near the injection point, which suggests
there was not enough driving longitudinal contraction colder temperatures near the injection point.
to generate more horizontal fractures. The lack of ther-
mal driving and multiple horizontal fractures could be
due to the shorter stimulation time – 11 min vs. 36 min 4 CONCLUSIONS
in Specimen 1 test, thus lower amount of LN2 applied
– 7.6 kg vs. 20 kg in Specimen 1 test, by the early Experiments were performed to study the develop-
depletion of the LN2 tank. ment and morphology of fractures generated by a
cryogenic thermal shock in a borehole geometry. We
designed our experimental apparatus and procedures
3.3 Effect of presence of fracture on temperature specifically for thermal stimulation using liquid nitro-
distribution gen. Direct observations of fracture formation were
Temperature distribution at the surface and inside the made possible by the use of transparent acrylic speci-
acrylic samples was dynamically coupled to the initia- mens. The study provides key observations on fracture
tion and growth of fractures. During the experiments, initiation and propagation when sufficient thermal
it was observed that liquid nitrogen moved into and contraction/tensile stress is achieved in a borehole.
flowed through created fractures, which helped frac- Cryogenic fracture growth was observed to be
tures to further propagate. This, in turn, accelerated abrupt due to the brittleness of the material at the
the temperature propagation. Figure 5 and 7 show that cryogenic temperature and the accumulation-release
some cracks approached the surface at the later stage of tensile stress coupled with fracture propagation
Figure 7. Crack development – Specimen 2.
215
and heat transfer. The area of fracture created at each Journal of Petroleum Science and Engineering, 124, 436–
growth event tends to decrease as the fractures become 450.
larger. Two distinctive patterns in the fracture devel- Grundmann, S. R., Rodvelt, G. D., Dials, G. A. & Allen, R.
opment were observed: one is horizontal, planar, and E. (1998) Cryogenic Nitrogen as a Hydraulic Fracturing
Fluid in the Devonian Shale. SPE-51067-MS. SPE East-
radial propagation created by longitudinal thermal ern Regional Meeting. Pittsburgh, Pennsylvania, Society
contraction, and another is vertical fractures created of Petroleum Engineers.
by circumferential contraction. The horizontal frac- Jenkins, D. R. (2005) Optimal spacing and penetration of
tures were initiated first and were more dominant than cracks in a shrinking slab. Physical Review E, 71.
the vertical fractures. This is perhaps because the bore- King, S. R. (1983) Liquid CO2 for the Stimulation of Low-
hole height is much greater than the borehole diameter, Permeability Reservoirs. SPE-11616-MS. SPE/DOE Low
which makes thermal contractions more pronounced Permeability Gas Reservoirs Symposium.
in the longitudinal direction. The horizontal fractures Lillies, A. T. & King, S. R. (1982) Sand Fracturing With
tend to be spaced by a certain length (exclusion dis- Liquid Carbon Dioxide. SPE Production Technology Sym-
posium, 8-9 November, Hobbs, New Mexico. Society of
tance), which exists because a set of fractures cannot Petroleum Engineers.
be created closer than a certain length due to limited Mazza, R. L. (1997) Liquid CO2 improves Fracturing. Hart’s
amount of thermal contraction. The vertical fractures Oil and Gas World, 22.
tend to initiate from or form between the horizontal Mcdaniel, B., Grundmann, S., Kendrick, W., Wilson, D. &
fractures and bridge them. We expect that the sequence Jordan, S. (1997) Field applications of cryogenic nitro-
of initiation and patterns will also depend on the stress gen as a hydraulic fracturing fluid. SPE Annual Technical
condition of the specimen and this will be examined Conference and Exhibition.
in future experiments. Nicot, J.-P. & SCANLON, B. R. (2012) Water Use for Shale-
Gas Production in Texas, U.S. Environmental Science &
Technology, 46, 3580–3586.
Shaefer, M. T. (2005) Are Slick Water-Fracturing Applica-
ACKNOWLEDGEMENTS tions Effective in the J-Sand Formation? SPEAnnualTech-
nical Conference and Exhibition, 9–12 October, Dallas,
Support for this research was provided by Research Texas. Society of Petroleum Engineers.
Partnership to Secure Energy for America (RPSEA) Sharma, M. M., Gadde, P. B., Sullivan, R., Sigal, R., Fielder,
(Grant no. 10122-20). R., Copeland, D., Griffin, L. & Weijers, L. (2004) Slick
Water and Hybrid Fracs in the Bossier: Some Lessons
Learnt. SPE Annual Technical Conference and Exhibition,
REFERENCES 26–29 September, Houston, Texas. Society of Petroleum
Engineers.
Buijse, M. A. (2000) Understanding wormholing mecha- Steward, D. B. (2013) George P. Mitchell And The Barnett
nisms can improve acid treatments in carbonate forma- Shale. Journal of Petroleum Technology, 65, 58–68.
tions. SPE Production & Facilities, 15, 168–175. Toramaru, A. & Matsumoto, T. (2004) Columnar joint mor-
Cha, M., Yin, X., Kneafsey, T., Johanson, B., Alqahtani, N., phology and cooling rate: A starch-water mixture experi-
Miskimins, J., Patterson, T. & Wu, Y.-S. (2014) Cryogenic ment. Journal of Geophysical Research: Solid Earth, 109,
fracturing for reservoir stimulation – Laboratory studies. B02205.
216
Interactions between hydraulic fracture and interfaces in layered formations
S.M. Ham & Tae-Hyuk Kwon

KAIST, Daejeon, Republic of Korea
Y.J. Sim
Korea Land and Housing Corporation, Gyeongsangnam-do, Republic of Korea
ABSTRACT: Hydraulic fracturing is widely used to enhance hydrocarbon productivity or to enhance heat
recovery. Predicting fracture propagation and geometry is an important but daunting task because of complexities
in natural geologic rock conditions. This study explores the interactions between hydraulic fracture and interfaces
in layered formations.A series of laboratory experiments were performed, in which hydraulic fracture propagation
behavior in 2D gelatin plates were monitored. It was found that the fracture propagation was heavily affected by
the stiffness of bounding layers. When the stiffness of the bounding layer was lower than that with the fracture,
it appeared that the fracture passed through the interface. Whereas, fracture propagation was confined when the
bounding layer had the same or higher stiffness. This study presents the simple but unique experiment data on the
interactions between hydraulic fracture and interfaces, and can be further used to develop models for geometry
prediction of hydraulic fracture.
1 INTRODUCTION modulus of bounding layers. Laboratory experiments

were performed, in which hydraulic fracture propaga-
Hydraulic fracturing is now an important technique tion behaviors in 2D gelatin plates were monitored.
for the oil and gas industry. It is a process where Stiffness of bounding layers was controlled by adjust-
viscous fluid is pumped into a borehole at a high injec- ing the concentration of gelatin, and three different
tion rate to generate a permeable fracture. The design bounding layer formations were tested. During frac-
of hydraulic fracture requires the geometry predic- ture propagation, the images were recorded at a high
tion of the induced fractures. Many researchers have speed to obtain the fracture geometry and the pressure
suggested models that predict the fracture geometry curves were measured.
in homogeneous rocks. For example, KGD (Khris-
tianovic & Zheltov, 1955, Geertsma & Klerk, 1969)
and PKN (Perkins & Kern, 1961, Nordgren, 1972)
2 EXPERIMENTAL PROCEDURE
are representative models widely used in oil indus-
try. However, natural rock has several discontinuities,
2.1 Gelatin plate
such as joints, flaws, and faults, and this draws a
significant attention for considering the interactions Gelatin was used as an analogue to a homogeneous,
between hydraulic fractures and natural discontinuities isotropic, and elastic rock medium. It stays as a viscous
to obtain a better design. liquid form at high temperatures and as a solid gel form
Accordingly, Simonson et al. (1978) have found at low temperatures. The most important characteris-
that when the shear modulus of the barrier forma- tic of gelatin is its transparent property. This allows us
tion is less than that of the payzone, hydraulic fracture to visualize fracture propagation during experiments.
can propagate through the interface easily. Teufel & The stiffness of the gelatin was controlled by gelatin
Clark (1984) have presentedthat the large normal stress weight percent: 7.41 wt% (low stiffness, L), 12.28 wt%
is required to make fracture cross the rougher dis- (medium stiffness, M), and 16.67 wt% (high stiffness,
continuity. Further, Warlinski (2011) stated that large H). Each gelatin solution was made by mixing the
permeability terminated the growth of fracture. How- dry gelatin powder with distilled water at 50◦ C. This
ever, there are only few studies that observed hydraulic solution was then poured in an acryl mold having a
fracture propagation because it occurs within a second dimension of 200 mm × 200 mm × 10 mm, as shown
in the core-scale experiments (Frash, 2007, Bohloli & in Figure 1a. The mold was kept in a refrigerator at
De Pater, 2006). 4◦ C for 24 h to cure the gelatin. For homogeneous
Thus, this study presents high-quality images of samples with no discontinuity, the gelatin was poured
hydraulic fracture propagation at difference shear in the mold at once. For samples with discontinuity,
217
Figure 1. (a) The acryl mold used in the experiment and
(b) the composition of galatin sample with interfaces (M2H
case).
Figure 2. Experiment setup.
the gelatin plate was prepared to have three layers

with two interfaces, as shown in Figure 1b. Each layer 2.4 Experiment procedure
was made 2 h after pouring the previous layer to allow
When the curing of gelatin was completed, at the
hardening of the previous layer. The first and third
borehole two initial cracks having lengths of ∼5 mm,
layers were 50 mm thick while the middle layer was
aligned perpendicular to the layer interfaces, were cre-
100 mm thick (Figure 1b). The stiffness of the bound-
ated by scratching the gelatin using a thin wire (thick-
ing layer was controlled in this study. Six samples
ness <1 mm). Thereafter, a copper tube of 6.4 mm
of different stiffnesses were used to examine fracture
diameter was inserted into the plate. The copper tube
propagation. Three homogeneous samples were pre-
had two vertical holes of diameter 1.5 mm, and through
pared with different stiffnesses of gelatin: low stiffness
these two holes the fracture fluid was injected at a
(L-gelatin), medium stiffness (M-gelatin), and high
constant flow rate of 10 mL/min for creating hydraulic
stiffness (H-gelatin). Three layered samples were pre-
fractures. The test stopped when the hydraulic fracture
pared to have different bounding layers: medium to low
reached the edge of the gelatin plate.
stiffness (M2L), medium to medium stiffness (M2M),
and medium to high stiffness (M2H).
3 RESULTS AND DISCUSSIONS
2.2 Injection fluid 3.1 Homogenous samples
A sugar solution dyed with blue ink was used as the For L- and M-gelatin cases, two fractures propagated
fracture fluid. The viscosity of the fluid was controlled along the directions of the perforated holes. However,
by changing the concentration of sugar. To obtain a only one fracture was generated in the H-gelatin case.
viscosity of 10 cp, a sugar solution of 50 wt% (100 g The pressure curves during the fluid injection were
of distilled water and 100 g of sugar) was prepared. plotted in Figure 3a. The pressure gradually increased
This is in line with the approach taken in Lide (2000). at the beginning of fluid injection, and the fracture
To aid visual observation of the fracture shape, blue was initiated just before the peak pressure, as denoted
ink was added to the sugar solution. in Figure 3a by points A, C, and E for L-, M-, and
H-gelatin cases, respectively. Those curves are similar
with the pressure results from field tests. Points B, D,
and F indicate the end points of the experiments. Fig-
ure 3b shows the developed fractures at point C and
2.3 Experiment setup
point D in M-gelatin. The curve shapes were similar
The whole setup for the experiment is shown in Fig- for the three different experimental cases. However,
ure 2. A high-speed camera (UI-3360CP-C-HQ, IDS the peak pressure increased as the stiffness increased.
Corp., Korea) was used to record images at 0.02 s Once a fracture was initiated, the pressure decreased
intervals (50 frames/s). An LED lamp was installed to smoothly and became almost constant during fracture
control the light intensity, facing down to the gelatin propagation.The fracture in L-gelatin initiated approx-
plate. A transfer vessel and a syringe pump (500HP, imately 5 s after the fluid injection at a breakdown
ISCO Teledyne) were used to inject the fracture fluid. pressure of 39 kPa. M-gelatin required a pressure of
The syringe pump provided water to the transfer vessel 66 kPa for initiation of fracture, which was higher than
at a constant flow rate of 10 mL/min. Thus, the fracture that required for L-gelatin. H-gelatin required a pres-
fluid was pumped to the perforated borehole from the sure of 149 kPa for fracture initiation, as denoted by
transfer vessel at the same flow rate. During the exper- point E in Figure 3a. More pressure was needed to ini-
iment, fluid injection pressure was measured at 0.1 s tiate the fracture at stiffer gelatin, therefore the higher
intervals by using a data logger (34972A, Agilent). initiation pressure was measured.
218
Figure 4. (a) Pressure curves for layered samples and (b)
fracture images of layered sample.
Figure 3. (a) Pressure curves for homogeneous samples and bounding layer with high stiffness (H-gelatin, point H
(b) fracture images of M-gelatin sample. in Figure 4a). Because of the containment, the pres-
sure increased at the interface, and as a result, the
second fracture was generated and then the pressure
decreased. When the second fracture reached the inter-
3.2 Layered samples face (point I in Figure 4a), it propagated along the
interface.
The different fracture behaviors and pressure curve
shapes were found in the layered samples. In the M2L
case, the fracture propagated through the interface, 3.3 Fracture length
and the fracture width became thicker after passing
through the interface, as shown in Figure 4b. How- The changes in fracture length with time were cal-
ever, the fractures were contained in the M2M and culated by conducting the image analysis, as shown
M2H cases. In essence, the fractures did not penetrate in Figure 5. In the analysis, the fracture length was
the layer interfaces, but instead, they either propagated summed if bi-wings developed. The growth rate of
along the interfaces or stopped propagating. This trend fracture was very low initially for the first couple of
supports the results from Simonson et al. (1978). The seconds, but it gradually increased and then became
pressure curves for layered cases are shown in Fig- constant. As shown in Figure 5a, the fracture propa-
ure 4a. The middle layers of gelatin plates were made gated faster in the softer gelatin. The fracture lengths in
with the M-gelatin; thus, the pressure curves were in the layered cases were plotted with the M-gelatin case
the similar shape with the homogenous M-gelatin case in Figure 5b. For the M2L case, as the fracture crossed
until the fracture approached the interfaces. The points from M-gelatin through the interface to L-gelatin, the
A, D, and G in Figure 4a denote fracture initiation growth rate significantly increased. For the M2M case,
points, and they have the similar values. as the fracture developed along the interface, the rate
The pressure responses varied with the stiffness of of propagation also increased. For the M2H case, as the
bounding layers. In the M2L case, the pressure dropped first fracture grew and approached the bounding layer,
significantly at point B in Figure 4a as soon as the the growth rate diminished followed by a stagnant
hydraulic fracture passed the interface because the phase where further growth was confined. Then, the
fracture moved to the softer layer. The pressure also second fracture was generated and the length increased
gradually drops in the M2M case when the fracture step-wisely. It is noteworthy that the growth rate of
met the bounding layer and followed the interface. In the first fracture was slightly slower than that of the
the M2H case, the fracture behavior was quite differ- second fracture. Comparing to the pressure curve, the
ent from others. One fracture initiated first within the pressure was built up until the second fracture growth
219
Figure 6. Fracture volume prediction from injection rate
and experiment data for H case.
3.4 Fracture volume

The fracture volume was estimated by multiplying
the fracture height (10 mm) by the projected fracture
area that is the maximum width of the fracture. The
fracture area was obtained from the image analysis,
and two extreme cross sections of the fracture were
assumed: a rectangular section and a diamond-shaped
section. Therefore, for a given fracture width, the frac-
ture volume of the rectangular section is twice that
of the diamond-shaped section, as shown in Figure 6.
Also, the fluid injection rate i.e., 10 mL/min was super-
imposed in Figure 6. The slope of the curve for the
diamond-shaped section appears to be similar to the
fluid injection curve. This implies that the actual cross
section of the generated fractures was likely close to
diamond-shaped. If the diamond-shape is unrealistic
then it is expected to have a barrel shape with the
maximum width at the middle of the section.
4 CONCLUSIONS
Figure 5. Fracture length comparison for (a) homogeneous This study explored the hydraulic fracture behaviors in
cases and (b) layered cases. homogeneous and layered media using the time-lapsed
imaging of fracture propagation in gelatin plates. It
was found that the pressure at the fracture initiation
point and the fracture growth rate increased with the
began. Then, as the second fracture grew, the elevated gelatin stiffness. The growth rate of hydraulic frac-
pressure was suddenly released. This sudden pressure ture increased as the gelatin stiffness decreased. The
release caused the faster growth of the second frac- actual cross section of the generated fractures appears
ture than that of the first. Thereafter a stagnant phase, more likely to be in a barrel shape. For layered cases,
the second fracture grew along the interface with the the propagation behavior of fracture was affected by
increased growth rate while the first fracture stopped the stiffness of bounding layers. When the stiffness
growing. of the bounding layer is lower than that with the
220
fracture, the fracture passed through the interface, rheology and confining stress. Journal of Petroleum
followed by the increased growth rate and rapidly Science and Engineering, 53(1), 1–12.
decreasing pressure. Whereas, the fracture propaga- Frash, L. P. (2007). Laboratory-scale study of hydraulic frac-
tion was confined and the pressure started building turing in heterogeneous media for enhanced geothermal
systems and general well stimulation.
up when it met a bounding layer having the same or Geertsma, J. & De Klerk, F. (1969).A rapid method of predict-
higher stiffness. But, as the fracture propagated along ing width and extent of hydraulically induced fractures.
the interface, the pressure dropped and it grew at a Journal of Petroleum Technology, 21(12), 1–571.
faster rate. This study presents the simple but unique Khristianovich, S. A. & Zheltov, Y. P. (1955). Formation of
experiment data on the interactions between hydraulic Vertical Fractures by Means of Highly Viscous Liquid.
fracture and interfaces, and can be further used to Fourth World Petroleum Congress, 579.
develop models for geometry prediction of hydraulic Nordgren, R. P. (1972). Propagation of a vertical hydraulic
fracture. fracture. Society of Petroleum Engineers Jour-nal, 12(04),
306–314.
Perkins, T. K. & Kern, L. R. (1961). Widths of hydraulic
fractures. Journal of Petroleum Technology, 13(09), 937–
ACKNOWLEDGEMENT 949.
Simonson, E. R., Abou-Sayed, A. S., & Clifton, R. J. (1978).
This research was supported by a grant(15SCIP- Containment of massive hydraulic frac-tures. Society of
B105148-01) from the Construction Technology Petroleum Engineers Journal, 18(01), 27–32.
Research Program funded by the Ministry of Land, Teufel, L. W. & Clark, J. A. (1984). Hydraulic fracture propa-
Infrastructure, and Transport of the Korean govern- gation in layered rock: experimental studies of fracture
ment and by a grant(13SCIPS04) from Smart Civil containment. Society of Petroleum Engineers Journal,
Infrastructure Research Program funded by Ministry 24(01), 19–32.
Warpinski, N. R. (2011). Measurements and observations of
of Land, Infrastructure and Transport(MOLIT) of fracture height growth. In US EPA Technical Workshop for
Korea government and Korea Agency for Infrastruc- the Hydraulic Fracturing Study: Chemical & Analytical
ture Technology Advancement(KAIA). Methods, Arlington, Virginia.
Lide, D. R. (2000). CRC Handbook of Chemistry and Physics,
edited, CRC Press, Boca Raton, Florida.
REFERENCES
Bohloli, B., & De Pater, C. J. (2006). Experimental study
on hydraulic fracturing of soft rocks: Influence of fluid
221
A parametric sensitivity study on CO2 injection in deep saline aquifers

accounting for hydro-mechanical microfracturing
Alexandru Bogdan Tatomir

Department of Applied Geology, University of Göttingen, Göttingen, Germany
Ingrid Tomac
Department of Structural Engineering, University of California San Diego, USA
Martin Sauter
Department of Applied Geology, University of Göttingen, Göttingen, Germany
ABSTRACT: CO2 -injection in saline reservoirs requires studying of the geomechanical responses to estimate
the mechanical integrity of the reservoir- and cap-rock. Pressure build-up depends on the permeability and
compressibility of the reservoir rocks and fluids. Additionally, CO2 injection-induced pressure increase can
enhance the pre-existing micro-fractures and/or lead to the formation of new cracks. The current study evaluates
the sensitivity of critical parameters for the multiphase CO2 and brine flow in a fractured porous system. Two
numerical simulators are used to solve the CO2 -injection processes at micro- and field-scale.The PFC2D simulator
based on the Discrete Element Method (DEM) is used to model micro-scale stress induced damage, while 2p2c-
DuMux model is used to simulate the multiphase (CO2 -brine) flow in the porous system. This paper demonstrates
the capability of the linked approaches with 2p2c-DuMux -DEM for one case study on the injection of supercritical
CO2 at a pilot site located at Heletz, Israel. First, a parameter sensitivity study on the CO2 injection pressure on
the damage of the reservoir sandstone and caprock is performed. Further, the hydro-mechanical effects of current
pilot scale (10,000 t) CO2 injection and those of an industrial scale (>100,000 t) are compared. The results show
that the sandstone reservoir rock, at in-situ stress and pore-pressures, is likely to suffer failure, which changes
the effective flow properties affecting the plume migration dynamics. The results improve the understanding of
the implications of industrial scale GSC.
Keywords: geological carbon dioxide storage (GCS), deep saline aquifers, discrete element method, micro-
fracturing, permeability change
1 INTRODUCTION (e.g., Luo and Bryant, 2014), while the mechanical

stresses are caused by increased reservoir pressure
Carbon capture and storage (CCS) in geological for- (Hou et al., 2012; Rutqvist et al., 2008). As a result,
mations, or geological carbon storage (GCS), repre- irreversible mechanical changes can create new frac-
sents a viable solution for reducing the increasing tures or reactivate old ones, which constitute leakage
carbon dioxide (CO2) levels in the atmosphere (Bachu, pathways for the CO2.
2015; Celia et al., 2015; Cook et al., 2014; IPCC, The elastic and strength properties of the reservoir
2005). Sandstone rock represents the lithology of rock and induced micro-cracks define the hydro-
approximately 50% of the small-scale GCS injection mechanical behaviour of the reservoir. Brittle and/or
projects (less than 100,000 tonnes of injected CO2), weakly consolidated sandstone reservoir rocks tend to
respectively, 75% of the large scale injection projects fail under the tensile-shear failure mechanism induced
(Cook et al., 2014). by CO2 injection. Real-time crack coalescence pro-
The CO2 injection rates must be selected appropri- cess in sandstone rock has been investigated in several
ately in order to satisfy the economic and operational experimental studies, e.g., (Yang et al., 2015a) and
considerations. However, the rate of injection should numerically using a two-dimensional particle flow
not affect the stability of the system (Eshiet and Sheng, code (PFC2D) (Yang et al., 2015b). Permeability
2013). Large masses of injected CO2 can induce in- changes during rock damage and cracking have been
situ stresses and deformations in the reservoir and the investigated through a number of laboratory experi-
cap rock, which can cause changes of the hydraulic ments (Hu et al., 2010) and studies (Homand-Etienne
properties (Hu et al., 2010). Cooling down of the stor- et al., 1998; Levasseur et al., 2013; Shao et al., 2005).
age formation leads to temperature induced stresses Yang et al., (2015a) define a relationship between
223
permeability and deformation, identifying five phases. over years (Liu et al., 2011). Rathnaweera et al. (2016)
A recent review of advancement of permeability evolu- found that the significant rock mass mineralogical
tion models for fractured porous media is given in (Ma, structure alterations enhance the permeability of the
2015). Fluid permeability of sedimentary rocks in a aquifer by the long term CO2 reaction. The pore struc-
complete stress-strain process is discussed in (Wang ture changes caused by the CO2 reaction also affect
and Park, 2002). the effective stress response of the aquifer rock mass
Standard macroscopic multiphase multi-component (Rathnaweera et al., 2016).
models usually do not take into account the changes In this study two numerical simulators are used to
in hydraulic parameters (e.g., permeability) occurring solve the CO2 -injection processes at micro- and field-
during the injection of fluid (Class et al., 2009, 2002; scale. The PFC2d simulator based on the Discrete
Nordbotten et al., 2012; Tatomir et al., 2011). Never- Element Method (DEM) is used to model micro-
theless, there are several coupled numerical models scale stress induced damage, while the two-phase
to account for thermos-hydro-mechanical processes two-component (2p2c) model within the numerical
under multi-phase conditions (e.g., Hou et al., 2012; toolbox DuMux is used to simulate the field-scale
Rutqvist et al., 2008, 2002; Rutqvist and Tsang, 2002). multiphase/multi-component (CO2 -brine) flow and
The literature review has shown that the rock transport in the (fractured) porous system.
permeability evolution is directly related to the dis- The objective of this paper is to investigate the capa-
tribution, opening and coalescence of induced micro- bility of the linked approaches with DuMux -DEM for
cracks. Furthermore, the determination of hydraulic one case study on the injection of supercritical CO2 at
and poroelastic properties with the evolution of rock a pilot site located at Heletz, Israel. First, a parame-
damage is a challenging research topic. ter sensitivity study on the CO2 injection pressure on
Numerical and theoretical work has been performed the damage of the reservoir sandstone and caprock is
to study damage of the basement rock and caprock performed. Further, the hydro-mechanical effects of
during a CO2 injection by several researchers. Rutqvist current pilot scale (10,000 t) CO2 injection and those
and Tsang (2005) used coupled TOUGH2-FLAC3D of an industrial scale (>100,000 t) are compared.
computer codes to study hydromechanical changes in
a caprock during CO2 disposal in brine formations.
They obtained a general reduction in the effective mean 2 METHODOLOGY
stress in the lower part of the caprock, indicating a
possible rock failure. Shear reactivation seems to be 2.1 Simulation environment
more likely than the probability of fracturing and shear
under poro-elastic stresses induced by slow increase The two codes are coupled sequentially; the output of
in pressure during the injection period (Rutqvist, Wu, one code being used by the other. DuMux two-phase
Tsang, & Bodvarsson, 2002). Cerasi and Walle (2016) two-component model simulates the reservoir-scale
conducted tests on weak sandstone outcrop samples, supercritical CO2-brine flow and uses an input func-
in order to assess whether sequences of injection of a tion obtained from the DEM model, which relates local
pore fluid and shut-in could have a destabilizing effect porosity and permeability to the fluid pressure and the
on the borehole wall of the hollow cylinder specimen. degree of damage. The general simulation procedure
The tests showed no signs of fatigue weakening under is illustrated in Figure 1.
cycling injection, for low injection pressure and high A relationship between permeability change and
confinement of the sample (Cerasi and Walle, 2016). stress is determined with the PFC2D model to be used
Large CO2 injection over 10 000 years’ period was by the large(field)-scale 2p2c-Dumux simulator. How-
modeled by Liu et al. (2010). However, the emphasis ever, PFC2D model uses as first input the pressure
is given to hydro mechanical and chemical processes ranges attained in the reservoir due to CO2 injection,
calculated with no hydro-mechanical effects by the
2p2c model.
2.2 Discrete element method (DEM)

The DEM defines a system of particles that are repre-
sented by finite spherical or discs particles and walls in
PFC2D . The calculation cycle in PFC2D (Cundall and
Strack, 1979; Itasca, 2004) is a time-stepping algo-
rithm that consists of the repeated application of the
Law of Motion to each particle, a Force-displacement
Law to each contact, and a constant updating of wall
positions.The motions of particles and walls are solved
using the explicit finite difference scheme.
The Bonded-particle model (BPM) for rock is intro-
duced in PFC2D with the pre-scribed procedures that
Figure 1. Schematic representation of the modelling work- are used to determine the mechanical parameters of
flow DuMux – PFC2D . the synthetic rock mass (Potyondy, 2007; Potyondy
224
and Cundall, 2004). The parallel bond is a component processes in porous media build as a modular tool-
within the particle contact and can be pictured as a box. The partial differential equations for multiphase
cementitious material at the particles contact that is flow and transport are for solved with grid-based meth-
able to transfer forces and moments from one parti- ods such as the vertex-centered finite volume scheme
cle to another. Such a bond can be envisioned as a (box method, Helmig, 1997; Huber and Helmig, 2000).
set of elastic springs with constant normal and shear The implicit Euler method is applied for the temporal
stiffness uniformly distributed over either a circular discretization. The 2p2c model was benchmarked in
or rectangular cross-section lying on the contact plane several studies, e.g. (Class et al., 2009; Nordbotten
and centered at the contact point. After the parallel et al., 2012).
bond has been created between neighboring particles, The mathematical model on which 2p2c is con-
the relative motion at the contact causes normal and structed comprise the mass balance equations for each
shear stresses to develop within the bonded material, as phase and each component. The starting assumption
a result of the parallel-bond stiffness. If either of these is that the medium is a continuum where the extended
stresses exceeds its corresponding maximum strength, Darcy’s law (Bear, 1972; Helmig, 1997) is valid.
the parallel bond breaks. The parameters required to The resulting set of equations are closed with the
model the parallel bond are normal and shear strength constitutive relationships for capillary pressure and
and stiffness. knowing the sum of saturations is 1. A description of
Mechanical parameters of the resulting solid com- the basic mathematical model for 2p2c is given in for
prised of the bonded particles are modeled indirectly instance in (Flemisch et al., 2011), or in (Tatomir et al.,
by implementing an iterative procedure for obtaining 2015) an extension of 2p2c to include reactive tracer
the parallel bond parameters. For rock mass modeling, transport. The 2p2c standard model as provided in the
usually the Direct tension test, the Brazilian test, the free open-source repository (www.dumux.org) is fur-
unconfined compression tests (UCS) and the triaxial ther developed to account for the permeability changes
tests are used. induced by the pressure effects of CO2 injection.
A fracture can propagate within the bonded parti-
cles assembly by breaking the bonds between parti-
cles. For low porosity solids the flow pathways may 2.4 Heletz sandstone reservoir
be assumed to consist of parallel-plate channels at Heletz, Israel (Niemi et al., 2016; Tatomir et al., 2016)
contacts accompanied with artificial fluid reservoirs is the location for an onshore deep saline CO2 stor-
scheme for calculation (Figure 2). age pilot site. ‘Heletz sandstone’ is the building unit
The aperture of such a channel is proportional to the of the deep saline CO2 storage pilot site. The physi-
normal displacement at corresponding contacts. In the cal and geomechanical properties of Heletz sandstone
case of bonded material, the channel opening will not are given in (Edlmann et al., 2016; Elhami et al.,
increase from its initial value unless the bond is broken, 2016; Niemi et al., 2016; Tatomir et al., 2016). A
and the adjacent particles distance increases. Pressures series of destructive and non-destructive tests have
stored in fluid reservoirs are updated during the fluid been performed on core sample material showing that
calculation, and act on the surrounding particles as the reservoir sandstone is extremely weak with respect
equivalent forces. Each channel is a link between two to its depth of deposition (>1600 m) (Elhami et al.,
adjacent fluid reservoirs. As far as the fluid is con- 2016). The sandstone is poorly cemented and little
cerned, the channel is equivalent to a parallel-plate consolidated as shown by SEM images in Figure 3.
channel, with the length L, the aperture a, and the unit Small micro-cracks (1–5 mm long) are present acting
depth in the out-of-plane dimension. as preferential flow paths and influencing the hydraulic
behaviour.
2.3 Two phase two component (2p2c) simulator
(DuMux )
The two phase two component simulator isother-
mal (2p2c) model is developed in the DuMux
(www.dumux.org) numerical toolbox (Flemisch et al.,
2011). DuMux is a simulator for flow and transport
Figure 3. Heletz sandstone core sample analysis from CO2

Pilot Site. Thin section and SEM images showing the mineral
conglomerate of quartz grains interconnected by clayey and
carbonatic cement. Right hand side: µCT image revealing a
Figure 2. The BPM scheme with fluid channels and reser- micro-crack (after Tatomir et al., 2016).
voirs for hydraulic fracturing.
225
From permeameter tests on the sandstone sample Viscous fluid flows into the specimen at a constant
tested under triaxial compression it was found that flow rate on the left boundary, and the pressure dif-
the permeability is not a constant, but changes with ferences across the model are recorded at prescribed
the stress and strain in the rock. (Yang et al., 2015a) times during the test. The final pressure difference is
came to similar observations. However, is varying recorded following the test convergence, as shown in
between 50 and 700 mD. An accepted value obtained Figure 6.
from pumping tests is 400 mD which is going to The numerical relationship between the initial
be used in this study. The porosity is 23.2%. The average synthetic rock permeability and the model
Brooks-Corey (BC) parameters for capillary pressure-
saturation relationship are λ = 0,762 and entry pres- Table 1. Micro-mechanical properties of BPM,
sure Pe = 3861,2 Pa. These parameters are obtained Rmin = minimum particle radius; Rmax = maximum par-
from core experiments and showed high variability ticle radius; fi = particle friction coefficient; Pb_kn = parallel
(Niemi et al., 2016). bond normal stiffness; Pb_ks = parallel bond shear stiffness;
Pb_sstr = parallel bond shear strength; Pb_nstr = parallel bond
normal strength.
3 RESULTS Parameter Value
3.1 DEM sandstone model calibration Rmin (mm) 0.00025

Rmax /Rmin (−) 1.66
Text PFC2D synthetic sandstone mechanical properties fi (−) 0.5
are matched with the available data from the labora- γ (kg/m3 ) 2670.0
tory tests performed on the Heletz sandstone and for Pb_kn (GPa) 10.0
weakly bonded sandstones (Tatomir et al., 2016. In Pb_ks (GPa) 3.5
press; (Elhami et al., 2016; Haimson and Lee, 2004; Pb_sstr (MPa) 44.0 ± 15
Lin et al., 2009). The obtained micro and macro param- Pb_nstr (MPa) 10.0 ± 3.5
eters are shown in Tables 1 and 2. The boundary and
initial conditions, as well as, the CO2 parameters are
shown in Tables 3 and 4. Figure 4 shows the initial and Table 2. Macro-mechanical properties of BPM compared
boundary conditions. to the average granite and sandstone (Elhami, Ask, & Matts-
The permeability of the synthetic sandstone son, 2016; Haimson & Lee, 2004; Lin, Fakhimi, Haggerty, &
depends on the particle size and the initial aperture Labuz, 2009; Tatomir et al., 2016. In press).
of the fluid channels. Therefore, it is not possible to
directly set the initial permeability. Modeling perme- Tensile Young’s Fracture Poisson’s
ability laboratory tests permits obtaining an average Setting Strength Modulus Toughness Ratio
DEM model permeability. Simulation of permeability σt E KIC √ ν
(MPa) (GPa) (MPa m) (−)
tests in PFC2D uses the constant flow technique, which
is suitable for testing low permeability geo-materials
PFC2D 3.1 31.9 0.24 0.23
(Nakajima, Takeda, & Zhang, 2007; Olsen, 1966). The Sandstone
Darcy flow equation is used for determining the aver- Heletz 1.8–4.4 24.0–32.0 0.28 0.22
age permeability of the tested rock sample with the
prescribed initial width of fluid flow channels. Dur-
ing cycling, the bond-breakage is disabled by setting
the bond strengths unrealistically high. The no bond- Table 3. Boundary conditions of the model, where x is hori-
breakage model enables use of high flow rates and low zontal axes and y is vertical axes in the model (representation
of two dimensional field conditions, x is in the direction of
fluid dynamic viscosities, which aids in computational the fluid flow and horizontal, y is vertical).
time saving. The tested rock sample has dimensions
0.015 × 0.015 m (Figure 5). Parameter Magnitude
σmin (MPa) 16.4

σmax (MPa) 1.66
pp (MPa) 15.7
Table 4. Fluid injection properties (single-phase flow).
Parameter Magnitude
µ (Pa·s) 52.64·10−6
ρ (kg/m3 ) 662.08
Figure 4. Initial, boundary conditions and the DEM model
E (MPa) 32.26
size for lightly cemented sandstone.
226
Figure 8. Gas flow into the saturated sample, where
the left hand boundary pressure is a) Pg = 17,0 MPa, b)
Figure 5. Initial and final DEM model snapshot for perme- Pg = 20,0 MPa, c) Pg = 25,0, MPa, d) Pg = 30,0 MPa and
ability measurement. reservoir initial pore pressure is Pp = 15,6 MPa.
Figure 6. Initial permeability testing example of the syn-

thetic sandstone without bond breakage.
Figure 9. Pressure profiles across the 50 cm wide sample at

early times (t1 = 1377 s, t2 = 2760 s) for the left-hand fixed
pressures Pg = 17.0 MPa, Pg = 20.0 MPa, Pg = 25.0 MPa
and Pg = 30.0 MPa.
Figure 4 (maximum compressive stress is in the verti-

cal direction σv,max = 42.7 MPa, minimum horizontal
compressive stress σv,min = 15.7 MPa). The horizontal
Figure 7. The numerical relationship between the initial in-situ total stress is estimated based on the earth pres-
average synthetic rock permeability and the initial fluid sure at rest K0 = 1 − sin φ, with the internal sandstone
channel aperture in the model.
friction angle of φ = 38◦ . The in-situ stresses are cal-
culated based on the total unit weight of the sandstone
parameter of initial fluid flow channel is estab- material at the depth of h = 1600 m. The initial pore
lished and shown in Figure 7. Initial synthetic pressure is obtained using the unit weight of the water
rock permeability is calibrated for published val- and depth h = 1600 m, and is Pp = 15.6 MPa.
ues of average grain size and measured permeability The wellbore is assumed to be on the left side of
(k = 3.95 · 10−13 m2 = 400 mD) of the Heletz sand- the model. Green colored area represents the saturated
stone (Tatomir et al., 2016. In Press; Elhami et al., sandstone, while blue lines are broken parallel bonds
2016). between DEM particles. The broken bonds denote the
micro-cracks and damage caused by infiltration of
pressurized gas into rock. The simulations were estab-
lished for a small time, in order to observe the effect
3.2 Sandstone permeability and gas pressure
of pressure to rock damage.
relationship DEM modeling
Figure 9 accompanies the Figure 8, where the actual
The DEM synthetic sandstone model was used to pressure profiles are plotted across the width of the
investigate effects of gas pressurization on micro- model at times 22.95 and 45.9 min after beginning of
mechanical damage and permeability evolution. Fig- injection. It can be observed that differences in the
ure 8a-d shows the vertical two dimensional section pressure profiles are indeed recognizable.
of the sandstone near the wellbore, with dimensions The area where micro cracks occurred due to
50 × 50 cm. Boundary far-field total stress field in gas pressurization is isolated for permeability mea-
imposed to the model are the same as specified in surements. First, a set of new significantly smaller
227
Figure 10. Figure 7. Micro cracks developed in the biaxi-
ally compressed sample with initial dimensions 15 × 15 cm,
where σv,max = 42.7 MPa and σh,min = 15,7 MPa, for gas
pore pressures a) Pg = 17.0 MPa, b) Pg = 20.0 MPa and c)
Pg = 25.0 MPa.
Figure 12. a) CO2 pressure spatial distribution after 10 hour

of injection; b) CO2 Saturation spatial distribution after 40
days of injection.
Figure 11. Horizontal permeability change in the

biaxially confined sample, where σv,max = 42.7 MPa
and σh,min = 15.7 MPa, for gas pore pressures Pg =
17.0–25.0 MPa (kin = 3.95 · 10−13 m2 = 400 mD).
synthetic sandstone models are subjected to the same

initial and boundary conditions. Second, the pore pres-
sure is increased thorough the entire model which is
cycled for some time to reach equilibrium. The devel-
opment of micro cracks is evident for even small
pore pressure increase, indicating breakage of bonds
between DEM particles or sandstone grains. Figure
10 shows micro cracks (blue lines) in the pressurized
models. It can be also observed that the higher pore
pressure in Figure 10c causes larger deformation of the
model in the horizontal direction with the minimum
far-field confinement.
After obtaining the equilibrium of the model, the
permeability is measured in the horizontal direction.
The obtained relationship between the gas pressure
and the new average model horizontal permeability
is shown in Figure 11.
Figure 13. Comparison between static and dynamic –
hydro-mechanical (HM) models expressed as parameter vari-
3.3 Field scale CO2 migration ation with time at location x = 50 m, y = 50 m from the
injection well on top of the reservoir (z = 15 m): a) CO2
Following the DEM micro-scale models, the forward saturation; b) CO2 pressure; c) change in permeability.
simulation of CO2 injection at the field-scale. Due to
its symmetry the domain size of the modeled reservoir
is reduced to a 400 m by 400 m (with 15 m thickness), function (hydro-mechanical model) and another one
which only one quarter of the entire domain. with no permeability change (static model).The results
The first (static) field-scale CO2 saturation and are depicted in Figure 13. It can be observed that the
pressure profiles resulted from 2p2c model with a CO2 front does suffer almost no change. However, the
constant injection rate of 10 t/h are shown in Fig- changes in non-wetting phase permeability are visible
ure 12a) and b) and the corresponding breakthrough even though they are quite small (16.15 MPa in the
curves (non-wetting saturation – Sn and pressure – pn) HM model and 16.05 in the static model). Permeabil-
in Figure 14. The influence of the hydro-mechanical ity changes are plotted in Figure 13c. As expected the
effects on the reservoir performance is identified by static model has no change due to pressure whereas
comparing a numerical model using the permeability the HM model increases from 4e-13 to 5e-13 m2 .
228
Bear, J., 1972. Dynamics of fluids in porous media. American
Elsevier, New York. London and Amsterdam.
Celia, M.A., Bachu, S., Nordbotten, J.M., Bandilla, K.W.,
2015. Status of CO2 storage in deep saline aquifers with
emphasis on modeling approaches and practical simu-
lations – Celia – 2015 – Water Resources Research –
Wiley Online Library. Water Resources Research 51.
doi:10.1002/ 2015WR017609
Figure 14. Breakthrough curves of CO2 and pressure in the Cerasi, P., Walle, L.E., 2016. Investigation of Potential Loss
location of the injection well. of Injectivity and Formation Integrity Due to Pressure
Cycling. Energy Procedia, The 8th Trondheim Confer-
ence on CO2 Capture, Transport and Storage 86, 420–431.
doi:10.1016/j.egypro.2016.01.043
Class, H., Ebigbo, A., Helmig, R., Dahle, H.K., Nordbotten,
J.M., Celia, M.A., Audigane, P., Darcis, M., Ennis-King,
J., Fan, Y., Flemisch, B., Gasda, S.E., Jin, M., Krug, S.,
Labregere, D., Naderi Beni, A., Pawar, R.J., Sbai, A.,
Thomas, S.G., Trenty, L., Wei, L., 2009. A benchmark
study on problems related to CO2 storage in geologic for-
mations?: Summary and discussion of the results (Original
paper). Computational geosciences 13, 409–434.
Class, H., Helmig, R., Bastian, P., 2002. Numerical sim-
ulation of non-isothermal multiphase multicomponent
processes in porous media.: 1. An efficient solution
technique. Advances in Water Resources 25, 533–550.
doi:10.1016/S0309-1708(02)00014-3
Cook, P., Causebrook, R., Gale, J., Michel, K., Watson, M.,
2014. What Have We Learned from Small-scale Injection
Figure 15. Comparison between CO2 plume profile at Projects? Energy Procedia, 12th International Conference
10,000 t (top) and 100,000 t (bottom) injected. on Greenhouse Gas Control Technologies, GHGT-12 63,
6129–6140. doi:10.1016/j.egypro.2014.11.645
4 CONCLUSIONS Cundall, P.A., Strack, O.D.L., 1979. A discrete numerical
model for granular assemblies. Géotechnique 29, 47–65.
doi:10.1680/geot.1979.29.1.47
In spite of the fact that lot of research has been done Edlmann, K., Niemi, A., Bensabat, J., Haszeldine,
on chemical damage of sandstone formations dur- R.S., McDermott, C.I., 2016. Mineralogical proper-
ing CO2 injections, relatively little effort has been ties of the caprock and reservoir sandstone of the
dedicated to better understanding of the mechanical Heletz field scale experimental CO2 injection site,
damage in weak sandstones due to high gas pressure Israel; and their initial sensitivity to CO2 injec-
during injection. This paper shows results of DEM tion. International Journal of Greenhouse Gas Control.
modeling of Heletz sandstone samples which are sub- doi:10.1016/j.ijggc.2016.01.003
jected to different pore pressure increases due to CO2 Elhami, E., Ask, M., Mattsson, H., 2016. Physical-
and geomechanical properties of a drill core sam-
injection. It was found that increased pore pressures
ple from 1.6 km depth at the Heletz site in Israel:
cause mechanical damage in forms of breaking bonds Some implications for reservoir rock and CO2 stor-
between sandstone and causing increase in formation age. International Journal of Greenhouse Gas Control.
permeability. Significant micro cracking was accom- doi:10.1016/j.ijggc.2016.01.006
panied with deformation of the samples in the direction Eshiet, K.I.-I., Sheng, Y., 2013. Carbon dioxide injection and
of minimum in-situ stress at high pressures. associated hydraulic fracturing of reservoir formations.
Environ Earth Sci 72, 1011–1024. doi:10.1007/s12665-
013-3018-3
Flemisch, B., Darcis, M., Erbertseder, K., Faigle, B., Lauser,
ACKNOWLEDGEMENTS A., Mosthaf, K., Müthing, S., Nuske, P., Tatomir, A.,
Wolff, M., Helmig, R., 2011. DuMux: DUNE for multi-
This research has received funding from the European {phase, component, scale, physics, …} flow and trans-
Community’s 7th Framework Programme through the port in porous media. Advances in Water Resources 34,
TRUST (Grant agreement no. 309067) project. 1102–1112. doi:10.1016/j.advwatres.2011.03.007
Haimson, B., Lee, H., 2004. Borehole breakouts and com-
paction bands in two high-porosity sandstones. Interna-
REFERENCES tional Journal of Rock Mechanics and Mining Sciences
41, 287–301. doi:10.1016/j.ijrmms.2003.09.001
Bachu, S., 2015. Review of CO2 storage efficiency in Helmig, R., 1997. Multiphase Flow and Transport Processes
deep saline aquifers. International Journal of Green- in the Subsurface: A Contribution to the Modeling of
house Gas Control, Special Issue commemorating Hydrosystems, 1st ed. Springer.
the 10th year anniversary of the publication of the Hingerl, F.F., Yang, F., Pini, R., Xiao, X., Toney, M.F.,
Intergovernmental Panel on Climate Change Special Liu, Y., Benson, S.M., 2016. Characterization of het-
Report on CO2 Capture and Storage 40, 188–202. erogeneity in the Heletz sandstone from core to pore
doi:10.1016/j.ijggc.2015.01.007 scale and quantification of its impact on multi-phase
229
flow. International Journal of Greenhouse Gas Control. of Rock Mechanics and Mining Sciences 44, 677–691.
doi:10.1016/j.ijggc.2015.12.037 doi:10.1016/j.ijrmms.2006.10.002
Homand-Etienne, F., Hoxha, D., Shao, J.F., 1998. A Potyondy, D.O., Cundall, P.A., 2004. A bonded-particle
continuum damage constitutive law for brittle rocks. model for rock. International Journal of Rock Mechan-
Computers and Geotechnics 22, 135–151. doi:10.1016/ ics and Mining Sciences, Rock Mechanics Results from
S0266-352X(98)00003-2 the Underground Research Laboratory, Canada 41, 1329–
Hou, Z., Gou, Y., Taron, J., Gorke, U.J., Kolditz, O., 2012. 1364. doi:10.1016/j.ijrmms.2004.09.011
Thermo-hydro-mechanical modeling of carbon dioxide Rathnaweera, T.D., Ranjith, P.G., Perera, M.S.A., 2016.
injection for enhanced gas-recovery (CO2-EGR): a bench- Experimental investigation of geochemical and miner-
marking study for code comparison. Environ Earth Sci 67, alogical effects of CO2 sequestration on flow character-
549–561. doi:10.1007/s12665-012-1703-2 istics of reservoir rock in deep saline aquifers. Sci Rep 6,
Huber, R., Helmig, R., 2000. Node-centered finite volume 19362. doi:10.1038/srep19362
discretizations for the numerical simulation of multiphase Rutqvist, J., Birkholzer, J.T., Tsang, C.-F., 2008. Coupled
flow in heterogeneous porous media. Computational Geo- reservoir–geomechanical analysis of the potential for
sciences 4, 141–164. doi:10.1023/A:1011559916309 tensile and shear failure associated with CO2 injection
Hu, D.W., Zhou, H., Zhang, F., Shao, J.F., 2010. in multilayered reservoir–caprock systems. International
Evolution of poroelastic properties and permeabil- Journal of Rock Mechanics and Mining Sciences 45,
ity in damaged sandstone. International Journal of 132–143. doi:10.1016/j.ijrmms.2007.04.006
Rock Mechanics and Mining Sciences 47, 962–973. Rutqvist, J., Tsang, C.-F., 2002. A study of caprock
doi:10.1016/j.ijrmms.2010.06.007 hydromechanical changes associated with CO2-injection
IPCC, 2005. IPCC special report on carbon dioxide cap- into a brine formation. Env Geol 42, 296–305.
ture and storage. Metz, B., Davidson, O., de Coninck, doi:10.1007/s00254-001-0499-2
H.C., Loos, M., Meyer, L.A. (Editors): Prepared by Work- Rutqvist, J., Wu, Y.-S., Tsang, C.-F., Bodvarsson, G., 2002.
ing Group III of the Intergovernmental Panel on Climate A modeling approach for analysis of coupled multiphase
Change. Cambridge University Press. fluid flow, heat transfer, and deformation in fractured
Itasca, 2004. PFC user manual: Minneapolis: Itasca Consult- porous rock. International Journal of Rock Mechanics and
ing Group Inc. Mining Sciences, Numerical Methods in Rock Mechanics
Levasseur, S., Collin, F., Charlier, R., Kondo, D., 2013. A 39, 429–442. doi:10.1016/S1365-1609(02)00022-9
micro–macro approach of permeability evolution in rocks Shao, J.F., Zhou, H., Chau, K.T., 2005. Coupling between
excavation damaged zones. Computers and Geotechnics anisotropic damage and permeability variation in brittle
49, 245–252. doi:10.1016/j.compgeo.2012.12.001 rocks. Int. J. Numer. Anal. Meth. Geomech. 29, 1231–
Lin, Q., Fakhimi, A., Haggerty, M., Labuz, J.F., 2009. 1247. doi:10.1002/nag.457
Initiation of tensile and mixed-mode fracture in sand- Tatomir, A.B.A., Halisch, M., Duschl, F., Peche, A., Wie-
stone. International Journal of Rock Mechanics and Min- gand, B., Schaffer, M., Licha, T., Niemi, A., Bensabat,
ing Sciences 46, 489–497. doi:10.1016/j.ijrmms.2008. J., Sauter, M., 2016. An integrated core-based analysis for
10.008 the characterization of flow, transport and mineralogical
Liu, F., Lu, P., Zhu, C., Xiao, Y., 2011. Coupled reactive parameters of the Heletz pilot CO2 storage site reser-
flow and transport modeling of CO2 sequestration in the voir. International Journal of Greenhouse Gas Control.
Mt. Simon sandstone formation, Midwest U.S.A. Inter- doi:10.1016/j.ijggc.2016.01.030
national Journal of Greenhouse Gas Control 5, 294–307. Tatomir, A.B., Schaffer, M., Kissinger, A., Hommel, J.,
doi:10.1016/j.ijggc.2010.08.008 Nuske, P., Licha, T., Helmig, R., Sauter, M., 2015.
Luo, Z., Bryant, S., 2014. Impacts of Injection Induced Frac- Novel approach for modeling kinetic interface-sensitive
tures Propagation in CO2 Geological Sequestration – Is (KIS) tracers with respect to time-dependent interfa-
Fracturing Good or Bad for CO2 Sequestration. Energy cial area change for the optimization of supercritical
Procedia, 12th International Conference on Greenhouse carbon dioxide injection into deep saline aquifers. Inter-
Gas Control Technologies, GHGT-12 63, 5394–5407. national Journal of Greenhouse Gas Control 33, 145–153.
doi:10.1016/j.egypro.2014.11.570 doi:10.1016/j.ijggc.2014.11.020
Ma, J., 2015. Review of permeability evolution model for Tatomir, A.B., Szymkiewicz, A., Class, H., Helmig, R., 2011.
fractured porous media. Journal of Rock Mechanics Modeling two phase flow in large scale fractured porous
and Geotechnical Engineering 7, 351–357. doi:10.1016/ media with an extended multiple interacting continua
j.jrmge.2014.12.003 method. CMES - Computer Modeling in Engineering and
Niemi, A., Bensabat, J., Shtivelman, V., Edlmann, K., Gouze, Sciences 77, 81–111.
P., Luquot, L., Hingerl, F., Benson, S.M., Pezard, P.A., Wang, J.-A., Park, H.D., 2002. Fluid permeability of sed-
Rasmusson, K., Liang, T., Fagerlund, F., Gendler, M., imentary rocks in a complete stress–strain process.
Goldberg, I., Tatomir, A., Lange, T., Sauter, M., Freifeld, Engineering Geology 63, 291–300. doi:10.1016/S0013-
B., 2016. Heletz experimental site overview, characteri- 7952(01)00088-6
zation and data analysis for CO2 injection and geological Yang, S.-Q., Huang, Y.-H., Jiao, Y.-Y., Zeng, W., Yu, Q.-L.,
storage. International Journal of Greenhouse Gas Control. 2015a. An experimental study on seepage behavior of
doi:10.1016/j.ijggc.2015.12.030 sandstone material with different gas pressures. Acta
Nordbotten, J.M., Flemisch, B., Gasda, S.E., Nilsen, H.M., Mech. Sin. 31, 837–844. doi:10.1007/s10409-015-0432-7
Fan, Y., Pickup, G.E., Wiese, B., Celia, M.A., Dahle, Yang, S.-Q., Huang, Y.-H., Ranjith, P.G., Jiao, Y.-Y., Ji, J.,
H.K., Eigestad, G.T., Pruess, K., 2012. Uncertain- 2015b. Discrete element modeling on the crack evolu-
ties in practical simulation of CO2 storage. Interna- tion behavior of brittle sandstone containing three fissures
tional Journal of Greenhouse Gas Control 9, 234–242. under uniaxial compression.Acta Mech. Sin. 31, 871–889.
doi:10.1016/j.ijggc.2012.03.007 doi:10.1007/s10409-015-0444-3
Potyondy, D.O., 2007. Simulating stress corrosion with a
bonded-particle model for rock. International Journal
230
Experimental study of proppant particle-particle interaction micromechanics

during flow and transport in the fracture
Ingrid Tomac & Lan Luo

University of California, San Diego, San Diego, USA
ABSTRACT: This paper presents analysis experimental study of the proppant behavior during settling in a
narrow smooth fracture. The objective of this study is to better understand particle interactions in a dense phase
slurry. The study focuses on identifying and observing proppant agglomerations in narrow fracture like slot in
the laboratory. The Geo Particle Image Velocimetry (GeoPIV) method was used for analyzing the movements of
proppants during settling on a particle scale. The proppant settling velocity depends on the relationship between
settlement, particle concentration and occurrence of particle agglomeration. In this study, the settling of single
particles was compared to the Stokes’ Law, as well as agglomerated particles. Due to the very small ratio of the
average proppant particle size and the slot aperture, deviations from the Stoke’s law increases significantly for the
case when particles agglomerate. However, the power law relationship between the settling and Stokes’ law ratio
and particle or agglomerate size was found to be valid. The significance of the visually observed agglomerates
in the slurry and their larger settling velocities indicate that the proppant initial volumetric concentration might
not be a good parameter for predicting the final slurry settling when significant agglomeration occurs.
1 INTRODUCTION Joseph (1994) focused on different particle interac-

tions or particle-wall interactions between Newtonian
Selection of both proppants and fracturing fluid affects and viscoelastic fluids. Joseph (1994) found the attrac-
the process and results of hydraulic fracturing. Prop- tion behavior of two particles in viscoelastic fluid but
pant is small granular material which keeps fractures the repulsion behavior in Newtonian fluid while the
open after fluid injection. A better understanding gap between the two particles is really small. Joseph
of micro-mechanics of proppant flow and transport (1994) also concluded that particles will be attracted
into hydraulic fracture has impact on both industrial by the wall in viscoelastic liquid but will be repelled in
and environmental applications. Many researchers Newtonian liquid as they are close to a vertical wall at
have studied proppant flow and transport in frac- the beginning. Malhotra & Sharma (2012) conducted
tures, but relatively little work has been done on an experimental study to qualify the wall factor for
micro-mechanical understanding of particle-particle shear thinning viscoelastic fluids. Malhotra & Sharma
interactions effects on the overall behavior of prop- (2012) kept the ratio of particle diameter to wall spac-
pant. Deng et al. (2014) studied the effects of proppant ing constant to find the retardation effect of confining
size,Young’s moduli and pressure levels on the size the walls due to the increasing of shearing thinning behav-
crack aperture. Hammond (1995) did numerical analy- ior of fluid. As for low proppant concentrations in flow
sis on the gravity-driven vertical motion of proppant in through horizontal or near horizontal pipes, Stevenson
a hydraulic fracture. Hammond (1995) concluded that et al. (2001) obtained a correlation for sand velocity
proppant settling processes significant contributed to while there is no sand movement by the intermittent
the proppant rearrangement on the time scale and the flow. Patankar et al. (2002) conducted both numer-
rearrangement was also severe at lower proppant con- ical and experimental studies on the sediment flow
centrations in homogeneous flows. Dontsov & Peirce and transport in pressure driven channel and obtained
(2014) analyzed the gravitational settling of spherical the power law correlations, which can be used as a
particles in steady flow. They introduced an approxi- basis for sediment transport models. Furthermore, to
mate solution to reduce the complexity of problem of enhance the design process of hydraulic fracturing, the
boundary conditions. Furthermore, Dontsov & Peirce migration process and the volumetric concentration of
(2015) extended to observe the tip screen- out effects proppant were also investigated by Zhao et al. (2008)
for proppant transport in hydraulic fractures while the by using numerical simulation. McClure et al. (2015,
ratio of facture width to diameter of sand particle 2016) also contributed to the implementing proppant
is small, the effect of the wall cannot be neglected. flow and transport in the numerical model to describe
231
propagation of hydraulic fractures and opening and shown in Figure 1. The plastic was clear enough to see
shear stimulation of natural fractures. They used non- through in order to let the proppants be recorded by
linear empirical equations are used to relate normal digital camera. One of the plastic plates which locates
stress, fracture opening, and fracture sliding to frac- on the back side had dark blue background in order to
ture aperture and transmissivity. Eskin & Miller (2008) enhance the color contrast. Two different tests with
presented governing equations which are composed of different viscosity fluids were conducted, one with
boundary conditions and constitutive relations for the 75% and other with 50% volumetric water-glycerol
proppant flow and transport model which takes the mixture.
micro-level particle dynamics into account. Eskin &
Miller (2008) concluded that the slurry dynamics is
governed by particle fluctuation in a high-shear-rate
3 EXPERIMENTAL METHODOLOGY
flow and that slurry flow in a fracture is charac-
AND ANALYSIS
terized by non-uniform solids concentration across
the fracture width. Roy et al. (2015) conducted both
The GeoPIV software was used for analysis of indi-
experimental and numerical analysis on the proppant
vidual particle movements in time. The GeoPIV uses
transport. The particles were separated as transparent
the principles of the Particle Image Velocimetry (PIV)
and opaque groups for easier tracking and finding out
method for obtaining the particle displacement data
different displacement behavior. They concluded that
from couple of digital images. Images can be extracted
the initially consolidated pack of soils at the top of
from camera directly when the time interval is suffi-
the cell tend to serve as a source of particles and it
cient, otherwise it should be extracted from videos if
will increase the duration and concentration of the set-
time interval is too small. Adrian (1991) first intro-
tling particle phase. In addition, in experiments with
duced the particle-imaging technique measure the
the lowest particle concentration particles ultimately
motion of small, marked regions of a fluid by observ-
had lowest settling velocity.
ing the locations of the images of the markers at two or
The objectives of the study presented in this paper
more times. Willert & Gharib (1991) further developed
is better understanding of the single particle and parti-
PIV method to make it easier and faster to handling
cle agglomeration settling rates in a 2 mm narrow slot.
of the whole series of operations. White (2001) used
Particularly, the agglomeration processes and agglom-
digital photography and PIV image processing to mea-
erate settling is compared to a single particle settling
sure displacements of partially obscured soils in area of
rate. An experimental setup was built at a small scale
high strain gradient. White (2002) presented a new sys-
in 20 × 40 cm slot between two parallel acrylic plates.
tem for deformation measuring in geotechnical tests
The proppants were stored up at one end of the closed
based on PIV method with improved the accuracy and
frame first and then the experiment was turned by 180◦
precision, adding the displacement array.
to let the proppants move downward. For recording
In this study, the digital image is first divided into
the movements of proppants, a 60 frame/s camera was
a mesh as shown in Figure 2. The GeoPIV software is
placed in front of the transparent side of the frame.
used for analyzing subsequent pictures of the frames
In this paper, the GeoPIV method was used to trans-
obtained from the recorded video. Particle position
fer the digital figures to manageable data and particle
comparisons between two subsequent frames permits
displacement analysis.
obtaining displacement and particle velocities.
Figure 3 shows displacement arrays of tracked par-
2 EXPERIMENTAL SETUP ticles in a predefined mesh in Figure 2. The GeoPIV
method is chosen because it obtains the particle
The components of the experiment for this test are movement arrays and clearly shows the movement
the 2 mm wide slot made of acrylic plates, water and directions. The displacement data can be converted to
glycerol mixture, and 20/40 mesh sand as proppant. velocity, knowing the speed of the camera recording,
The 20 × 40 cm frame was used to hold two transpar- which is 60 frames per second.
ent plastic plates with 2 mm spacing between them, as
Figure 1. Experiments of narrow slot with proppants. Figure 2. Meshes of digital images.
232
Figure 3. Displacement arrays of tracked meshes.
Figure 6. Average velocity in y direction.
Figure 4. Particle group settling at low concentrations.
Figure 7. Average velocity in x direction.
In Figures 6 and 7 the average values of particle

settling velocities for each patch in Figure 4 are plot-
ted against time. Figure 6 shows the vertical particle
velocities in the direction y, and Figure 7 shows the
horizontal particle velocities in the direction x. As
Figure 5. Movement vectors for particle at low concentra-
shown in both Figure 6 and Figure 7, both of the
tions.
velocities of group of particles in x direction and in
y direction were larger at the beginning of the settling
which indicates that the movement at the beginning
4 ANALYSIS OF PARTICLES SETTLING is random and circular. After about 2 seconds, the
velocity in x direction almost disappears, which means
4.1 Particle settling 75% glycerol-water mixture that the movement can be treated as settling and the
Particle settling was analyzed in a 75% glycerol steady settling velocity is established. At that time,
water mixture, with the fluid dynamic viscosity is the movement of sand particles can be treated as
µ = 0.0355 Pa · s (Glycerine Producers’ Association, settlement.
1963). However, the movement of proppants is not sta-
ble during the entire test. Thus it is necessarily to find
out the particle movements for the whole period of 4.1.1 Velocity of a single particle
settlement to find out the stable period. The average During the process, there are both single particle and
velocities of the four patches are analyzed with loca- agglomerated particles settling in the fluid. Figure
tions shown in Figure 4. For analysis, the movement 8 shows the analyzed single particle with different
vectors are extracted for every time interval between diameters.
two subsequent frames. Figure 5 shows one example The settling velocity against size can be extracted
of the movement vectors for the first time interval. during the stable settling period as shown in Figure 9.
233
Figure 8. Different sizes of one single particle. Figure 10. Different sizes of agglomerated particles.
Figure 9. Velocities of one single particle in y direction vs. Figure 11. Velocities of agglomerated particles in y direc-
particle size. tion vs. particle size, compared with the Stoke’s law predic-
tion of sphere settling in unbounded fluid.
The expression of Stokes’ law which can determine
the terminal velocity of sphere falling in fluid is: agglomerates are observed with different shapes and
irregular sizes in fluid.
As shown in Figure 11, different sizes of agglom-
erated particles have different average velocities and
with larger diameter, the velocity becomes higher.
Comparing the data from the lab test with those from
where Vt is the flow settling velocity (m/s), g is the Stokes’ law, the difference becomes significant while
gravitational acceleration (m/s2 ), ρp is the particle the particle size increases. Similar like for the single
mass density (kg/m3 ), ρf is the fluid mass density particle, the wall effect plays a role in retarding the
(kg/m3 ) and µ is the fluid dynamic viscosity (Pa · s). particle settling velocities.
The proppant is medium 20/40 mesh sand with
approximate density of 2.65 kg/m3 . When the per-
centage of glycerol is 75 %, the density of fluid is
4.2 Particle settling 50% glycerol-water mixture
1198.45 kg/m3 and the dynamic viscosity is 0.0355
Pa · s. By using the Stokes’ law, the settling velocity of Conducting the same analysis as for the 75% glycerol-
ideal spherical particle with respect of particle sizes water solution, it is visible from the recorded video that
was presented in Figure 9. It can be seen that the veloc- the movements of proppants in 50% glycerol-water
ity extracted from the test is consistent with the results fluid are even faster than that in 75% glycerol-water
from Stokes’law, but with the increasing of the particle fluid. As shown in Figures 12 and 13, the velocities in
size, the velocity from tests did not change as much as both x direction and y direction are large, after around
the Stokes’ law would have predicted. If the wall effect 3 seconds the velocities becomes stable.
is considered, one of the conclusion is that when the Figure 14 shows particle vector arrays extracted
ratio of particle diameter with slot aperture is as small from the beginning of the test at 6 different times. It
as 0.2 or smaller, the results in tests can be consistent is reasonable that at the very first part of the test, the
with the prediction by Stokes’ law, otherwise the wall horizontal movement cannot be ignored to treat it as
effect cannot be ignored. settlement.
Figure 15 shows the displacement vectors at the end
4.1.2 Velocities of agglomerated particles part of the test. It shows that the vectors are approx-
In viscoelastic fluid, agglomerated groups of particles imate to be vertical. Thus after periods of time, the
are observed during settling besides single particles. movement of sands can be treated as settlement.
Figure 10 shows particle agglomerates with different
approximate diameters. The diameter approximation 4.2.1 Velocity of a single particle
is obtained for easier comparison of the agglomerate Similar with the settling in fluid with higer viscosity,
settling velocity with the Stoke’s law. However, the there are still both single particle and agglomerated
234
Figure 12. Average velocities in x direction.
Figure 15. Displacement vectors of particles.
Figure 16. Different sizes of one single particle.

Figure 13. Average velocities in y direction.
Figure 17. Velocities of one single particle in x direction vs.

particle size.
particle exsisting. Figure 14 shows pictures for anal-

ized single particles with different sizes, taken from
the camera.
The velocity of x direction shown in Figure 17
is relatively small compared with the velocities of y
direction in Figure 18, thus it is reasonable to treat the
movement of particles as settlement. It is obvious that
the velocity of settling varies with different sizes of
Figure 14. Displacement vectors of particles. particles.
235
Figure 18. Velocities of one single particle in y direction vs. Figure 20. Velocities of one agglomerated particles in y
particle size. direction vs. particle size.
Figure 19. Different sizes of agglomerated particles.
The density of the 50% glycerol solution is

1126.30 kg/m3 and the dynamic viscosity is 0.006
Pa · s. Compared the test data with the prediction by
Stokes’ law, it is also the same rule as in 75% glyc-
erol. The prediction only works when the particle size
is small compared to the slot size.
4.2.2 Velocities of agglomerated particles

Although in the 50% glycerol-water solution the vis- Figure 21. Ratio of settling velocity and Stokes’ law.
cosity is lower than that in 75% glycerol-water solu-
tion, there are still agglomerated particles. Figure 19
shows three selected agglomerates of particles in fluid. 5 CONCLUSION
The approximate diameter of the agglomerates was
estimated and shown in Figure 19, in order to obtain This paper shows analysis of experimental data of
the Stoke’s law settling velocity prediction. proppant settling in a narrow slot with smooth walls.
The settling velocity increases as the agglomerate The objective of the study was to better understand
size increases as shown in Figure 20. Comparing the occurrence and settling of single and agglomerated
test results with predictions from Stokes’law, it is obvi- particles in two glycerol-water solutions, with 75%
ous that with the increasing of the particle size, the and 50% volume concentrations. Settling velocities of
difference between practical results and the predictions particles were analyzed during predominantly verti-
is also becoming larger, indicating the importance of cal settling, in spite of the fact that observed motions
the wall effect. Comparing Figure 20 with Figure 18, it were also circular and erratic. The interruption at the
shows that the general velocity is faster if the particles beginning of the movements of proppants could not be
agglomerate with each other. ingnored. The main issue was that the average veloc-
Figure 21 summarizes the analyzed single and ities in x direction was almost the same as those in y
agglomerated particle settling velocities normalized direction, indicating that the proppant motion was not
with the Stoke’s law. The particle diameter and slot simple settling. After the proppants started to move
aperture ratio is plotted on the horizontal axis. It can as simple settling which means the velocities in x
be seen that the single particle settling velocities fol- direction can be ignored, the effects of particle size
low power law relationship in both glycerol solutions. compared to the slot diameter were found significant
However, the agglomerates settling velocities devi- in both solutions. Consequently, it was observed that
ate more from the idealized power law relationship particle settling velocities are similar to the Stokes’law
presented with the fit line in Figure 21. As the parti- prediction only when the ratio of particle diameter and
cle or agglomerate diameter increases, the wall effect slot aperture is as small as 0.1–0.2. The observed set-
becomes more significant and the ratio of the settling tling velocity of agglomerated particles was larger than
velocity and Stokes’ law decreases. single particles, where the deviation from the Stokes’
236
law was also bigger. However, it was possible to obtain Hammond, P.S. 1995. Settling and slumping in a Newtonian
a power-law relationship when plotting the ratio of slurry, and implications for proppant placement during
both single particle or different sizes of agglomerates hydraulic fracturing of gas wells. Chemical Engineering
and Stokes’ law versus the particle size. The experi- Science. 50: 3247–3260.
Joseph, D.D. 1994. Aggregation and dispersion of spheres
mental results identify the occurrence of agglomerated falling in viscoelastic liquids. Journal of Non-Newtonian
particles and prove that the settling rates of agglom- Fluid Mech. 54: 45–86.
erated particles are larger than single particles. In the Malhotra, S. & Sharma, M.M. 2012. Settling of spherical
performend experiments, it was visually observed that in unbounded and confined surfactant-based shear thin-
the agglomerates form significant portion of the slurry. ning viscoelastic fluids: An experimental study. Chemical
Future work will be focused on establishing the under- Engineering Science. 84: 645–655.
standing of how agglomerated particles influence the Patankar, N.A., Joseph, D.D., Wang, J., Barree, R.D., Conway,
average settling of the slurry and if the volumetric con- M. & Asadi, M. 2002. Power law correlations for sediment
centration of proppant, along with the fluid properties transport in pressure driven channel flows. International
Journal of Multiphase Flow. 28: 1269–1292.
and the particle-fracture size ratio solely can or cannot Roy, P., Wyatt, L. & Stuart, D. 2015. Proppant transport at
predict the slurry settling. the fracture scale: simulation and experiment. American
Rock Mechanics Association Conference. San Francisco,
28 June–1 July 2015.
REFERENCES Stevenson, P., Thorpe, R.B., Kennedy, J.E. & McDermott,
C. 2001. The transport of particles at low loading in
Adrian, R.J. 1991. Particle-imaging techniques for experi- near-horizontal pipes by intermittent flow. Chemical Engi-
mental fluid mechanics. Annual Reviews of Fluid Mechan- neering Science. 56: 2149–2159.
ics. 23: 26304. White, D.J., Take, W.A., Bolton, M.D. & Munachen, S.E.
Deng, S., Li, H., Ma, G., Huang, H. & Li, X. 2014. Simula- 2002. A deformation measurement system for geotech-
tion of shale–proppant interaction in hydraulic fracturing nical testing based on digital imaging, close-range pho-
by the discrete element method. International Journal of togrammetry, and PIV image analysis. 15th International
Rock Mechanics & Mining Sciences. 70: 219–228. Conference on Soil Mechanics and Geotechnical Engi-
Dontsov, E.V. & Peirce, A.P. 2014. Slurry flow, gravitational neering. Pp: 539–542. Balkema, Rotterdam.
settling and a proppant transport model for hydraulic White, D.J., Take, W.A. & Bolton, M.D. 2003. Soil deforma-
fractures. Journal of Fluid Mechanic. 760: 567–590. tion measurement using particle image velocimetry (PIV)
Dontsov, E.V. & Peirce, A.P. 2015. Proppant transport in and photogrammetry. Géotechnique. 53: 619–631.
hydraulic fracturing: Crack tip screen-out in KGD and Willert, C.E. & Gharib, M. 1991. Digital particle image
P3D models. International Journal of Solids and Struc- velocimetry. Experimental Fluids. 10: 181–193
tures. 63: 206–218. Zhao, Z., Cui, B.,Yue,Y., Wang, L. & Wu,Y. 2008. Numerical
Eskin, D. & Miller, M.J. 2008. A model of non-Newtonian simulation of horizontal migration of proppant. Journal of
slurry flow in a fracture. Powder Technology. 182: Hydrodynamics. 20: 74–80.
313–322.
237
Modeling of discrete intersecting discontinuities in rock mass using

XFEM/level set approach
Kamal C. Das, Sandeep S. Sandha, Eduardo Rodrigues & U. Mello

IBM Research
Ignacio Carol
ETSECCPB (School of Civil Engineering), UPC (Technical University of Catalonia), Barcelona, Spain
Pablo E. Vargas, Nubia A. González, J.M. SeguraSerra & M.R. Lakshmikantha

REPSOL CTR, Madrid, Spain
ABSTRACT: Modeling of discontinuities is of major importance to assess the geomechanical behavior of oil
and gas reservoirs. Traditionally, discrete discontinuities have been introduced in the numerical analysis using
continuum with modified constitutive laws (e.g. multi-laminate model (ML) or zero-thickness interface elements
(IE’s)). More recently, there have been several attempts to use extended finite element method (XFEM). The
development of an XFEM tool could lead to improved predictions for porosity/permeability changes in coupled
geomechanical reservoirs. Multiple intersecting faults are often the case in complex Reservoir and has been
explored less in the literature. In this work, we have presented a novel methodology based on XFEM to analyze
the behavior of multiple intersecting faults. Detailed mathematical framework has been derived by using the
concept of level-set and is implemented using distributed computing to solve complex geo-mechanical problems.
For validation purpose, we analyze the different examples and comparison them with standard finite element
method using zero-thickness IE’s (IE-FEM).
1 INTRODUCTION (for implicit representation methods), include: (a)

Extended finite element (XFEM) method (Belytschko
Mechanical discontinuities (fractures, faults, cracks et al., 2001; Fries and Belytschko, 2010; Das, 2013)
etc) affect the strength and deformability of the rock and, (b) Assumed enhanced strain (AES) method,
mass and may play a crucial role in the geome- which are reviewed and compared in (Borja, 2008;
chanical behaviour of reservoirs. Depending on the Das et al., 2015). ‘Enrichment’ signifies the addi-
procedure used to represent discontinuities, numerical tional problem specific functions for displacement
techniques can be classified into two different cat- approximation. For example, for rock faults modeling,
egories in the finite element literature. In the first, ‘Heaviside function’is used to enrich nodes (Watanabe
discontinuities are represented in a smeared fash- et al., 2012).
ion, by using constitutive equation for the continuum In Geomechanics, multiple discontinuities often
which incorporates also the additional deformations arise in rock medium. Many of these discontinuities
due to opening/slip. In the second one, discontinu- are arbitrary and can be intersecting each other in
ities are represented explicitly. This in turn may be the actual reservoir. To the authors’ knowledge, there
done in various ways, the more classical one is via are only a few studies that have applied the XFEM
zero-thickness (or sometimes thin layer) elements to multiple discrete geomechanical discontinuities in
inserted in-between element faces/edges while all 2D (Watanabe et al., 2012; Deb et al., 2015) and
nodal variables maintain the traditional meaning as 3D (Das et al., 2015). This paper develops the con-
regular displacements (Goodman et al., 1968; Gens cept of XFEM by incorporating arbitrary multiple
et al., 1988). In a second, more recent approach, intersecting rock faults in 3D. A standard weak for-
we do implicit representation of the discontinuities mulation method is used for developing the discrete
using level set functions. The discontinuity can cut the equations. A finite element can be defined as regular,
elements in arbitrary ways, which is captured using partially, singly or doubly enriched depending upon
zero values of level set function. This approach gives the number of discontinuities intersecting the element.
the flexibility to define the discontinuities with com- Based on enrichment, we have derived the element
plex geometries. An appropriate enrichment function stiffness matrices. The proposed procedure has been
is introduced near the discontinuity to capture the validated by solving different examples and their com-
jump in the displacement field. Enrichment techniques parison with standard FEM using zero-thickness IE’s
239
(Goodman, 1968). The examples show that the devel- 2.1 Discrete variational formulation
oped method produces results which are in strong
Consider a finite element interserted by two faults 1d
agreement with reference solutions.
and 2d of as shown in Figure 1(a) for 2D quadri-
lateral element. For representation, 1d and 2d are
2 MATHEMATICAL AND NUMERICAL assumed to be made up of triangles or quadrilaterals
SETTING in 2D, and tetrahedra or hexahedra in 3D. The total
number of elements in the partition will be denoted
The finite element setting is briefly summarized here. by ne and the total number of vertices by np . The dis-
First, the displacement field is obtained as the sum of crete form of problem (1) proceeds by considering a
a regular and discontinuous parts, such that, the vari- discrete space Vh ⊂ V for the displacement field. In
ational formulation for the problem then reads: Find this work, chosen displacement field belong to the
u ∈ V such that following space
∀ v ∈ V , where V = [H 1 ()]d is the space of vec- The space Vdi (Ti ) is not other than P1 (Ti ) for the case
d d
tor square integrable functions whose gradients are of triangular/tetrahedral meshes or Q1 (Ti ) for the
d
also square integrable. To capture the discontinuous case of quadrangular/hexahedral meshes. In this for-
behavior of the displacement field in the presence of j
mulation, we have: Vdi (Ai ) = { j∈J bj Md }, where
two intersecting discontinuities, 1d , 2d , we start by d
J is the set of vertices shared by elements in Ai
considering the following form for u. j
d
and Md are appropriate nodal enrichment functions.
Before detailing the enrichment functions, we present
the discrete variational formulation:
where ur is continuous and u1d , u2d , and u12 d ) ∈ Uh such that
Find (ur , u1d , u2d , u12
d are discon-
tinuous across 1d , 2d . For convenience we write the
discontinuous part as follows
where Mi = [Hi (x) − Hi (xi )] and Hi (x) is the
d d d d
Heavyside function (Hi (x) = 1 if x ∈ + , Hi
d d
(x) = 0 if x ∈ − ), whose distributional gradient is
∇Hi (x) = δi n̆, with δi the Dirac function and
d d d
Hi (xi ) is heaviside function value at xi . In (3), ũid
d
is continuous across i and ui (x) = ũi (x), x ∈ i ,
d d d d
where · denotes the jump of any quantity at id . Now,
we choose as weighting function
As for the traction law ti , a constitutive behav-
d
ior needs to be assumed for ti = Di · uid if uid · n ≤ 0
d d
for ∀ i = 1,2,12. In the next subsection, we describe
whose distributional symmetric gradient is given by constitutive laws which are used in this work
2.2 Enrichment functions

The displacement field on the standard elements is
written as
Denoting by Njr , j = 1, . . . , d + 1 the standard

The variational formulation can be obtained by con- P1 /Q1 basis functions for an element, we define as
sidering independent variations of vr , v1d , v2d and v12 usual
d
in the usual way. For the sake of brevity we ommit
the details here and focus on the discrete variational
problem.
240
2.3 XFEM element equations
For the sake of simplicity we restrict here the atten-
tion to the linear case. The element stiffness matrix is
written as
where the index r refers to the standard degree of

freedoms (DOFs) and the index d 1 , d 2 , and d 12 are
the enriched DOFs for 1d , 2d and 12
d respectively.
Computation of matrices Kα,β proceeds as usual:
Each matrix block reads
Figure 1. A quadrilateral element (a) intersect by two faults,

and (b)–(c) are sum of the nodal enrichment functions cut by
the faults 1d and 2d respectively.
where npe is the number of nodes per element (4 for
tetrahedra and 8 for hexahedra), α refers either to r
where {e } is the canonical basis,
k
= δik , i, k =
eki
(regular) or d 1 , d 2 , and d 12 (enrichment) functions, and
1, . . . , d. As for the elements cut by two faults (Figure
the matrices Bi , i = 1, . . . , npe are computed as usual
1) the displacement field is written as a linear combi-
with the spatial derivatives of the shape functions.
nation of (8) plus a linear combination of n enrichment α,β
functions for fault-1, fault-2 and their interaction: Computation of surface integrals KS in Equation 14
explained in Das et al., 2015.
3 IMPLEMENTATION DETAILS
In this work, we adopted a shifted-type enrichment
scheme in order to avoid additional computation in In the follwing subsections, we have explained imple-
the partially enriched elements (Das et al., 2015) as mentation of XFEM for 3D intersecting fault.
follows:
3.1 Fault representation using level set function
The level set parameterization which is also adopted
i.e., we use the product of the shifted Heavyside in some commercial software is used to represent the
function and the standard shape functions. As an illus- faults. The idea consists of defining a scalar function
tration, in Figure 1(b)–(c), we plot the sum of the φj : → R, such that
enrichment functions for the case of a quadrilateral
element in 2D. Similary, sum of the interaction enrich-
j
ment function (M12 (x)) is defined as the product of
d
enrichment functions of fault (1d ) and fault (2d ) as where Nf is the number of faults presents in the
shown in equation domain. A practical choice is to define the level set
function as a signed distance function to the fault lin-
early interpolated on Ti . In this case the fault has
d
241
a simple planar representation on each element and
conformity across interelement faces is in principle
ensured (Dompierre et al., 1999).
The level set representation adds flexibility and elle-
gance to the mathematical/computational formulation
even in the case of static faults. For those faults that
end inside the domain, an auxiliary level set function
is introduced so as to define the boundary of Th . This
is illustrated with an example in Figure 2. We consider
the computational domain = [0, 1]3 and two level Figure 2. Example to define a fault whose boundary lies
set functions with a given hierarchy among them inside the computational domain by means of two different
level sets.
The first level set has the highest hierarchy, while

the second, that is used to define the boundary of the
first one, has lowest hierarchy. The discrete interface
Th and its boundary ∂Th are drawn in the right part
of Figure 2. In order to represent multiple intersecting
faults, different level-sets are used for each fault.
Figure 3. Elemental/nodal enrichment scheme in 2D.

3.2 Classification of element and nodes
Elements and nodes close to the fault need to be iden-
tified so as to perform the necessary local operations
related to the enrichment terms computation define in
integral from in Equation 14. This is done only once
during the preprocessing stage of the computation and
the enrichment type is stored. For example, a finite ele-
ment intersected by single/double faults is considered
as a “singly” or “doubly” enriched element. Similarly,
an element sharing a node of an enriched element is
termed as “partially” enriched. If an element is not in Figure 4. Eight possible cases for the smallest vertex in the
any of the above situations then it is termed as regular left bottom corner.
element. However, the enrichment functions are iden-
tically zero at the nodes (Figure 1-(b)-(c)). Therefore, subdivide enriched element into a number of triangu-
we do not need to compute additional stiffness matri- lar subdivisions in 2D and tetrahedral subdivisions in
ces for partially enriched elements. The enrichement 3D. In order to explain this more clearly, we will elab-
scheme for multiple discontinuities is illustrateded in orate one of the most complex cases of hexahedron
Figure 3 for quadrilateral elements in 2D by consid- subdivision. Firstly, a hexahedron is subdivided into 5
ing three faults in the domain. In Figure 3, faults are or 6 tetrahedrons in several ways. An important ques-
shown with red lines. Singly enriched elements are tion when dealing with hexahedral meshes is to ensure
shown in yellow color which are enriched by any of conformity of the fault across inter-element faces after
the fault. Doubly enriched elements are shown in green subdivision. Since in our case the fault is represented
color which are enriched by any two faults. It is noted by a level set function defined as nodal values, and the
that, regular and partially enriched elements are shown interface is reconstructed by linear interpolation of this
with white color. Similarly, regular, singly and doubly level set along the element edges, care must be taken
enriched nodes are shown with white, yellow and red to ensure that subdivision of a given face shared by
color respectively. two hexahedrons is made through the same diagonal.
This is done by making a remapping of the element
incidencies so as to end up with the smallest incidence
3.3 Element subdivision
id at the left bottom corner as indicated in Figure 4
Subdivision of elements that are cut by faults is a nec- (Dompierre et al., 1999).
essary step to redefine the quadrature rules in order There are 8 cases, depending on the smallest id
to integrate the submatrices defined in Equation 14. of a indices in given element that are summarize in
There are several ways in which quadrilateral elements Table 1. Once the remapping is performed, subtetra-
can be cut by faults as shown in Figure 3. The idea is to hedra are defined. Then fault plane is introduced and
242
Table 1. Remapping of incidences for an hexahedral ele-
ment to avoid nonconformity across interelement faces.
each of these subtetrahedra’s are further subdivided

as shown in Figure 5(a). Similar, steps are followed
for doubly enriched elements in which we introdue
the second fault after subdivision is done by the first
fault. The different cases which might arise due to
the introduction of second fault are shown in Figure
5(b)–(d). In this manner we introduce different faults
through the element one after another, which reuse the Figure 5. Elemental subdivisions scheme of a hexahedron
cut by (a) single fault, (b) two non-intersecting faults, (c) two
subdivision algorithms recursively. The implemented
intersecting faults (X-shaped) and (d) T-shaped faults.
subdivision scheme is general and can handle various
intersecting/non-intersecting faults (X and T shape),
and their all the different possible configurations in 3D.
a) Surface is not intersected by any faults: In this
case, dirichelet boundary conditions are imposed
similar as in the standard finite element method
3.4 Gauss point generation (FEM).
b) Surface is intersected by single fault: In this
Subdivision of singly and doubly enriched elements is
case, dirichelet boundary conditions are imposed
performed in order to generate Gauss points which
as reported author’s earlier work (Das et al., 2015).
facilitates different integrations to be performed at
c) Surface is intersected by two faults: Dirichlet
element level. Each subdivision can have a number
boundary conditions (DBC) are imposed on reg-
of Gauss points depending on the integration rule
ular as well as additional (enriched by individual
(linear/quadratic). For example, each of the subte-
faults and their interaction) DOFs. Imposition of
trahedra can either have a single Gauss point in the
DBC on regular DOFs is idential to the case (a).
linear case or three Gauss points in the quadratic
Imposition of DBC on enriched DOFs for indi-
case. The Gauss point coordinates and their corrspond-
vidual faults is identical to case (b). However, on
ing weights are generated in physical space using the
interaction DOFs we might impose or leave free
physical shape function of the subdivided tetrahedrals.
depending upon the physical meaning of the DBC.
Similar computations can also be performed in the
For example, if a surface has uniform DBC, then
master space and then corresponding transformation
the solution of interaction DOF is constrained to
to physical space using isoparametric mapping. How-
have zero value. However, for non-uniform DBC,
ever, we have avoided the mapping step by performing
multiple sub cases arises which are individually
direct computations in the physical space. The con-
handled.
cept of Gauss quadrature scheme in the physical space
using mapping has been explained in Das et al., 2015.
4 NUMERICAL EXAMPLES: RESULTS AND

DISCUSSIONS
3.5 Boundary conditions on the fault
Neumann boundary conditions are imposed by com- We have developed in-house XFEM code which runs
puting enriched force vectors (Fid ∀i = 1, 2, 12) which over distributed environment and can handle very large
are surface integrals defined in Equation 14. Details models. The code is conceived to solve complex geo-
about the imposition of Dirichlet and Neumann bound- mechanical problems with different loading conditions
ary conditions on the enriched element has been and in order to simulate the behavior of reservoir rock
reported in the authors’ earlier work (Das et al., 2015). mass with faults. The first verification example con-
Here, we have extended similar concept for dou- sidered is a column (Figure 6) with two orthogonal
bly enriched element. Dirichlet boundary conditions intersecting fault planes. Body is subjected to differ-
applied on the surface of a doubly enriched element are ent displacement boundary conditions on the lateral
handled depending upon number of faults intersecting surfaces (Table 2), in order to create different displace-
the surface. There can be three different cases: ment mechanisms and evaluate the influence of faults
243
Figure 7. (a) Horizontal and (b) vertical displacement
profile for case-A.
Figure 6. Example definition and dimensions.
Table 2. Boundary conditions on lateral surfaces.
δx δy
Surface 1 2 3 4 5 6 7 8
Case-A 0 0 −δ −δ 0 0 −δ −δ
Case-B 0 δ −δ 0 0 δ −δ 0
Case-C 0 δ −δ 0 0 0 0 0
Case-D 0 δ −δ 0 0 0.5δ −δ 0 Figure 8. (a) XFEM and (b) IE-FEM normal stress on fault
planes for Case-A.
on stresses field. For further validation purpose, a sec-

ond example considers arbitrary inclined faults in a
column (Figure 15) with Triaxial 3D loading condi-
tions. In both the examples, results are compared with
existing analytical solutions and with standard FEM
using zero-thickness IE’s code results.
4.1 Orthogonal fault planes

Figure 9. (a) Horizontal and (b) vertical displacement
In this example, a three-dimensional column 10 m × profile for case-B.
10 m × 20 m is cut by two orthogonal intersecting fault
planes as shown in Figure 6. In this subsection, all
test cases assume following mechanical properties
of rock mass and faults stiffness parameters: Young
modulus (E) 10 GPa, Poisson ratio (ν) is 0.0, fault
normal and shear stiffness are kn = 10000 GPa/m, and
kt1 = kt2 = 1 GPa/m respectively. In all the cases, lat-
eral displacement profiles along with deformed bodies
are shown with 10 times magnification. The normal
stress on the fault plane is also plotted.
Case-A: Uniform displacement of amount (δ) has
been applied in the horizontal (on the faces 3-4)
and vertical (on the faces 7-8) directions respectively, Figure 10. (a) Horizontal and (b) vertical displacement
whereas no-slip boundary conditions are imposed on profile for case-C.
opposite faces in the horizontal (on the faces 1-2) as
well in the vertical directions (on the faces 5-6). There- in Figure 8(a). It noted that, computed normal stress
fore, uniform displacement profile as shown in Figure is in good agreement with the IE-FEM results shown
7 is observed in both the directions (horizontal and in Figure 8(b).
vertical) as expected. In other words, constant normal Case-B: Rigid DBC are imposed on surfaces 1, 4,
jump has occurred on each face of fault planes and uni- 5, 8 whereas uniform displacement (δ) is applied on
form normal stress of 50 MPa is observed as shown surfaces 2, 3, 6, 7 in order to form direct shear state
244
Figure 11. Displacement along the line ABC (Figure 9)
(inset showing zoomed view of displacement jump on fault). Figure 13. (a) Horizontal and (b) vertical displacement
profile for case-D.
Figure 12. Normal stress on fault planes for Case-B and

Figure 14. (a) XFEM and (b) IF-FEM normal stress on fault
Case-C respectively.
planes for Case-D.
of movement with respect to both the faults planes as
the fault plane x = 5 whereas δ amount of jump has
specified in Table 2. Displacement profiles in the lat-
occurred in all other lateral surfaces. Due to different
eral directions are shown in Figure 9(a) and (b) which
movement of lateral surfaces, normal stress distribu-
are evidence of successful creation of the direct shear
tion on the fault planes is non-uniform as shown in
kind of movement as imposed through DBC. In these
Figure 14(a). Normal stress of ∼25 MPa is observed on
conditions, each face of the fault plane has uniform
the face of fault plane located in the middle of surfaces
normal displacement and hence, uniform normal stress
6 and 8 and all other faces on faults plane experienced
(∼50 MPa) is observed as shown in Figure 12(a).
uniform normal stress of ∼50 MPa. These results are
Case-C: Simplified version of Case-B is consid-
in close agreement with the interface element results
ered here by restricting no-slip boundary in vertical
shown in Figure 14(b).
direction. This case represents direct shear movement
with respect to the fault plane x = 5. The computed
displacement profiles are shown in Figure 10. Hori-
4.2 Column with inclined intersecting faults
zontal displacement along the line ABC which crosses
the fault plane x = 5 at point B is shown in Figure 11 In this subsection, we studied column example with
along with the theoretical result (Das, 2013). In the two intersecting faults plane at an angle β1 = 46.94◦
inset of the Figure 11 the jump is plotted in the hori- and β2 = −60◦ with horizontal axis as shown in Figure
zontal displacement. Excellent agreement is observed 15. A total of 2000 hexahedral elements are consid-
with the theoretical solution (XFEM: 0.000049, ana- ered for this column example. The width, length and
lytical: 0.00005). Uniform normal stress of 50 MPa height of the column are 10 m × 10 m × 20 m respec-
occurs on fault plane x = 5 whereas negligible normal tively. Mechanical properties of rock mass and faults
stress occurs on fault plane y = 5 as shown in Figure stiffness parameters are assumed as: Young modulus
12(b). This is as expected due to the nature of applied (E) 10 GPa, Poisson ratio (ν) is 0.25, fault normal stiff-
DBC as uniform jump occurs in the normal direction of ness (kn ) is 100 GPa/m and shear stiffness (kt1 = kt2 )
one fault plane (x = 5) whereas negligible jump occurs are 10 GPa/m respectively. No-slip boundary condi-
on the other. tions are applied on the bottom surface and uniform
Case-D: This case is similar to the Case-B except pressure (σ1 ) 30 MPa is applied on top surface in z-
for the applied DBC on surface 6. On Surface 6 dis- direction. Confining pressure of (σ2 = σ3 ) 20 MPa is
placement of 0.5 × δ is applied in the vertical direction applied on lateral surfaces. Normal and shear stresses
which creates asymmetric movement w.r.t fault planes. on inclined fault planes are shown in Figure 15 and
Due to non-uniform DBC, interaction DOFs plays 16 respectively. In static condition, analytical solution
active role. Distribution of lateral displacement pro- for stress on faults (Jaeger, 1960) is also applicable for
files are plotted in the Figure 13 which replicates the multiple fault planes. The traction profile on inclined
applied DBC mentioned in Table 2. It is noted that, the fault planes matches the theoretical results and has less
surface y = 0 experience 0.5 × δ amount of jump along than 1% error which is reported in Table 3.
245
In XFEM, FE mesh suffices and the faults can arbi-
trarily cut the elements. This advantage of the XFEM
has been leveraged for many applications in fracture
mechanics, and in this paper we have shown that it is
equally advantageous for modeling intersecting faults.
Three-dimensional benchmark cases are presented to
validate the accuracy of the approach, and the potential
benefits in applications. The present study shows that
interaction DOFs and their associated surface stiffness
matrices plays a significant role in accurate modeling
of cohesive faults via XFEM.
REFERENCES
Belytschko, T., Moës, N., Usui, S., & Parimi, C. (2001).
Arbitrary discontinuities in finite elements. International
Figure 15. Normal stress on fault with two inclined faults Journal for Numerical Methods in Engineering, 50(3),
(46.94 deg and 60 deg respectively). 993–1013.
Borja, R. I. (2008). Assumed enhanced strain and the
extended finite element methods: A unification of con-
cepts. Computer Methods in Applied Mechanics and
Engineering, 197(33), 2789–2803.
Das, K.C. (2013), Enriched Finite Element Method and
Applications to Reinforced Jointed Rock Mass, PhD
Thesis, Indian Institute of Technology Kharagpur, India.
Das, K. C.,Ausas, R. F., Serra, J. M. S., Narang,A., Rodrigues,
E. R., Carol, I., & Mello, U. T. , EFEM vs. XFEM:
A comparative study for modeling strong discontinu-
ity in geomechanics. 13th International ISRM Congress,
Montreal, May 10–13, 2015.
Das, K.C., R. Ausas, I., Carol, E. Rodrigues, S.S. Sandha,
P.E. Vargas, N. A. González, J. M. Segura, M.R. Laksh-
mikantha & U.T. Mello, “Discrete Modeling of Multiple
Discontinuities in Rock Mass using XFEM.” 49th US
Rock Mechanics/Geomechanics Symposium, American
Figure 16. Resultant shear stress on fault with two inclined Rock Mechanics Association (ARMA), San Francisco, 28
faults (46.94 deg and 60 deg respectively). Jun–1 July, 2015.
Deb., D., Pramanik, R., Das, K.C., A generalized XFEM pro-
Table 3. Comparison of stresses on faults plane: theoretical cedure for analyzing intersecting joints in rock masses
solution vs. XFEM results. with excavation. Engineering Computations 2015; 32(3):
806–833
Inclination (β) (Deg) 46.54 −60 Dompierre, J., Labbé, P., Vallet, M. G., & Camarero, R.
(1999, October). How to Subdivide Pyramids, Prisms, and
Theoretical Normal stress (MPa) 24.660 22.500 Hexahedra into Tetrahedra. In IMR (195–204).
Shear stress (MPa) 4.988 4.330 Jaeger, J.C. (1960). Shear fracture of anisotropic rocks,
XFEM Normal stress (MPa) Min 24.611 22.490 Geological Magazine 97: 65–72.
Max 24.704 22.500 Fries,T. P., & Belytschko,T. (2010).The extended/generalized
Shear stress (MPa) Min 4.988 4.330 finite element method: an overview of the method and its
Max 4.989 4.330 applications. International Journal for Numerical Meth-
Average Normal stress 0.378 0.044 ods in Engineering, 84(3), 253–304.
error (%) Shear stress 0.020 0.004 Gens, A., Carol, I., and Alonso, E.E., 1988. An interface ele-
ment formulation for the analysis of soil-reinforcement
interaction. Computers and Geotechnics, 7:133–151.
Goodman, R. E., Taylor, R. L., & Brekke, T. L. (1968).
A model for the mechanics of jointed rock. Journal of
5 CONCLUSIONS Soil Mechanics & Foundations Div., 94 (SM3): 637–659,
1968.
In this paper, we have introduced a novel formulation Watanabe, N., Wang, W., Taron, J., Gorke, U.J., Kolditz, O.
of XFEM which can handle arbitrary multiple inter- (2012). Lower dimensional interface elements with local
secting cohesive faults in 3D. Due to different possible enrichment: application to coupled hydro-mechanical
problems in discretely fractured porous media. Interna-
intersections of multiple faults within elements, sev-
tional Journal for Numerical Methods in Engineering,
eral types of enrichments are defined and handled with 90(8), 1010–1034.
“Shifted Heaviside function”. A generalized subdivi-
sion scheme has been described to capture the complex
geometries of multiple faults using Level set function.
246
DEM modeling of hydraulic fracturing in fractured shale formation: Effect

of inherent anisotropy and induced anisotropy
K. Duan & C.Y. Kwok

State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics,
Chinese Academy of Sciences, Wuhan, China
Department of Civil Engineering, the University of Hong Kong, Hong Kong
ABSTRACT: Hydraulic fracturing has been widely used in the oil and gas industry to increase the recovery
of hydrocarbons from low permeability formations. The hydraulic fracturing process can be very complex in
naturally fractured reservoirs due to the anisotropy of material properties and existing of natural fractures.
The aim of this study is to better understanding the mechanics of hydraulic fracturing in fractured reservoirs
by performing numerical simulations in a two-dimensional discrete-element particle flow code (PFC2D). In
the numerical model, rock matrix is represented by bonded-particle model (BPM). The intrinsic anisotropy is
explicitly represented by imposing smooth-joint models. Any pre-existing horizontal or vertical natural joints
are added by superimposing continuous smooth joint contacts onto the BPM to form a Synthetic Rock Mass
(SRM) in which fluid injection and rock fracturing can be modeled in a fully coupled manner. The models are
first calibrated to reproduce the mechanical behaviors of the isotropic and anisotropic rock in field. Effects of
inherent anisotropy, joints orientation, and joint apertures on the hydraulic fracturing growth are investigated by
conducting a series of comparative simulations. Results of these simulations show that the formation’s anisotropy
plays an important role in the orientation of hydraulic fractures and it promotes horizontal fracture growth. Joints
properties (orientation and aperture) could have a major impact on the behavior of rock mass during hydraulic
fracturing operations.
1 INTRODUCTION nature of hydraulic fracturing. Field studies (Britt et al.,

1994, Rodgerson, 2000, Azeemuddin et al., 2002) con-
Hydraulic fracturing has been widely used in the indus- ducted in naturally fractured formations reveal that
try of oil and gas production to increase the recovery the effects of natural fractures on hydraulic fracture
of hydrocarbons from low permeability formations. propagation are enhanced fluid leak-off, premature
The objective of hydraulic fracturing is to increase the screen-out, arrest of the fracture propagation, forma-
permeability of the surrounding rock mass. In nat- tion of multiple fractures, fracture offsets, and high net
urally fractured reservoirs, due to the heterogeneity pressure.
and anisotropy of material properties and existing of Various numerical approaches have been devel-
naturally fractures, hydraulic fracturing processes are oped to estimate the development of hydraulic
highly complex (Germanovich et al., 1997, Maxwell fractures, e.g., the Khristianovic-Geertsma-deKlerk
et al., 2002). Understanding the mechanics of propa- (KGD) model (Geertsma and De Klerk, 1969), the
gation of hydraulic fracturing in fractured reservoirs Perkins-Kern-Nordgern (PKN) model (Perkins and
is important to the design and operation of hydraulic Kern, 1961, Nordgren, 1972), pseudo-3D models
fracturing. (P3D)(Mack and Warpinski, 2000) and planar 3D
In the laboratory, a number of small-scale hydraulic models (PL3D) (Advani et al., 1990). However, all
fracturing experiments have been conducted on either of these continuous methods are based on several
natural rock samples (Matsunaga et al., 1993, Falls assumptions such as isotropic and homogeneous mate-
et al., 1992) or artificial materials (Groenenboom and rial, linear elastic deformation, and assumption of
van Dam, 2000, Zhou et al., 2008). Various interac- linear elastic fracture mechanics (LEFM) for fracture
tion types (cross, dilate and arrest) between induced growth. In fact, most rock mass formations in field
fractures with pre-fracture have been observed (Zhou are inhomogeneous with abundant bedding planes and
et al., 2008), controlled by the differential horizontal natural fractures which will significantly affect the
stress, angle of approach and shear strength of pre- hydraulic fracturing trajectory (Damjanac et al., 2013).
fracture. In the field, microsecimic monitoring was The discrete nature of Discrete Element Method
often used to provide information about the extent and (DEM) allows for explicitly modeling of out-of-plane
247
Figure 1. Schematic diagram of smooth-joint contact (after
Ivars et al., 2011).
fracture propagation, fracture branching and reorien-

tation, and shearing of natural fractures ahead of the
hydraulic fracture front. In this study, the propagation Figure 2. PFC model used in the hydraulic fracture sim-
of hydraulic fracturing in fractured shale formation is ulation. Reservoirs (black dots), flow paths (black lines)
investigated by DEM modeling. and bonds (while lines) in compacted bonded assembly of
particles (after Itasca, 2010).
2 NUMERICAL METHODOLOGIES and that these channels connect up small reservoirs that
store fluid at some pressure, as illustrated in Figure 2.
DEM simulations in this study are performed in two- Each pipe is assumed to be a set of parallel plates with
dimensional Particle Flow Code (PFC2D). In the some aperture. The flow rate in a pipe is given by:
numerical model, intact rock is represented as an as-
sembly of rigid disks bonded at their contact points
(Bonded Particle Model). Any pre-existing horizontal
bedding and vertical joints are added by superimpos-
ing smooth joint contacts onto the BPM to form the where µ is fluid viscosity, (P1 − P2 ) is the pressure
synthetic rock mass (SRM). A brief introduction to difference between the two adjacent domains. L and a
the smooth joint model and the fluid flow model-ing represent the length and aperture of pipe, respectively.
algorithm in PFC2D are provided in this section. Each pipe has an aperture with it which is defined
by:
2.1 Smooth joint mode
The smooth-joint contact model was first proposed
by Cundall et al. (1996) to represent fractures. The where a0 is the residual aperture when the two particles
smooth-joint contact model allows particles at the joint are just in contact, σ is the normal stress at the con-
surface experience relative slip on the specified joint tact and σ0 is the normal force at which the aperture
surface rather than sliding along the particle surface decrease to a0 /2
as depicted in Figure 1. The behavior of joints can be Each domain receives flows from the surrounding
modeled by assigning smooth-joint models to all con- pipes: q. In one time step, t, the increase in fluid
tacts between particles that lie on opposite side of the pressure P is given by:
joint.
Most recently, the authors have adopted the indi-
vidual smooth joint model to represent the intact
anisotropic rock with great success (Duan and Kwok,
2015, Duan et al., 2015). In this study, the inherent where Kf is the fluid bulk modulus and Vd is the appar-
anisotropy of intact rock is simulated by imposing indi- ent volume of the domain. The change in fluid pressure
vidual smooth joint models onto the bonded particle will exerts forces on enclosing particles causing defor-
model. The induced anisotropy originated from joints mation and subsequent particle movement alters the
is modeled by assigning continuous smooth-joint mod- contact forces, which affects the fluid flow by altering
els to all contacts that lie on the opposite sides of the the channel aperture.
joints.
3 SETUP OF NUMERICAL MODEL

2.2 Fluid flow modelling in PFC
Fluid flow is simulated in PFC2D (Itasca, 2010) by The DEM model is constructed to simulate a 2D
assuming that each particle contact is a flow channel slice of the Haynesville shale reservoir approximately
248
Table 1. Micromechanical parameters for the intact
isotropic model.
Particle parameters Parallel bond parameters
E0 28 GPa Ēc 28 GPa

kn /ks 2.5 k̄n /k̄s 2.5
µ 0.5 σ̄c 56 ± 11.2 MPa
Rmax /Rmin 1.66 τ̄c 56 ± 11.2 MPa
ρ 3169 kg/m3 λ̄ 1
Table 2. Mechanical properties for the anisotropic model.
Modelled
formations Numerical model
Reduction
Property H/V Vert. Horiz. H/V factor
UCS, MPa 0.96 74.19 70.89 0.96 2.5

Young’s 1.6 20.98 33.53 1.60 3.0
Modulus, Figure 3. Setup of numerical model for the simulation of
GPa hydraulic fracturing in anisotropic m odel with open joints in
vertical direction.
3,739 m subsurface. A model size of 50 × 50 m with four different cases are considered: intact isotropic
particle diameter of 0.5 m is used. Particles in this model, intact anisotropic model, anisotropic model
model do not represent single mineral grains in the with open vertical joints, and anisotropic model with
rock. They are simply a way to discretize the medium. open horizontal joints. Open vertical and horizontal
The micro parameters used in the DEM model are joints are represented in the model by unbounded
calibrated to match the mechanical responses of the continuous smooth joints. The spacing between each
intact rock under uniaxial compression test (uniax- joint is 2.5 m, which is 5 particles apart. Three dif-
ial compression strength 73.53 MPa, Young’s modulus ferent initial apertures, 0.1 mm, 0.7 mm and 1.0 mm,
32.9 GPa, and Poission’s ratio 0.27). The size of sample are assigned to the open joints to test the sensitivity of
used for calibration is 10 × 20 m with particle diameter these parameters.
of 0.5 m. Corresponding micro-parameters obtained As illustrated in Figure 3, the vertical in-situ stress
from calibration are illustrated in Table 1. Sv = 93 MPa and the horizontal maximum principal
The anisotropic model is generated by removing stress SH max = 80 MPa. The pore pressure is 75 MPa.
any sub-horizontal parallel bonds (those dipping less These initial conditions are applied to the DEM models
than 20◦ ) and replacing them with horizontal smooth prior to fluid injection. In all simulations, the location
joint contacts. The properties of the smooth joint con- of the injection point is at the bottom of the simu-
tacts, except for dip angle which is zero in this case, lated domain, assuming the model is symmetric along
are inherited from the propertied of the deleted par- the boundary. The outside model boundary is fixed
allel bond contact and the two contacting particles. and impermeable. The average operational parameters
Theses smooth joints used to represent the bedding used in the simulation are: the injection fluid viscosity
planes have an initial aperture of 0.5 mm. The val- 0.001Pa.s; average injection rate 6.3588 m3 /min. Fig-
ues of stiffness and bond strength for the smooth ure 3 illustrates an example of the anisotropic model
joints are lowered with different reduction factors by with vertical joints before injection.
trial-and-error to achieve the desired anisotropy (ratio
between properties obtained from horizontal and ver-
tical direction, H/V listed in Table 2) Virtual uniaxial 4 SIMULATION RESULTS
compression tests are carried out in vertical (per-
pendicular to beddings) and horizontal (parallel with Results of the simulations are presented in this section.
beddings) direction of the anisotropic model in order For each case, two sets of plots are included: (1) Injec-
to obtain the reduction factor for Young’s Modulus tion pressure history; (2) The results of fluid injection
and parallel bond strength of bedding plane contacts. showing induced fractures, pore pressures and smooth
The obtained macroscopic mechanical properties and joint slip at the last stage of modeling
related reduction factors for the anisotropic model are
shown in Table 2.
After the anisotropic model is calibrated, natural
4.1 Hydraulic fracturing in isotropic model
fractures including vertical or horizontal joints are
added to the assembly to create different models that In the isotropic rock mass without beddings and frac-
mimic the individual lithology in field. In this study, tures, the hydraulically induced fracture propagates
249
difference can be found by comparing the fluid pres-
sure in Figure 5(b): in the case of anisotropic rock, the
induced fracture shows less pressure differential along
its length than the fracture in the isotropic rock which
means that the average aperture in the former case
is larger, leading to better hydraulic communication
along the fracture.
Boundary effects can be observed from the
anisotropic model at the beginning of the injection pro-
cess as shown in Figure 5 (b) (horizontal part of the
fracture near the injection point). However, later in the
injection process, the fracture propagates within the
model, and become inclined as it is affected by both
the in-situ stress and the inherent horizontal bedding.
It is clear that in this case, vertical fracture growth is
hindered by the modeled bedding. For this anisotropic
model, the final fracture height achieves after fluid
injection is less than 1/3 of the fracture height achieved
in the case of isotropic model as shown in Figure 4(b)
(with injection time being about 600 sec in both cases).
4.3 Hydraulic fracturing in anisotropic model with

open vertical joints
The presence of open vertical joints is modeled in
this part and the effect of the changes in the aper-
ture of such joints on fracture propagation is evaluated.
Three different apertures are tested: 0.1 mm, 0.7 mm
and 1.0 mm.
The injection pressure histories for the three aper-
tures are presented in Figure 6(a). The models run
long enough to see fluid pressure breakthrough into
the upper edge of the model. Very similar shape and
magnitude in their pressure response can be observed
Figure 4. Simulation results of hydraulic fracturing in between the cases with different apertures, which mean
isotropic model. (a) Injection pressure history. (b) Induced that comparable amounts of energy were necessary in
cracks and induced pore pressure. The radii of the blue circles each case in order to sustain fracture propagation.
are proportional to the magnitude of induced fluid pressure. In this series of tests, and despite the fact that the
simulation for aperture 0.1 mm fails (Figure 6(b)),
nearly vertically upward as dictated only by the ini- modeling shows that the hydraulic fracture tends to
tial stress state (Figure 4(b)). The radii of the blue propagate along the open vertical joints once they are
circles are proportional to the induced fluid pressure contacted (see Figure 6(c) and (d)). The time it takes to
magnitude, thus, the pressure differential between the reach upper boundary is, in general, proportional to the
injection point and fracture tip is large. From the injec- apertures assigned to the vertical joints. For 0.7 mm
tion pressure history (Figure 4(a)), there is a steady aperture, it takes 56.5 sec to reach the upper model
increase of injection pressure after the initial break- boundary, while it takes 46.18 sec for the case with
down. This can be explained by the fact that as the aperture equals to 1.0 mm.
fracture propagates farther away from the injection
point, more energy is necessary to propagate the crack.
4.4 Hydraulic fracturing in anisotropic model with
open horizontal joints
4.2 Hydraulic fracturing in anisotropic model
The presence of open horizontal joints is modeled in
Simulation results of hydraulic fracturing in anisotropic this part. The effect of changes in the apertures on
model are illustrated in Figure 5. The history of injec- fracture propagation is evaluated. Same with that of
tion pressure (Figure 5(a)) reveals a similar behavior to vertical joints in Section 4.3, three magnitudes of joint
the one found in the case of isotropic model (see Figure aperture are tested: 0.1 mm, 0.7 mm and 1.0 mm.
4(a)): a breakdown peak, followed by a continuous lin- As illustrated in Figure 7(a), the injection pressure
ear pressure increase over time. In this case, however, histories of the simulations for the three apertures show
the slope of such increase is higher; in other words, the the same behavior at the beginning of the injection
amount of energy needed for sustaining fracture prop- (about the first 130 sec of injection), probably show-
agation is higher than that of isotropic rock. Another ing rock failure between the injection point and the
250
Figure 5. Simulation results of hydraulic fracturing in
anisotropic model. (a) Injection pressure history. (b) Induced
cracks and induced pore pressure. The radii of the blue circles
are proportional to the magnitude of induced fluid pressure.
first open horizontal bedding plane. But after such

plane is contacted, the amount of energy needed for
fracture propagation becomes a direct function of the
assigned joint aperture. Larger joint aperture in gen-
eral demands less injection pressure to sustain the same
injection rate.
The case with joints apertures equal to 0.1 and
0.7 mm are run for about 600 sec of injection time.
However, for the case with the largest aperture, the
simulation is stopped when the fracture extends hor-
izontally to reach the model lateral edge, which
occurs about 519 sec of injection. In all these tests, Figure 6. Simulation results of hydraulic fracturing in
it is observed that the injected fluid tends to follow anisotropic formation with vertical open joints. (a) Injection
the horizontal joint planes once they are contacted; pressure history. Induced cracks and induced pore pressure
thus, severely impending fracture vertical growth. It when the aperture of joints equals to (b) 0.1 mm, (c) 0.7 mm
and (d) 1.0 mm. The radii of the blue circles are proportional
appears, at least for these aperture magnitudes, that
to the induced fluid pressure.
the effect of applied stress on fracture propagation is
overcome by the presence of the horizontal open joints.
251
5 CONCLUSIONS
The effects of inherent anisotropy (beddings) and

induced anisotropy (pre-existing joints) on the propa-
gation of hydraulic fracturing in fractured shale forma-
tion are investigated in this study. In the DEM model,
rock matrix is represented by bonded particle model.
The existing of inherently anisotropy is represented
by imposing individual smooth-joint contacts horizon-
tally. Any pre-existing horizontal or vertical joints are
added by superimposing continuous smooth joint con-
tacts onto the sample. A series of comparative studies
are conducted based on different combinations of open
horizontal and vertical joints.
Simulation results show that the formation’s
anisotropy plays an important role in the orientation
of hydraulic fractures and it promotes horizontal frac-
ture growth. Likewise, the presence of open vertical
fractures in the rock creates the distinct possibility for
uncontrolled height growth. This is a function of ver-
tical joints aperture: larger apertures allow for easier
flow transmission within the joints and consequently,
make fracture propagation in the vertical direction
easier.
The joints properties (orientation and aperture)
could have a major impact on the behavior of rock
mass during hydraulic fracturing operations. Thus, the
need for a comprehensive characterization of the frac-
tures in the rock is not only apparent but critical. The
numerical method presented in this study has shown
the flexibility to explicitly model the presence of such
fractures and also the effects that they could have on
the fracturing operation.
ACKNOWLEDGEMENT
This study is supported by the National Natural

Science Foundation of China (NSFC) (Grant no.
51428902) and Open Research Fund of State Key
Laboratory of Geomechanics and Geotechnical Engi-
neering, Institute of Rock and Soil Mechanics, Chinese
Academy of Sciences, Grant NO. Z014004.
REFERENCES
Advani, S., Lee, T. & Lee, J. 1990. Three-dimensional mod-
eling of hydraulic fractures in layered media. I, Finite
element formulations. Journal of energy resources tech-
nology, 112, 1–9.
Azeemuddin, M., Ghori, S., Saner, S. & Khan, M. Injection-
induced hydraulic fracturing in a naturally fractured
carbonate reservoir: a case study from Saudi Arabia.
International Symposium and Exhibition on Formation
Figure 7. Simulation results of hydraulic fracturing in Damage Control, 2002.
anisotropic formation with open joints in horizontal direction. Britt, L., Hager, C. & Thompson, J. Hydraulic fracturing
(a) Injection pressure history. Induced cracks and induced in a naturally fractured reservoir. International Petroleum
pore pressure when the aperture of joints equals to (b) 0.1 mm, Conference and Exhibition of Mexico, 1994.
(c) 0.7 mm and (d) 1.0 mm. The radii of the blue circles are Cundall, P., Potyondy, D. & Lee, C. Micromechanics-based
proportional to the induced fluid pressure. models for fracture and breakout around the mine-by
tunnel. Proceedings, International Conference on Deep
Geological Disposal of Radioactive Waste, Winnipeg.
252
Edited by JB Martino and CD Martin. Canadian Nuclear Itasca 2010. PFC2D Particle Flow Code in 2 Dimensions. 4.0
Society, Toronto, 1996. 113–122. ed. Minneapolis.
Damjanac, B., Detournay, C., Cundall, P. A. & Varun 2013. Ivars D M, Pierce M E, Darcel C, et al. 2011. The synthetic
Three-Dimensional Numerical Model of Hydraulic Frac- rock mass approach for jointed rock mass modelling.
turing in Fractured Rock Masses. International Journal of Rock Mechanics and Mining
Duan, K. & Kwok, C. Y. 2015. Discrete element modeling of Sciences, 48, 219–244.
anisotropic rock under Brazilian test conditions. Interna- Mack, M. G. & Warpinski, N. 2000. Mechanics of hydraulic
tional Journal of Rock Mechanics and Mining Sciences, fracturing. Reservoir Stimulation 6–1.
78, 46–56. Matsunaga, I., Kobayashi, H., Sasaki, S., et al Studying
Duan, K., Kwok, C. Y. & Pierce, M. 2015. Discrete element hydraulic fracturing mechanism by laboratory experi-
method modeling of inherently anisotropic rocks under ments with acoustic emission monitoring. International
uniaxial compression loading. International Journal for journal of rock mechanics and mining sciences & geome-
Numerical and Analytical Methods in Geomechanics chanics abstracts, 1993. Pergamon, 909–912.
n/a–n/a. Maxwell, S.C., Urbancic, T.I., Steinsberger, N., et al. Micro-
Falls, S., Young, R., Carlson, S. & Chow, T. 1992. Ultrasonic seismic imaging of hydraulic fracture complexity in the
tomography and acoustic emission in hydraulically frac- Barnett shale. SPE Annual Technical Conference and
tured Lac du Bonnet grey granite. Journal of Geophysical Exhibition, 2002.
Research: Solid Earth (1978–2012), 97, 6867–6884. Nordgren, R. 1972. Propagation of a vertical hydraulic
Geertsma, J. & De Klerk, F. 1969. A rapid method of predict- fracture. Old SPE Journal, 12, 306–314.
ing width and extent of hydraulically induced fractures. Perkins, T. & Kern, L. 1961. Widths of hydraulic fractures.
Journal of Petroleum Technology, 21, 1571–1581. Journal of Petroleum Technology, 13, 937–949.
Germanovich, L. N., Astakhov, D. K., Mayerhofer, M. J., Rodgerson, J. Impact of natural fractures in hydraulic frac-
Shlyapobersky, J. & Ring, L. M. 1997. Hydraulic fracture turing of tight gas sands. SPE Permian Basin Oil and Gas
with multiple segments I. Observations and model for- Recovery Conference, 2000.
mulation. International Journal of Rock Mechanics and Zhou, J., Chen, M., Jin, Y. & Zhang, G.-Q. 2008. Analysis
Mining Sciences, 34, 97.e1-97.e19. of fracture propagation behavior and fracture geometry
Groenenboom, J. & van Dam, D. B. 2000. Monitoring using a tri-axial fracturing system in naturally fractured
hydraulic fracture growth: Laboratory experiments. Geo- reservoirs. International Journal of Rock Mechanics and
physics, 65, 603–611. Mining Sciences, 45, 1143–1152.
253
Comparison of discrete and equivalent continuum approaches to simulate

the mechanical behavior of jointed rock masses
N.A. González & P.E. Vargas

REPSOL Technology Center, Mostoles, Madrid, Spain
I. Carol
ETSECCPB (School of Civil Engineering), UPC (Technical University of Catalonia), Barcelona, Spain
K.C. Das, S.S. Sandha, E. Rodrigues & U. Mello

IBM Research
J.M. Segura & M.R. Lakshmikantha

REPSOL Technology Hub, The Woodlands, Houston, Texas, USA
ABSTRACT: Modeling of rock masses is of major importance to assess the Geomechanical behaviour of oil
& gas reservoirs, especially for fractured tight reservoirs. The presence of discontinuities will significantly influ-
ence the general behavior of the rock masses, in particular introducing strength reduction, enhanced/reduction
permeability, anisotropic behavior and a non-linear response. In the present study, Discrete and Equivalent
continuum approaches have been used to simulate the mechanical behavior of jointed rock masses. Discrete
approach uses the eXtended Finite Element Method (XFEM) and the Zero-thickness interface elements, while
the Equivalent continuum approaches uses an elastic-viscoplastic constitutive model of the multilaminate type
to represent the rock mass behavior. Advantages and limitations of each approach are identified and some hints
for their practical use are given. Although the discrete approach is sometimes preferred for being based on a
mature theory, the equivalent continuum analysis seems to be more often applicable for usual geomechanical
analyses from engineering practice.
1 INTRODUCTION the permeability depending on the deformation of the

rock mass.
Modeling of rock masses is of major importance to Finite Element Method (FEM)-based models for
assess the Geomechanical behaviour of mining/civil the analysis of the mechanical behavior of rock
tunnel, oil & gas reservoirs, etc. Rock masses are in masses can be split into two main groups: those
most cases composed of an assembly of rock blocks that explicitly discretize each discontinuity using spe-
separated by sets of discontinuities such as joints, cial joint/interface/enriched finite elements, and those
bedding planes, fractures and faults that will condi- based on a continuum-type representation of the
tion the behavior of the system. Global behavior of medium. In the present study, discrete and equiva-
the rock mass depends on the relative importance lent continuum approaches based on FEM have been
of both components (matrix and discontinuities), the implemented in a Parallel in-house-Finite element
number, nature and characteristics of discontinuities code.
and the scale of the analysis. The presence of dis- Discrete representation of discontinuities can be
continuities will significantly influence the general done using both explicit and implicit ways. The more
behavior of the rock masses, in particular introducing classical explicit representation is the interface ele-
strength reduction, enhanced/reduction permeability, ment method (Goodman at al., 1968; Gens et al.,
anisotropic behavior and a non-linear response. 1998; Garolera et al., 2013) in which the disconti-
In the Oil & Gas industry, an accurate descrip- nuity is explicitly discretized with special elements
tion of the discontinuities behavior is a fundamen- inserted in-between element faces/edges. In this case
tal aspect during reservoir development and pro- additional degrees of freedom are introduced at the
duction including well planning, completion opera- nodes to capture the discontinuous behavior of the
tions such as hydraulic fracturing interaction with displacement. On the other hand, the eXtended Finite
pre-existing natural fractures/faults, and produc- Element Method (XFEM) (i.g. Deb & Das, 2010) rep-
tion/injection operations that might reduce/ maintain resents implicitly the discontinuities. In this method,
255
the discontinuity is given as the zero set of a level
set function which cuts the elements in principle
in arbitrary ways. The main advantage is therefore
the flexibility to define discontinuities with complex
geometries. In this case, appropriate enrichment func-
tions have to be introduced near the discontinuity so
as to capture the jump in the displacement field.
In the Equivalent Continuum approach, the original
discontinuous medium is substituted by a contin-
uum one with a constitutive model that incorporates
the effect of the intact material and that of the dis-
continuities, in other words, it defines an equivalent
material in which the properties of the joint system are
smeared out over a unit volume of rock. An elastic-
viscoplastic constitutive model of the multilaminate
type (Zienkiewicz and Pande, 1977; Caballero et al.,
2009) has been implemented, it includes the possibility
of incorporating up to three discontinuity planes with
given orientation and elastic and strength parameters.
Advantages and limitations of each approach are
highlighted by means of the numerical simulation of a
jointed rock block subject to different loading condi-
tions. Two academic cases are presented, one simulat-
ing the intersection of two discontinuities in X-shape
subjected to different boundary conditions on the lat-
eral faces to capture the jump in displacement/stresses
across and at the intersection discontinuities. Sec-
ond example simulates a Triaxial 3D loading on the
rock block with a single discontinuity at different
inclinations. Comparisons with theoretical solution
demonstrate the accuracy and robustness of the two
numerical approaches.
2 GENERAL DESCRIPTION OF NUMERICAL

APPROACHES
2.1 Equivalent continuum approach: Multilaminate

model
The model formulation is composed of two main parts: Figure 1. General formulation of multilaminate model.
the continuum (matrix) and the discontinuities, each
of them defined by elastic-visco-plastic models which • Elastic-perfectly plastic behavior defined by the
are then combined additively in terms of strains. Model (i) (i)
elastic stiffnesses: normal (kn ) and tangential (kt )
is based on Caballero et al. (2009), it exhibits up to four and the strength properties: apparent cohesion (c(i) )
hyperbolic yield surfaces, one for the continuum with and friction angle (φ(i) ) for each set of discontinu-
particular hardening/softening laws and flow rules, ities.
plus up to three more for limiting the stresses on the • Non-associated flow rule (P (i) ) in order to avoid
discontinuity planes. Specific details can be found in excessive dilatancy upon shear sliding of the dis-
Caballero et al. (2009) and González, et al. (2015). continuity plane.
Figure 1 illustrates the general formulation of the νρ(i)
• Viscoplastic strain (ṫcε ) following the classical
model. As usual in multilaminate formulations (e.g.
Zienkiewicz et. al., 1977), the static constrain is Perzyna formulation in terms of rates, with partic-
assumed, i.e the local stresses on the plane (tσ = [σn , ular viscoplastic parameters for each discontinuity
τs , τt ]) are simply assumed equal to the projection of set as the viscosity (η(i) ) and the Perzyna exponent
the global stress tensor (σ) on that plane by using the (N (i) ).
stress transformation matrix (T ). Matrix behavior is characterized by:
Each set of discontinuities is characterized by:
• Hyperbolic yield surface (F) expressed in terms of
• Dip direction (α) and dip angle (β) of the normal to the three stress invariants (p , J , θ).
the plane (i). • A non-associated flow rule (P) in order to reducing
• Loading function (F (i) ) of the hyperbolic type in volumetric dilatancy for high compressive confine-
terms of normal and shear stresses on the plane. ments.
256
• Hardening and softening behavior controlled by the
evolution of the strength parameters (c , φ and pT )
in terms of the deviatoric plastic strain.
• Viscoplastic strain (ε̇vp Matrix ) following the classical
Perzyna formulation in terms of rates, with vis-
coplastic parameters for the matrix as the viscosity
(η) and the Perzyna exponent (N ).
As usual in multilaminate formulations, the strain
rate of the plane is converted to a work-equivalent
strain tensor rate of the continuum, which turns out
related by the transposed matrix used for projecting
the stress tensor, T (i) .
The overall strain tensor rate of the system is then
obtained by simple addition of those of the continuum Figure 2. Sum of the nodal enrichment functions in a typical
triangular element cut by the interface h .
plus the discontinuity families. The same procedure is
applied to determine the elastic strain of the system. As for the elements cut by the interface (K ∩ h =
The global strain tensor (ε) is then obtained as the ∅) the displacement field is written as a linear combi-
sum of the contributions of the elastic and visco-plastic nation of (1) plus a linear combination of n enrichment
strains. functions,
A stress-prescribed scheme is used for the numeri-
cal integration of the Multilaminate model following
the proposal by Caballero et al., (2009). The model
can be applied to materials exhibiting rate-dependent There are various possibilities for the enrichment
behavior, but it can also be used to recover an inviscid shape functions to capture the strong discontinuities.
elastoplasticity solution when stationary conditions In this work the following nodal enrichment functions
are reached. is applied for simplicity,
2.2 Discrete approach: XFEM

The eXtended Finite Element Method (XFEM) is an e.g., it uses the product of the shifted Heavyside
implicit representation method to capture disconti- function and the standard shape functions. As an illus-
nuities in the mesh using the concept of level-set tration, in Figure 2 the sum of the three enrichment
without a discrete localization needed by explicit rep- functions is plotted for the case of a triangular element
resentation methods. The main advantage of XFEM is in 2D.
the flexibility to define discontinuities with complex
geometries. In this case, appropriate enrichment func- 2.2.2 Matrix form
tions have to be introduced near the discontinuity so For the sake of simplicity the elementary matrix for
as to capture the discontinuities in the displacement the linear case is written as,
field. Mathematical details of the XFEM formulation
employed in this work and their implementation are
given in Das et al., (2015) and Das et al., (2016).
The most important concept in XFEM method where the index r refers to the standard degrees of
is ‘enrichment’ which means that the displacement freedom and the index a to the enrichment ones.
approximation is enriched (incorporated) by addi- Computation of matrices Kαβ proceeds as usual:
tional problem-specific functions. For example, for
strong discontinuities such as rock joint/faults mod-
eling, ‘Heaviside function’ is used to enrich nodes
whose support element is cut by the joint/fault (Suku-
mar et al., 2003) whereas the near tip asymptotic
functions are used to model the crack tip singularity
(Sukumar et al., 2003). Each matrix block reads,
2.2.1 Enrichment functions
The displacement field on the standard elements is
written as, where npe is the number of nodes per element (4 for
tetrahedra and 8 for hexahedra), α refers either to r
(regular) or d 1 , d 2 , and d 12 (enrichment) functions,
and the matrices Bi , i = 1, . . . , npe are computed as
usual with the spatial derivatives of the shape func-
α,β
Denoting by Nj , j = 1, . . . , d + 1 the standard basis tions. Computation of surface integrals KS in (5) are
functions or shape function for an element. explained in detail in Das et al., (2015).
257
Figure 5. FE Geometry of example 1: (a) Multilaminate
material to represent intersecting faults; (b) Fault planes in
IE-FEM.
Figure 3. Zero thickness interface element inserted into the
continuum FE mesh. code. In this section, two academic examples are pre-
sented in order to highlight advantages and drawbacks
of both approaches.
The first example simulates the intersection of two
faults in X-shape subjected to different boundary con-
ditions on the lateral faces to capture the jump in
displacement/stresses across and at the intersection
fault planes. Second example simulates a Triaxial 3D
loading on a rock block with a single discontinuity
Figure 4. Stresses and relative displacements in at different inclinations. Comparisons with theoretical
zero-thickness interface element formulation (IE-FEM).
Adapted from Garolera et al., (2013).
solution demonstrate the accuracy and robustness of
the two numerical approaches.
2.3 Discrete approach: Zero-thickness interface 3.1 Intersection of two faults: X-shape
elements
In this example, a column with two orthogonal inter-
The more classical discrete representation of discon- secting fault planes as shown in Figure 5 is studied.
tinuities is the interface element method (Goodman The width, length and height of the column are
at al., 1968; Gens et al., 1998) in which the discon- 10m × 10m × 20m respectively. Mechanical proper-
tinuity is explicitly discretized with special elements ties of rock mass and faults stiffness parameters are
inserted in-between element faces/edges. Mathemati- the following: Young modulus (E) is 10 GPa, Pois-
cal details of the interface element method (IE-FEM) son ratio (ν) is 0.0, fault normal and shear stiffness
employed in this work and their implementation is are KN = 1e4 GPa/m, and KT 1 = KT 2 = 1e-6 GPa/m
given in Garolera et al., (2013) and Garolera et al., respectively.
(2014). In order to create a jump in stresses and dis-
Zero-thickness joint or interface elements are finite placements through the faults, two different types
elements introduced between adjacent continuum ele- of boundary conditions on the lateral surfaces have
ments, with the special feature that they have one less been prescribed as shown in Figure 6. Prescribed
dimension than the standard continuum elements, that displacements are of δx = δy = 0.050005 m. These dis-
is, they are lines in 2D, or surfaces in 3D (Figure 3).The placements create a direct shear state of movement
integration of these elements is done through a local with respect to the faults planes. In case-1 (Fig. 6a)
orthogonal coordinate system defined on the interface a displacement of 0.5 × δy is applied in Y-direction
line or surface. on the bottom right face which creates an asymmetric
The interface constitutive behavior is formulated in movement with regards to the fault plane XZ, while in
terms of the jump of the main variable across the mid- case-2 (Fig. 6b), asymmetric movements are created
plane of the interface, and the corresponding force- with respect to both fault planes (XZ and YZ). These
type conjugate variable. In the standard mechanical configurations have been created in order to evaluate
problem, those variables are the normal and tangential the ability of the numerical approaches to represent
components of the relative displacements, and their the correct kinematic of the blocks.
counterpart stress tractions (Figure 4).
3.1.1 Results of Case-1
Distributions of lateral displacement profiles for the
3 NUMERICAL EXAMPLES: RESULTS AND case-1 are plotted in the Figure 7 which replicates the
DISCUSSIONS applied boundary conditions showed in Figure 6a. A
sudden jump in displacements field across the faults is
Discrete approaches using XFEM and Zero thickness well captured using discrete IE-FEM (Fig. 7a). Using
interface elements (IE-FEM) as well as the equivalent multilaminate model a smoothing jump is observed
continuum approach using the multilaminate model (Fig. 7b), which is a function of the element size
have been implemented in an in-house-Finite element affected by the faults.
258
Figure 6. Lateral boundary conditions of example 1. (a)
Case-1, (b) Case-2.
Figure 8. Lateral displacements along the lines ABC and

DEF (Figure 7) for case-1 (inset showing zoomed view
of displacement jump on faults). (a) X-Displacement, (b)
Y-Displacement.
Figure 7. Lateral displacements for case-1 (deformed mesh

with 10x factor). (a) Discrete IE-FEM, (b) Multilaminate
model.
Figure 9. YY-Stress variation for Case-1. (a) Discrete
Computed displacement profiles along the line IE-FEM (b) Multilaminate model.
ABC (in Figure 7) which crosses the fault plane YZ
(x = 5) at point B and along the line DEF which crosses stress results using Multilaminate model (Fig. 10c)
the fault plane XZ (y = 5) at point E, are shown in are in close agreement with the IE-FEM results (Fig.
Figures 8a and 8b, respectively. In the inset of Fig- 10a), however, XFEM results show a little dispersion
ure 8 there is a zoomed view showing the theoretical (Fig.10b). A revision of the numerical implementation
jump in displacements. It is noted that IE-FEM shows of XFEM approach is therefore required.
a perfect agreement with theoretical solution (X-jump
of 5E-06 m and Y-jump of 2.5E-06 m). Multilami- 3.1.2 Results of case-2
nate model shows an average displacement through Lateral displacement profiles for the Case-2 are plot-
the fault planes which is accurate enough to capture ted in the Figure 11, which replicates the applied
the kinematic movement of the blocks. boundary conditions showed in Figure 6b. It is noted
Due to asymmetric movement of lateral surfaces that multilaminate model (Fig. 11b) predicts a smooth
in the Y-direction, the resulting stresses in this direc- displacement field through the fault planes and is able
tion are non-uniform as shown in Figure 9. IE-FEM to capture the kinematic movement of the blocks which
predicts a stress jump of 25 MPa between the blocks are asymmetric with respect to both fault planes.
separated by the fault YZ (Fig. 9a). Multilaminate Computed displacement profiles along the line
model also can predict the correct stress distribution ABC and line DEF (in Figure 11), are shown in Fig-
at the left and right blocks (Fig. 9b) with an average ures 12a and 12b, respectively. Again, it is observed
stress along the elements representing the fault YZ. that IE-FEM shows a perfect agreement with the theo-
Normal stress distribution on the fault planes is retical jump solution (X-jump of 1E-05 m and Y-jump
also non-uniform as shown in Figure 10. Results of of 5.0E-06 m) and Multilaminate model shows an
discrete XFEM approach are included in Figure 10 average displacement through the fault planes.
(panel b); they are taken from Das et al. (2016). Nor- Resulting stresses in X and Y direction are non-
mal stress of 25 MPa is observed on the face of fault uniform for this case. Figure 13 shows the stress
plane XZ located at the right side of the intersecting variation in X direction. It is noted that results of both
faults, all other faces on faults plane experienced uni- IE-FEM and Multilaminate model are in good agree-
form normal stress of 50 MPa. It is noted that normal ment. An abrupt stress jump of 50 MPa between the
259
Figure 12. Lateral displacements along the lines ABC and
Figure 10. Normal stress on fault planes for Case-1. (a) DEF (Figure 11) for Case-2 (inset showing zoomed view
Discrete IE-FEM, (b) Discrete XFEM and (c) Multilaminate of displacement jump on faults). (a) X-Displacement, (b)
model. Y-Displacement.
Figure 13. XX-Stress variation for Case-2. (a) Discrete

IE-FEM (b) Multilaminate model.
Figure 11. Lateral displacements for Case-2 (deformed

mesh with 10x factor). (a) Discrete IE-FEM (b) Multilam-
inate model.
blocks separated by the fault XZ is predicted by IE-

FEM (Fig. 13a), while multilaminate model predicts a
smoothed jump of stress (Fig. 13b).
Finally, normal stress distribution on the fault
planes is shown in Figure 14. Using IE-FEM a jump of
the normal stress of 50 MPa is observed on both fault
planes at the intersection between them (Fig. 14a). Figure 14. Normal stress on fault planes for Case-2. (a)
Discrete IE-FEM, (b) Multilaminate model.
Using multilaminate model an average normal stress
is predicted at the elements located in the intersec-
tion between both fault planes allowing a transitional with the minor principal stress direction (σ3 ), inclina-
normal stress variation on the fault planes. tion is varied between 20◦ to 85◦ . A hexahedral mesh
of 9537 nodes and 8192 elements is employed.
3.2 Strength of a fault in triaxial compression The shear strength of the fault is defined by
the Mohr Coulomb law in elastic perfectly plas-
In this example the numerical simulation of a triax- tic conditions. Fault properties are normal and tan-
ial test on a rock block with a single fault at different gential stiffness of KN = 1e6 MPa/m and, KT = 1e4
inclinations is performed. Figure 15 shows the finite MPa/m, apparent cohesion cf = 1 MPa and friction
element geometry using multilaminate and XFEM angle φf = 15◦ . Intact rock is elastic with a Young’s
approaches. The fault plane is inclined at an angle β modulus of 1e4 MPa and Poisson’s ratio of 0.3.
260
This tendency could be an effect of the proximity of
the elements representing the fault plane to the bound-
ary where lateral confinement is applied. A different
behavior could be expected if the fault inclination is
pivoting at the center of the column.
4 DISCUSSION
Academic examples presented in previous section

Figure 15. FE Geometry: (a) Multilaminate material; (b) show that both discrete and equivalent continuum
Fault planes with different angles in XFEM. approaches are able to represent the behavior of
discontinuities in a good approximation.
Main advantage of continuum approach is their sim-
plicity, it is adequate for most practical analyses, and
it delivers reasonably accurate results at a lower cost
than is needed by discrete models (see also Lhasa
et al, 2012). However, the response is smeared and
the results are strongly dependent on the element size
particularly in case of softening materials. In addi-
tion, regular meshes are recommended to discretize
the elements representing the discontinuity.
Discrete approach is theoretically more suit-
Figure 16. Comparison of Fault strength obtained from
analytical solution and XFEM and multilaminate analyses. able to capture the localized opening/closing/sliding
and stress representation along discontinuity planes.
The model is constrained at the bottom surface in XFEM method allows an implicit representation of
normal direction; a uniform vertical displacement δz the discontinuities without changing the background
of –0.024 m is prescribed on the top of the model and a mesh discretizing the continuum rock; this is their
constant confinement of σ3 = 5MPa is applied on the main advantage over the IE-FEM. However, discrete
lateral surfaces. approaches are usually more demanding in the sense
The strength of the rock sample with fault is con- that they need more specialized software. Additional
sidered to be the minimum pressure that has caused degrees of freedom (IE-FEM) and enrichment nodes
the slip of the fault plane based on Mohr-Coulomb (XFEM) are introduced to capture the discontinu-
criterion (onset of yielding). It is defined as, ous behavior of the displacement field, increasing the
number of elements and nodes of the problem.
For Oil & Gas industry a combined approach is rec-
ommended. Discontinuities (faults and fractures) will
be represented differently depending on the scale with
respect to the grid size (see Figure 17). Discontinu-
Results of the rock mass strength using XFEM ities larger than the grid cell or element size (typically
and multilaminate model are plotted in Figure 16. faults) will be represented using discrete approaches
Rock mass strengths obtained from both numerical (XFEM or IE-FEM). Discontinuities smaller than the
approaches are in good agreement with those of the grid cell size (typically fractures) will be represented
analytical solution given in equation (7). There is a with the equivalent continuum approach (multilam-
threshold value of the fault inclination under which inate model), which can represent several sets of
the plane is not activated (this threshold value is the fractures contained in a cell in a smeared way. This
friction angle), and, immediately after that thresh- combined methodology has been also recommended
old, strength decreases with inclination increases until in the literature (e.g. Nguyen and Selvadurai, 1995;
reaches the minimum strength at an angle of, Bai et al. 1995) to simulate coupled Hydro-Mechanical
behavior of fractured reservoirs.
5 CONCLUSIONS
Computed errors predicted by XFEM and multil-
aminate solutions are showed in Table 1.An error lower The basic features of the equivalent continuum and
than 2% was obtained using XFEM, while multilam- discrete approaches have been presented and discussed
inate model gives larger errors, which are in general using two examples analysis. The applicability and rel-
lower than 6. In spite of the higher errors, multilami- ative merits and limitations of both of the approaches
nate model can deliver reasonably accurate results. It for the simulation of jointed rocks were presented. It
is noted in Table 1 that using multilaminate model, the was observed that both the approaches are reasonably
error increases as the fault inclination also increases. good in predicting the real response.
261
Table 1. Results of theoretical solution against XFEM and Multilaminate analyses.
Strength Error
Inclination (β) Theoretical Multilaminate XFEM Multilaminate XFEM

(Deg) (MPa) (MPa) (MPa) (%) (%)
20 27.60 26.92 27.12 2.44 1.71

30 10.08 9.77 10.00 3.13 0.84
40 6.98 6.74 6.94 3.51 0.55
45 6.39 6.14 6.37 3.88 0.43
50 6.13 5.87 6.09 4.17 0.70
52.5 6.10 5.83 6.07 4.34 0.53
60 6.39 6.11 6.37 4.42 0.40
70 8.07 7.64 8.04 5.29 0.36
80 14.36 13.58 14.35 5.43 0.10
85 27.60 26.07 27.59 5.53 0.03
set approach. 1st International Conference on Energy

Geotechnics ICEGT 2016.
Deb, D., and Das, KC. (2010) Extended finite element
method for the analysis of discontinuities in rock masses.
Geotechnical and Geological Engineering, 28, 643–659.
Garolera, D., Aliguer, I., Segura, J.M., Carol, I., Laksh-
mikantha, M.R. and Alvarellos, J. (2013). “Zero-thickness
interface elements with H-M coupling, formulation and
Figure 17. (a) Fault representation using XFEM. (b) Frac- applications in geomechanics”. In XII International Con-
ture set representation using Multilaminate model. ference on Computational Plasticity. Fundamentals and
Applications COMPLAS XII. E. Oñate, D.R.J. Owen, D.
Peric and B. Suárez (Eds). (pp. 1372–1383).
Garolera, D.,Aliguer, I., Carol, I., Segura, JM.,Lakshmikantha,
The selection of one approach or the other is largely M R. and Alvarellos, J. (2014). Hydro-mechanical cou-
influenced by the scale of the problem. For Oil & Gas pling in zero-thickness interface elements, formulation
industry, behavior of faults can be modeled using a dis- and applications in geomechanics. Rock Engineering and
crete approach when the number of faults is relatively Rock Mechanics: Structures in and on Rock Masses
small and their scale is larger than the background (EuroRock-2014).
numerical mesh discretizing the domain. On the other Gens, A., Carol, I., and Alonso, E.E., (1988). An interface
hand, equivalent continuum formulations are more fit- element formulation for the analysis of soil-reinforcement
ted to reproduce the behavior of large numbers of interaction. Computers and Geotechnics, 7:133–151.
González, N.A. et al. (2015). Multilaminate model for
fractures and/or fractures at a smaller scale than the
fractured rocks. Repsol-IBM internal report.
numerical discretization. Goodman, R. E., Taylor, R. L., & Brekke, T. L. (1968). A
model for the mechanics of jointed rock. Journal of Soil
Mechanics & Foundations Div., 94 (SM3): 637–659, 1968.
REFERENCES Latha, G.M. & Garaga, A. (2012). Elasto-plastic analysis
of jointed rocks using discrete continuum and equivalent
Bai, M., Roegiers, J.C., and D. Elsworth. (1995). Porome- continuum approaches. Int. J. Rock mechanics. 53(2012)
chanical response of fractured-porous rock masses. Jour- 56–63.
nal of Petroleum Science and Engineering, 13, 155–168. Nguyen,T.S., and Selvadurai, A.P.S. 1995. Coupled thermal-
Caballero, A. Garolera, D., Carol, I. and Gens, A. (2009). mechanical-hydrological behaviour of sparsely fracture
Viscoplastic multilaminate model for jointed rock with rock: implications for nuclear fuel waste disposal. Inter-
stress-prescribed using a stress-prescribed LEGI scheme. naional Journal of Rock Mechanics and Mining Sciee &
International Conference on Rock Joints and Jointed Rock Geomechanics Abstracts, 32(5), 465–479.
Masses, Tucson, Arizona, USA, January 7–8, 2009. Perzyna, P. (1966). Fundamental problems in viscoplasticity.
Bai, M., Roegiers, J.C., and D. Elsworth. 1995. Poromechan- Advances in Applied Mechanics, 9(C):243–377, 1966.
ical response of fractured-porous rock masses. Journal of Sukumar, N., Chopp, D., Fatigue, L. (2003). Crack Prop-
Petroleum Science and Engineering, 13, 155–168. agation of Multiple Coplanar Cracks with the Coupled
Das, K., Ausas, R., Carol, I. Rodrigues, E. Sandha, S., Vargas, Extended Finite Element/Fast Marching Method. Interna-
P.E., González, N.A., Segura-Serra, J.M., Lakshmikantha, tional Journal of Engineering Science, 41(8), 845–869.
M.R. and Mello, U. (2015). Discrete Modeling of Multiple Zienkiewicz, O. C. & Pande, G. N. (1977). Time dependent
Discontinuities in Rock Mass using XFEM. ARMA 15- multilaminate model of rocks: a numerical study of defor-
ID46. mation and failure of rock masses. Int. J. Numer. Anal.
Das, K., Carol, I. Sandha, S., Vargas, P.E., Rodrigues, Meth. Geomech. 1, No. 3, 219–247.
E., González, N.A., Segura-Serra, J.M., Lakshmikantha,
M.R. and Mello, U. (2016). Modeling of discrete inter-
secting discontinuities in rock mass using XFEM/Level
262
Thermal-hydraulic modelling a Canadian deep geological repository
P. Abootalebi & G.A. Siemens

GeoEngineering Centre at Queen’s-RMC, Kingston, ON, Canada
ABSTRACT: Within Canada there are more than 2.5 million bundles of spent nuclear fuel with another
approximately 2 million bundles to be generated in the future. Canada, and every country around the world
that has taken a decision on management of spent nuclear fuel, has decided to on long-term containment and
isolation of the fuel within a deep geological repository. At depth, a deep geological repository consists of a
network of placement rooms where the bundles will be located within containers and surrounded by an engineered
barrier system. Amongst other design aspects, the engineered barriers will transfer the thermal energy from the
spent nuclear fuel to the surrounding geosphere. The barriers will be placed in a complex thermal-hydraulic-
mechanical-chemical environment. The environment will include competing gradients of groundwater pressure
driving moisture into the repository and thermal gradients driving moisture out. A current design criterion of
the repository is to keep temperatures of the container surface below 100◦ C. Therefore the thermal properties
of the engineered barriers are a critical component of the system in terms of its efficiency and size. Barrier
materials will be at variable saturation levels and temperatures over their design life. An experimental program
was initiated to measure the thermal properties of a number of potential barrier materials under variable moisture
and temperature conditions. In this paper the experimental methodology for thermal property measurement is
presented along with preliminary results. Preliminary modeling of the Canadian concept for the deep geological
repository is also presented. The results show the impact of coupled thermal-hydraulic properties on the surface
temperature of the container.
1 INTRODUCTION of the used fuel container below 100◦ C (Maak 2006).

Therefore the thermal properties of the engineered bar-
Global energy needs continue to rise along with the riers are a key attribute of the system. The engineered
need for carbon-reduced energy sources to limit cli- barriers, composed of bentonite-based materials, will
mate change effects. One viable energy source is be at variable saturation levels and temperatures over
nuclear power, which generates power using a carbon their design life. Thus, it is necessarily to determine
neutral source. Half of the electricity requirements of thermal properties of engineered barriers over a wide
the province of Ontario, Canada (pop. 13 million) are range of moisture contents (degree of saturations) and
met with nuclear power. Along with the benefits of variable temperatures (25◦ C and 80◦ C).
nuclear power comes the ethical and environmental In this paper, the experimental methodology for
responsibility to safely care for the waste products. thermal property measurement is presented along
Within Canada there are more than 2.5 million bundles with preliminary results. Preliminary modeling of the
of spent nuclear fuel with another approximately 2 mil- Canadian concept for the deep geological repository is
lion bundles to be generated in the future. Canada, and also presented. The results show the impact of coupled
every country around the world that has come to the thermal-hydraulic properties on the container surface
decision on a long-term solution for nuclear waste, has temperature over the lifetime of the deep geologic
decided deep geological repository. A deep geological repository.
repository (Figure 1) consists of a network of place-
ment rooms at depth where the bundles will be located
within containers and surrounded by an engineered
barrier system. 2 METRIALS AND METHODS
Amongst other design aspects, the engineered bar-
riers will transfer the thermal energy from the spent 2.1 Material description
nuclear fuel to the surrounding geosphere. The envi- The engineered barriers that surround the used fuel
ronment would likely include competing gradients of containers within the Canadian concept include highly
groundwater pressure driving moisture into the reposi- compacted bentonite, dense back fill and gap fill.
tory and thermal gradients driving moisture away from The container is surrounded by highly compacted
the used fuel container. A design criterion of the repos- bentonite, which is compacted to a dry density of
itory system is to keep the external surface temperature 1.7 Mg/m3 . Between adjacent containers is placed a
265
Figure 2. Compaction results to achieve dry density
ρd = 1.7 Mg/m3 in terms of compaction pressure and force
versus degree of saturation and gravimetric water content.
Figure 2. Compaction force is a maximum at low sat-

uration and decreases asymptotically from 90 kN to
over 40 kN at Sr = 100%.
Two test configurations are implemented for mea-
suring thermal properties depending on the moisture
Figure 1. a) Canadian concept of a deep geological reposi- content of the specimen. For higher moisture contents a
tory and b) Mark II placement room (after NWMO 2015). two-sided test is used (Figure 3a) and at lower moisture
contents a one-sided test methodology is used (Fig-
spacer block made of dense backfill, which is com- ure 3b). In the two-sided test, two 50 mm-diameter by
posed of a mixture of crushed granite, illite clay and 20-mm pucks are compacted and the sensor is placed
bentonite, compacted to a dry density of 2.1 Mg/m3 . between. A nominal normal stress is applied through
The gap between bedrock and buffer box will be the frame to ensure good contact between the soil and
filled with gap fill material which is 100% bentonite the sensor. For one-sided tests a special mould (Fig-
composed of pellet size grains in the matrix with a ure 3b) was constructed to allow for thermal property
dry density of 1.41 Mg/m3 . The gap fill and highly measurement without removal from the compaction
compacted bentonite specimens described within this mould. In some materials at low saturations, excessive
paper were prepared using National Standard Ben- disturbance occurred during removal from the two-
tonite acquired from Bentonite Performance Minerals sided compaction mould. In Figure 3b the one-sided
LLC. test compaction mould is shown with a slot cut into the
side to allow for placement of the thermal conductiv-
ity sensor on the top face of the compacted specimen.
2.2 Specimen preparation
During compaction the slot is covered to avoid soil
Specimens are prepared using a methodology from spillage and loss of confinement. Figure 3b shows
Man et al. (2011) to achieve reliable moisture con- the compacted specimen with the thermal conductiv-
tent and density conditions. The constituent bentonite ity sensor in place. A block of insulation is placed
is placed within an oven at 105◦ C for at least 24 hours. above the specimen to limit thermal energy loss and
The material is removed from the oven, sealed, and an applied normal stress encourages contact between
allowed to come to thermal equilibrium with the lab- the sensor and the specimen. This type of test is termed
oratory. The mass of bentonite required for specimen a ‘one-sided test’ as the thermal conductivity sensor is
preparation is removed and placed in a mixing bowl. only in contact with the test specimen on one side.
Water is added to the bentonite via misting with a spray A third methodology is used for testing thermal
bottle to achieve the target moisture content. Follow- properties at elevated temperature. The same com-
ing mixing, the soil is placed within two sealed bags paction method was followed for both one-sided and
in a fridge for at least 48 hours for moisture equilib- two-sided tests, but in this case the sample is wrapped
rium. Each weekday the soil is mixed inside the bag to up with plastic wrap and placed in ziploc (Figure 3c).
encourage suitable moisture distribution. The bag is sealed around the wire with tuck tape. Then
Static compaction was selected for preparing ther- the sample holder along with the specimen is placed
mal conductivity specimens in order to achieve uni- in oven at 80◦ C along with a bowl of water to keep
form density over the wide range of moisture con- relative humidity close 100% and reduce moisture gra-
tents. However, due to high target dry density value dients. Thermal property measurement is performed
(ρd = 1.7 Mg/m3 ) required for highly compacted ben- in a standard manner after specimen equilibrates with
tonite, low saturation specimens were compacted in a the elevated thermal environment. Following place-
General Electric 30,000 lb (130 kN) electromatic uni- ment in the oven thermal property measurements are
versal testing machine. Typical results in terms of performed periodically until thermal equilibrium is
compaction pressure and force versus degree of sat- confirmed. Figure 4 shows an example of tempera-
uration and gravimetric water content are plotted in ture drift detected by sensor over the 80 second test
266
Figure 4. An example of temperature drift detected by sen-
sor during thermal property measurement at different times
following placement in the oven.
where Tavg (τ) is the average temperature increase of

the specimen surface in contact with the sensor sur-
face, Ti is the constant temperature difference that
develops over the thin insulation layer covering the
sensor, R0 is the resistance of the disk prior to heating
at time (t) = 0 and α is the temperature coefficient of
resistivity which is well-known for nickel.
The temperature difference across the sensor’s insu-
lating layer, Ti , becomes constant over a short period
of time t, which can be estimated as:
where δ is the thickness of the sensor’s insulating

layer and κj is the thermal diffusivity of the insulating
Figure 3. Test procedures for measuring thermal properties: layer. Generally t is less than 10 seconds in the
a) one-sided test and b) two-sided test and c) 80◦ C test. experiments.
By introducing dimensionless integration variable
during 4 different time intervals. For this test, the time termed the characteristic time ratio, τ:
required to reach thermal equilibrium was more than
three hours. After the specimen is taken out of the oven
the final moisture content of the specimen is measured
and used to plot the results.
where t is the time measured from the start of transient
recording, κ is the thermal diffusivity of the specimen
2.3 Thermal properties test interpretation and α is radius of sensor, it is possible to solve the
differential equation of heat conduction in an isotropic
The device used in the thermal property testing is a Hot material. Then temperature increase can be determined
Disk Thermal Constants Analyzer (Hot Disk 2014). as a function of characteristic time:
The sensor (visible in Figure 3b) consists of an electri-
cal conducting pattern in the shape of a double spiral
etched out of the thin sheet of nickel. The nickel plate
is covered on both sides with thin sheets of insula-
tion. The sensor applies a constant thermal energy to
the specimen and measures the surface temperature
change. The experimental methodology and interpre-
tation framework allows for measurement of thermal
conductivity, thermal diffusivity as well as volumetric
heat capacity during a single test.
The temperature increase of the sample surface is
measured by monitoring the total resistance of the hot where P0 is power output of hot disk sensor, K is
disk sensor: thermal conductivity of sample, m is a number of
concentric rings of sensor and I0 is the first kind mod-
ified Bessel function and D(τ) is geometric function
of sensor.
267
Figure 5. Thermal conductivity measurements of highly Figure 6. Thermal conductivity measurements of gap fill as
compacted bentonite as a function of degree of saturation a function of degree of saturation at 20◦ C and 80◦ C.
at 20◦ C and 80◦ C.
and continued by considering a representative cross-
By knowing the relationship between t and τ, T (τ) section. Comsol version 5.2 was used as a finite
can be plotted as function of D(τ) which result in element software to model heat transfer.This is the first
a straight line and from slope of that line which is time numerical simulations of the current Canadian
P0 /(π3/2 aK), thermal conductivity of the sample can concept for deep geologic repository have considered
be calculated. However the proper value of τ is gener- the effects of moisture on the thermal response. The
ally unknown since it is related to thermal diffusivity of critical design consideration is surface temperature
the sample. Thus, the final straight line from which the of the container. These models with constant thermal
thermal conductivity is calculated, is obtained through properties are used to bound the maximum and mini-
an iterative process. mum temperature of the container at critical times of
its use.
3 EXPERIMENTAL RESULTS
4.1 Buffer box model in air
Selected thermal conductivity results are plotted in The finite element mesh of the first model is shown
Figure 5 and Figure 6 for highly compacted bentonite in Figure 7a. Due to the symmetric shape of container
and gap fill respectively. The blue circles are room half of the container in buffer box was modelled. In the
temperature data and the red diamonds are the ther- buffer box model, there are two materials, which are
mal conductivity results for tests performed at 80◦ C. highly compacted bentonite block and the outer cop-
Room temperature tests were performed at 11 degree per surface of the used fuel container. Perfect contact
of saturations in triplicate. The 80◦ C tests were per- between the container surface and the highly com-
formed at 4 degree of saturations again in triplicate. pacted bentonite is assumed. The initial temperature
The test results show thermal conductivity increases is set to 20◦ C of both materials. Boundary conditions
with increasing saturation in both highly compacted for the transient model include heat energy applied to
bentonite and gap fill. Good repeatability is indi- the inside of the container and the temperature at the
cated in the results with R2 > 0.97 for both materials. outside of the buffer box. The heat source was modeled
Comparing the Sr = 0% and Sr = 100% results shows as a decay function based on the number of bundles
indicates a 2-3 fold increase in thermal conductivity within the container. The outside of the buffer box was
from dry to saturated. In highly compacted bentonite set to 20◦ C. Results are given in Figure 7 including
the relationship is linear. A Botzmann sigmoidal curve the temperature regime of the model at maximum con-
was used to fit the gap fill results. The gap fill, com- tainer surface temperature (Figure 7b) and temperature
posed of bentonite pellets is prone to non-uniform versus time (Figure 7c). To verify the model, thermal
distributions of saturation given the high density of properties from a Nuclear Waste Management Orga-
the pellets and the low overall density of the material. nization report (NWMO 2015) were initially used.
The effect of temperature is notable in the gap fill Figure 7b shows maximum container surface temper-
but is undetectable in the highly compacted bentonite. ature after 20 days was equal to 37.1◦ C. Figure 7c
The gap fill results indicate a 15–30% increase in compared these results for container surface tempera-
thermal conductivity over the 60◦ C increase in tem- ture for 50 years with NWMO (2015) and showed that
perature. The highly compacted bentonite results show results are in agreement.
the same trend and quantitative measurements were After verification, thermal properties of highly
recorded in the 20◦ C and 80◦ C tests. compacted bentonite based on experimental data for
four different degrees of saturation were applied to the
buffer box and the container surface temperature was
4 THERMAL-HYDRAULIC MODEL modelled for 30 days. These results bound the pos-
sible maximum and minimum temperatures for this
The thermal-hydraulic modelling efforts began with initial configuration. Results for this model show-
modeling an isolated component of the repository ing the impact of saturation on the container surface
268
Figure 8. Container Surface Temperature over 30 days-
buffer box model in air.
boundary conditions. Figure 9b shows mesh developed

for repository and bedrock around, again extremely
fine mesh was used for buffer box and container com-
pared to surrounding geosphere. The height and width
of the section were assumed to be 1000 m and 10 m
respectively. The placement room at a depth of 500 m
consists of two one-quarters of the buffer box on top of
each other with the quarter of spacer block is attached
to the side of each buffer box. A thin layer of gap
fill material with thickness of 0.1 m was applied to
the perimeter of both buffer boxes. An initial temper-
ature that varied across the depth of the model from
5◦ C to 21◦ C (Guo 2010) was assumed for whole sys-
tem. Constant temperature equal to 5◦ C and 21◦ C were
applied at the surface and bottom of model during the
transient phase. These symmetric boundary conditions
were applied to the four left vertical boundary surfaces.
The heat source from the decaying waste was applied
to the containers and container surface temperature
observed over the time.
At first, the thermal properties of highly compacted
bentonite, dense back fill, gap fill and rock were
assumed to be the same as a report (NWMO 2015)
Figure 7. Buffer box model in air: a) Mesh developed at 0% degree of saturation for verification. The model
through half of container b) Maximum container surface tem- was run model for 100 years and the results are com-
perature reached after 20 days equal to 73.1◦ C c) Container pared in Figure 9c. The blue triangles and the green
surface temperature over 50 years compared with report by circles previous data on simulations performed by
NWMO (Gue 2015). Comsol and Ansys software respectively. The red line
is data from the current model, which agrees with the
temperature are plotted in Figure 8. The direct relation- previous simulations and verifies the results.
ship between saturation and thermal conductivity is For considering the effect of a static moisture
reflected in the model results. In the model where ther- regime on the surface temperature of the container, the
mal conductivity associated with Sr = 0% was applied experimental data for all materials, including highly
the maximum surface temperature of the container was compacted bentonite, dense backfill and gap fill at
36.6◦ C. When increased to Sr = 100% the maximum 100% were used as for the thermal properties. Mod-
temperature decreased to 29.2◦ C. els were run for 100 years assuming and plotted as
the blue line in Figure 10. This serves as a theoreti-
4.2 Buffer box model in deep geological repository
cal lower bound on the modelled temperatures for the
The geometry and dimensions of one tunnel within configuration. Figure 10 shows that if all the material
the repository in room placement concept is shown are assumed saturated, the maximum container sur-
in Figure 1b including the locations of the used fuel face temperature is 86.9◦ C, which is reached after 52.5
container, highly compacted bentonite, dense back fill years (blue line).
and gap fill within the insitu rock. Because of the sym- Further models considered the effect of degree of
metric shape of the part of the repository where the saturation of the highly compacted bentonite on the
containers placed, a section of the repository based container. Focus was placed on this material since
on Figure 1b was selected which represents a deep highly compacted bentonite surrounds the initially hot
geological repository at sedimentary rock. The section container and there is more potential for this mate-
is shown in Figure 9a as well as the dimensions and rial to dry out from the thermal gradients compared
269
Figure 10. Container Surface Temperature over 100 years-
buffer box model in deep geological repository.
than 100◦ C, the characterization of thermal proper-

ties of engineered barriers as a function of moisture
content and temperature is a key component of the
modelling. More efficient thermal transfer to the geo-
sphere could impact the overall design and have direct
impacts to the cost of the repository. In this paper, ther-
mal properties of engineered barriers were reported
and then used in thermal models of the Canadian con-
cept of the deep geologic repository. The results show
the impact of saturation on the maximum container
temperature and can be used to bound the potential
thermal regimes. Future models will consider the cou-
pled response of the repository to competing thermal
and hydraulic gradients.
ACKNOWLEDGMENT
Figure 9. Deep geological repository model: a) geome-
try and boundary condition b) Mesh developed through The support of the Nuclear Waste Management
repository c) Container surface temperature over 100 years Organization (NWMO) in this research is gratefully
compared with previous models. acknowledged.
to other materials. The results are plotted in Figure

REFERENCES
10 illustrating the increase in container temperature
associated with decreasing the degree of saturation of Guo, R. 2015. Thermal Modelling of a Mark II con-
the highly compacted bentonite. The maximum sur- tainer, Nuclear Waste Management Organization techni-
face temperature increases to 92.4˚ C after 40 years cal Report NWMO-TR-2015-06.
for Sr = 0% in the highly compacted bentonite (red Guo, R. 2010. Coupled Thermal-Mechanical Modelling
line). Increasing the thermal properties of the highly of a Deep Geological Repository using the Horizontal
compacted bentonite (associated with Sr = 20% and Tunnel Placement Method in Sedimentary Rock using
Sr = 60%) caused notable decrease in the maximum CODE_BRIGHT. Nuclear Waste Management Organiza-
tion Technical Report NWMO-TR-2010-22.
container temperature and also delayed the timing of
He, Y., 2005. Rapid thermal conductivity measurement with
the maximum temperature. Thus there are incremental a hot disk sensor: Part 1. Theoretical considerations. Ther-
benefits to configuring the repository to increase the mochimica acta, 436(1), pp. 122–129.
saturation of the bentonite barriers in both maximum NWMO. 2015. Progress Through Collaboration – Annual
temperature and timing. Report. 173 pp.
Hot Disk. 2014. Hot Disk Thermal Constants Analyser –
Instruction Manual. 135 pp.
Maak, P. 2006. Used fuel container requirements. Ontario
5 CONCLUSION Power Generation Preliminary Design Requirements.
06819-PDR-01110-10000 R02.
The current concept for Canada’s inventory of spent Man, A., Martino, J.S., Kim, CS., Priyanto, D.G. 2011.
nuclear fuel is a deep geological repository. The engi- Characterization and improving the thermal conductivity
neered barriers within repository will be under a wide of engineered clay barriers for sealing a deep geolog-
range of phenomena, one of which is transient thermal ical repository. Waste Management, Decommissioning
gradients as the energy from the used fuel containers and Environmental Restoration for Canada’s Nuclear
is transfer to the surrounding geosphere. Given design Activities, Toronto, Canada.
criterion of repository for temperatures to remains less
270
Behaviour of kaolinite and bentonite at elevated and lowered temperatures
T. Ring, C. Leo & S. Liyanapathirana

Western Sydney University, Sydney, NSW, Australia
ABSTRACT: Engineering applications which experience a change in the natural ground temperature are
increasing. There is an established a link between mechanical behaviour, temperature, and stress history. Some
models/relationships have been postulated, however, very few studies have been undertaken to compare and
contrast responses based on the distinctive soil compositions. This paper presents some initial investigations into
consolidation rate and thermal volume change for kaolinite and bentonite, two distinctive clays. Step loading tests
were performed at various overconsolidation ratios, at temperatures between 5 to 50◦ C. One temperature cycle
from 20 to 50◦ C was used to ascertain volume change. A clear influence of temperature on behaviour of both clays
was observed, though not always in the same way. Elevated temperatures resulted in an increase in consolidation
rate under normal consolidation predictably for both clay materials, however, at high overconsolidation rates,
the nature of response appears to be more complex.
1 INTRODUCTION types and procedures varied, they found consolidation

rate increased with an increase in temperature.
Engineering applications experiencing a variation of This work presents some initial findings from
the natural thermal state are increasing. These include a planned comprehensive thermo-poromechanical
high-level nuclear waste disposal (Ojovan, 2010), investigation on the influence of mineral structure as it
energy geostructures (Knellwolf et al., 2011), landfills relates to the poromechanical behaviour under various
(Abuel-naga and Bouazza, 2013), as well as advanced thermal paths. Consolidation rates and volume change
construction techniques such as soil freezing. The are the focus of this preliminary stage using a Rowe
potential for heated prefabricated vertical drains to cell and a temperature of 5 to 50◦ C. One expansive and
increase consolidation rate is also being explored one non-expansive clay were chosen for this research
(Pothiraksanon et al., 2010b). to investigate the mineralogy/structural differences
Design rules and guidelines related to thermal con- possibly contributing to different characteristics.
siderations are still limited (Laloui and Di Donna,
2011). This is a function of the complex nature
of the thermo-poromechanical behaviour. Over the
years, research has been expanding the understanding;
however, it is still evolving. 2 MATERIAL, EQUIPMENT AND
There have been a number of experimental works EXPERIMENTAL PROCEDURE
investigating the volume change produced by an alter-
ation of the thermal state (Abuel-Naga et al., 2007, 2.1 Material
Sultan et al., 2002, Burghignoli et al., 2000, Cekere- Two different clays were used in this study. They were
vac and Laloui, 2004, Cui and Tang, 2013). While commercially available under the brand names ‘Clay
the specifics of each test program varied, two general Ceram’ and ‘Trugel 100’. X-ray diffraction was used
conclusions were drawn. When the effective pressure to identify the major structures; they were kaolinite
remains constant and the sample is under drained con- and quartz, and beidellite and quartz. Beidellite is
ditions; (1) normally consolidated clays contract when colloquially known as bentonite.
heated and the magnitude of contraction decreases Due to kaolinite’s atomic and sheet structure (1:1
with an increase in OCR (over consolidation ratio) up tetrahedral to octahedral), it has a very low cation
to a point where it can become dilated, (2) the volume exchange capacity (10–100 mmol/kg), specific sur-
change induced by heating is largely non-reversible face area (5–20 m2 /g), and c-spacing (0.72 nm). This
(plastic). leads to a non-expansive and stable nature. Conversely,
Some laboratory (Delage et al., 2000, Di Donna and bentonite has a 2:1 sheet structure (tetrahedral to
Laloui, 2015) along with one field study (Pothirak- octahedral), this results in a high; cation exchange
sanon et al., 2010a), have investigated the clay consol- capacity (800–1200 mmol/kg), specific surface area
idation rate under elevated temperatures. While the soil (600–800 m2 /g), and c-spacing (1.2-2.1 nm) (Strawn,
271
2015). These result in both its value as an engineer- rebound creating an overconsolidation ratio (OCR) of
ing product and as a problematic soil with expansive 8, with the temperature set to the desired level (5, 20,
behaviour. 35, or 50◦ C).
The kaolinite and bentonite plastic limits are: 30.0 After temperature stabilization, the effective pres-
and 46.4%, and the liquid limits are: 45.5 and 453%. sure was applied in steps at: 100, 200, 400, 800, and
1200 kPa. Constant temperature was maintained for
the duration of the consolidation rate tests. Due to the
2.2 Equipment low permeability of the bentonite, two-way drainage
A 75.5 mm Rowe cell was used for all consolida- was used during testing. One-way drainage was used
tion tests. To control the test temperature the cell for kaolinite. Duration of each pressure step was con-
was fully immersed in a temperature controlled water sistent for each material; 1 hour for kaolinite and
bath. Temperature was logged separately using a dig- 24 hours for bentonite.
ital temperature probe with a resolution of 0.1◦ C. Upon completion of the test, the effective pressure
Cell and back pressures were regulated by 3 MPa was reduced to 10 kPa and temperature set to 20◦ C, the
automatic pressure-volume controllers. The vertical samples removed, final heights and moisture content
displacement was monitored by a 25 mm linear vari- recorded.
able displacement transducer. The data acquisition
system was setup to provide a displacement resolu- 2.3.2 Volume change test and specimens
tion of 0.001 mm, although a drift of ±0.002 mm was Separate specimens were created for investigating the
common. When applicable, a 1 MPa pore pressure volume change under thermal loading. Both kaolin-
transducer was employed. Clisp Studio software was ite and bentonite were prepared in the same manner,
used to automate the controllers and log the data. similar to that of the bentonite consolidation rate sam-
Calibrations were performed on all instruments at ples. Firstly, a slurry was created then placed directly
ambient temperatures prior to testing. Calibrations into the Rowe cell and consolidating to an OCR = 8.
of the complete system were also run at elevated Two-way drainage was used for sample preparation
and lowered temperatures to assess the impact on the and testing.
equipment and instrumentation. As expected the tem- Samples were prepared at a constant temperature
perature did affect some readings. Given the small of 20◦ C and pressures applied to achieve the desired
magnitude (0.0013 mm/◦ C), it was only necessary overconsolidation ratio. Once primary settlement was
to correct the vertical displacement for the volume complete the temperature was increased to 35 then
change tests. to 50◦ C, and lowered back to 20◦ C before the next
pressure step. At both temperature extremes the dis-
placement was monitored and once stabilized the test
2.3 Experimental procedure was continued. A majority of the volume change
occurred during the heating and cooling process.
2.3.1 Consolidation rate tests and specimens
To create the test specimens a slurry was mixed using
dried powder and de-ionized water to a moisture con-
3 MAIN RESULTS AND DISCUSSION
tent of 2x the liquid limit for kaolinite and 1.5x for
bentonite. The slurry was then deaired under vac-
3.1 Consolidation rate
uum and agitated to remove air bubbles. From this
point, two separate methods were used to prepare sam- The consolidation rate of each loading step was calcu-
ples to the desired initial overconsolidation ratio. The lated using the log time method. These results for each
kaolinite slurry was placed in a mold and under a temperature are presented in Figures 1 and 2. A trend
pressure of 200 kPa. The sample was then trimmed line was used to determine the average percentage
to approximately 23 mm and placed in the Rowe cell. change per one degree temperature change and is pre-
The bentonite slurry was placed directly into the Rowe sented in Table 1.All load histories produce an increase
cell. The cell was placed in the water bath set to 20◦ C. in the consolidation rate with the notable exception
Placing the bentonite directly into the Rowe cell of the largest OCR for bentonite which exhibited the
without prior consolidation was desirable for two rea- opposite response.
sons; it allowed for easier sample preparation (the low The normally consolidated (NC) stages show a sim-
permeability made preconsolidation using a mold dif- ilar response for both materials with the consolidation
ficult and time consuming), the resulting final height rate increasing 1.1–1.9%/◦ C. When overconsolidated,
(≈6–7 mm) rendered manageable test durations. Orig- the relation experiences greater variability, with an
inally, kaolinite was prepared in the same manner, opposite behaviour for bentonite with an OCR = 4.
however at high overconsolidation ratios a longer con- This reversal in trend is not consistent with the kaoli-
solidation time and magnitude was desired which was nite along with other findings that the settlement rate
the impetus for the change. increases with temperature, although direct compar-
Once in the Rowe cell an effective pressure of isons were not established. Therefore, this may reflect
400 kPa was applied until primary consolidation was the true nature of highly overconsolidated bentonite
complete as observed by settlement. The effective or could be a function of other factors. Care was
pressure was then reduced to 50 kPa and allowed to taken to reduce the impact of factors such as time and
272
is the opposite to that of the other overconsolidated
stages.
A lesser understood factor is the molecular inter-
action between clay platelets and water molecules.
This likely will vary based on the atomic structure
but the direction of change should remain consistent
throughout the tested temperature range within the
context of each material. The physical expansion of
the constituent soil parts may also contribute.
These factors can act in a competing fashion with
the overall behaviour dictated by the particulars of each
case; material type, stress history, and temperature.
Similarly, volume change with temperature is known
to have a relation to stress history, which impacts
Figure 1. Kaolonite consolidation rate with temperature. the magnitude and direction of change (expansion or
contraction).
Changes in the compression index (Cc ) and recom-
pression index (Cr ) were also investigated. No sig-
nificant changes in Cc or Cr were noted for either
material.
3.2 Volume change

Separate specimens were tested to assess the volume
change characteristics of both kaolinite and bentonite
at multiple overconsolidation ratios; 8, 4, 2, and 1.
Samples were prepared at a constant temperature of
20◦ C and pressures applied to achieve the desired
overconsolidation ratio. Once primary settlement was
Figure 2. Bentonite consolidation rate with temperature. complete, the temperature was increased to 50◦ C then
lowered back to 20◦ C before the next pressure step.
Table 1. Consolidation average rate change in percent per A majority of the volume change occurred during the
degree increase. heating and cooling process. The temperature was left
at the end of the range until the strain stabilized.
Loading Kaolinite Bentonite During heating kaolinite experienced a reduction in
volume for both normally consolidated stages (NC800
OCR = 4 1.7 –2.7 and OCR = 1) and the previous two stages (OCR = 2
OCR = 2 4.2 2.5 and 4). The initial stage (OCR = 8) experienced min-
OCR = 1 2.4 1.9 imal volume change. During cooling, the decrease
NC 800 1.1 1.1 in volume was plastic for all stages except OCR = 8
NC 1200 1.1 1.9
where it experienced swelling during cooling.
Bentonite stages experienced similar behaviour
during heating with little change for OCR = 1 and 4.
Consolidation of 0.2% and 0.3% for OCR = 2 and
loading rates but variations in sample heights can com- NC800 respectively. The OCR = 8 stage experienced
plicate duplicating exact initial starting conditions. swelling during heating to 0.7% strain and further
Additional tests are required to validate this response. swelling during cooling for a total strain of 1.2%.
Also, new preparation methods and test parameters Under cooling the remaining stages experienced a
will be considered to reduce these concerns. contraction of 0.6–0.9%.
One clear contributor to the consolidation rate is As observed in Figure 3 and Figure 4, both mate-
the viscosity of the pore medium, in this case water. rials had a trend related to the pressure history. The
Values are readily available in engineering tables, with higher the overconsolidation ratio the lower the ver-
the viscosity increasing with temperature. Therefore, tical contraction, transitioning into a dilation at the
if this were the only factor, the change in rate of each highest OCR. Similar behaviour for saturated clays
pressure step should be the same for any material at has been noted in multiple sources (Abuel-Naga et al.,
the same void ratio. Each bentonite stage and kaolinite 2007, Sultan et al., 2002, Burghignoli et al., 2000,
overconsolidated stages had similar void ratios as the Cekerevac and Laloui, 2004).
settlement caused from each pressure step was small. A noticeable difference between both materials was
However, the increase in consolidation rate is differ- observed during cooling. Kaolinite stages which con-
ent amongst OCR = 1, 2, and 4 for both materials. tracted exhibited little response during the cooling
Furthermore, the behaviour of bentonite at OCR = 4 cycle, while the corresponding bentonite stages further
273
Figure 3. Kaolinite thermal induced vertical strain.
Figure 4. Bentonite thermal induced vertical strain.
contracted. There is limited data available for satu-

rated bentonite, however, those available are mainly
conducted on unsaturated bentonite (Tang et al., 2008, exception requires further investigation. The magni-
Ye et al., 2013). tude of rate change was similar for normally con-
Volume changes can take place in two fashions, solidated stages but varies when overconsolidated.
reversible (elastic) and non-reversible (plastic). The Given the opposite response at high OCRs, multiple
elastic components comprise of the physical swelling mechanisms must be responsible for the rate change
of the constituent clay parts (soil particles and water). (viscosity of pore fluid and intermolecular bonds are
Both the soil particles and water will expand during suggested).
heating and contract during cooling. Given the drained No significant change in the compression index
test conditions the water only contributes to an overall (Cc ) or recompression index (Cr ) were noted for either
increase in volume if it is locked within the structure material.
(diffuse double layer – bonded to the clay platelets) The stress history, namely overconsolidation ratio is
and not expelled to maintain constant void space. This responsible for either a thermal contraction or dilation.
may be the case at high OCR resulting in the dilation Normally consolidated stages experienced consolida-
during heating. At lower OCR the expansion of the tion, while highly overconsolidated stages (OCR = 8)
adsorbed water along with the soil particles may be experienced dilation.
offset by the expulsion of water not so tightly adhered
to the soil structure.
The plastic component requires a reorganization of REFERENCES
the soil fabric. This is only possible where the pore
water is expelled. Therefore, the forces bonding the Abuel-Naga, H. M., Bergado, D. T. & Bouazza, A. (2007)
water to the platelets needs to be less than the change Thermally induced volume change and excess pore water
in viscosity of the water to allow it to escape. This pressure of soft Bangkok clay. Engineering Geology, 89,
144–154.
is the case under normally consolidated conditions as
Abuel-Naga, H. M., Bergado, D. T., Rujivipat, S. & Soralump,
the pore water is already at the tipping point of reten- P. (2005) Thermal consolidation of soft Bangkok clay.
tion within the soil structure and expulsion. As the Lowland Technology International, 7, 13–21.
OCR increases this becomes more difficult. Another Abuel-Naga, H. M. & Bouazza,A. (2013) Thermomechanical
consideration is an increase/decrease in the ability of Behavior of Saturated Geosynthetic Clay Liners. Journal
the soil structure to retain water (adsorbed water – of Geotechnical and Geoenvironmental Engineering, 139,
diffuse double layer) which may change with temper- 539–547.
ature. Although, it has been suggested that the effect Burghignoli, A., Desideri, A. & Miliziano, S. (2000) A labo-
of temperature on the thickness of the diffuse double ratory study on the thermomechanical behaviour of clayey
soils. Canadian Geotechnical Journal, 37, 764–780.
layer is insignificant (Abuel-Naga et al., 2005).
Cekerevac, C. & Laloui, L. (2004) Experimental study
of thermal effects on the mechanical behaviour of a
clay. International Journal for Numerical and Analytical
4 CONCLUSIONS Methods in Geomechanics, 28, 209–228.
Cui, Y.-J. & Tang, A. M. (2013) On the chemo-thermo-
Thermal consolidation and volumetric tests were per- hydro-mechanical behaviour of geological and engineered
formed on saturated kaolinite and bentonite clays at barriers. Journal of Rock Mechanics and Geotechnical
various overconsolidation ratios. Engineering, 5, 169–178.
Delage, P., Sultan, N. & Cui, Y. J. (2000) On the thermal con-
Elevated temperature has a direct influence on the solidation of Boom clay. Canadian Geotechnical Journal,
consolidation rate of bentonite and kaolinite. Increased 37, 343–354.
temperature increases the consolidation rate with the Di Donna, A. & Laloui, L. (2015) Response of soil subjected
notable exception of highly overconsolidated ben- to thermal cyclic loading: Experimental and constitutive
tonite, which experienced the opposite response. This study. Engineering Geology, 190, 65–76.
274
Knellwolf, C., Peron, H. & Laloui, L. (2011) Geotechnical Strawn, D. (2015) Soil chemistry, Chichester, West Sussex,
Analysis of Heat Exchanger Piles. J. Geotech. Geoenviron. UK, Hoboken, NJ, USA : Wiley Blackwell, 2015.
Eng., 137, 890–902. Sultan, N., Delage, P. & Cui, Y. J. (2002) Temperature
Laloui, L. & Di Donna, A. (2011) Understanding the effects on the volume change behaviour of Boom clay.
behaviour of energy geo-structures. Proceedings of ICE, Engineering Geology, 64, 135–145.
164, 184–191. Tang, A. M., Cui, Y. J. & Barnel, N. (2008) Thermo-
Ojovan, M. I. (2010) An Introduction to Nuclear Waste mechanical behaviour of a compacted swelling clay.
Immobilisation, Burlington, Burlington: Elsevier Science. Geotechnique, 58, 45–54.
Pothiraksanon, C., Bergado, D. T. & Abuel-Naga, H. M. Ye, W., Zhang, Y., Chen, Y., Chen, B. & Cui, Y. (2013) Exper-
(2010a) Full-scale embankment consolidation test using imental investigation on the thermal volumetric behavior
prefabricated vertical thermal drains. Soils and Founda- of highly compacted GMZ01 Bent. Appl. Clay Sci., 83-84,
tions, 50, 599–608. 210–216.
Pothiraksanon, C., Saowapakpiboon, J., Bergado, D. T., Voot-
tipruex, P. & Abuel-Naga, H. M. (2010b) Soft ground
improvement with solar-powered drainage. Proceedings
of the Institution of Civil Engineers: Ground Improvement,
163, 23–30.
275
Analysis of cement-based thermal energy storages

considering natural convection
H. Hailemariam, D. Shrestha & F. Wuttke

Marine and Land Geomechanics and Geotechnics, Kiel University, Kiel, Germany
ABSTRACT: One of the common technologies for balancing the energy demand and supply in district heating,
domestic hot water production, thermal power plants and thermal process industries in general is thermal energy
storage. In this study, the coupled thermo-hydro behaviour of a fully saturated cement-based thermal energy
storage system for domestic applications modeled with the Finite Element method by taking in to account the
effect of buoyancy-driven convection on the temperature and heat distribution within the storage system is
presented. Thermal energy storage systems in water saturated condition are commonly analysed considering
pure solid-solid, solid-water and water-water conduction. The significant heat transfer contribution from the
buoyancy-driven convection induced due to flow in saturated media arising naturally from the effect of a density
difference, resulting from a change in temperature or concentration of a medium, is commonly neglected. This
consequently leads to the underestimation of the actual loading/unloading rates as well as the overdesigning of
such systems. The results of an extensive experimental program for the determination of parameters used for the
FE modeling are also presented.
1 INTRODUCTION system, IGLU project1 , which aims at developing a

solar thermal powered heat energy storage system for
Thermal energy storage (Braun et al. 1981, Hesaraki domestic dwellings, by considering natural convec-
et al. 2015), in particular sensible heat storage (Dincer tion is presented. The finite element software package
et al. 1997) as compared to latent heat storage and COMSOL Multiphysics2 was used to couple the gov-
thermo-chemical storage, has recently gained much erning thermo-hydro equations discussed in Section 2
interest in the renewable energy storage sector due to of this paper.
its comparatively low cost and technical development. The loading/unloading of the system is operated
Sensible heat storages work on the principle of storing via an embedded heat exchanger system with a car-
thermal energy by raising or lowering the tempera- rier fluid (in this case water). The system provides
ture of liquid (commonly water) or solid media, and efficient flexibility in terms of economy, space require-
do not involve material phase change or conversion ments and operation as compared to other existing heat
of thermal energy by chemical reactions or adsorp- storage systems (Fig. 1). Two porous cement-based
tion processes as in latent heat and thermo-chemical commercial thermal energy storage materials namely,
storages respectively. Füllbinders L and M (SCHWENK Zement KG3 ) were
Environmentally friendly renewable energy sources used for our investigations.
such as solar, geothermal and wind energy are typically The experimental scheme for the determination of
harnessed for the purpose of sensible heat storage. parameters included: thermal conductivity and spe-
Solid sensible heat storage systems (Laing et al. 2006, cific heat using a transient line-source measurement
Laing et al. 2012) in the form of storage media such technique; semi-quantitative phase content determi-
as concrete, rocks, geomaterials and cemented satu- nation by combined X-ray diffraction (XRD) and
rated porous media are nowadays typically used to X-ray fluorescence (XRF) analysis; mechanical
capture solar thermal power. Such systems when used
in domestic thermal energy storage applications are 1
IGLU is an analysis, modeling and assessment of an intel-
preferable when compared to sensible heat storage via ligent and environmentally friendly geothermal long-term
liquids such as water as they provide low investment heat storage system project funded by the German Federal
costs with high operability (Laing et al. 2006) and can Ministry of Economy and Energy (BMWi)
be designed to bear loads as part of the sub-structure 2
COMSOL Multiphysics is a finite element analysis, solver
of buildings. and simulation software package for various coupled phe-
In this research, the numerical analysis of a satu- nomena in physics and engineering applications
rated cement-based porous media sensible heat storage 3
http://www.schwenk-zement.de/
277
combining heat transfer equations in solid and fluid
phases (Bejan 2004, Bejan & Lorente 2004).
where, the subscripts m, s and f refer to the porous

medium, solid and ?uid phases, respectively, n is poros-
ity, ρ is density, C is the specific heat of the solid, Cp
is the specific heat at constant pressure of the ?uid, λ
is the thermal conductivity, Q is the heat source, T is
temperature, t is time and ν is the flow velocity field.
Figure 1. Schematic representation of the sensible heat The buoyancy-driven convective flow velocity ν
storage system of IGLU project. in Equation 1 is obtained by solving porous media
fluid flow equations. Three fluid flow in porous media
equations namely, the Richards equation (for flow in
saturated and un-saturated porous media), the Darcy’s
law (preferable for flow in porous media that is com-
strength parameters such as uni-axial and shear
paratively slow) and the Brinkman equations (for fast
strength using uni-axial, one dimensional confined
flow in porous media) are commonly used to analyse
compression and tri-axial tests; and hydraulic prop-
fluid velocity and pressure in porous media. In this
erties such as hydraulic permeability or conductivity
study the Richards equation for a single phase flow
under controlled stress and temperature conditions
(water) has been used to obtain the convective flow
using a specialised hydraulic permeability meter.
velocity ν and pressure p.
According to Darcy’s law, the net flux across a face
of porous media is given by:
2 FE MODELING
2.1 Heat transfer governing equations

Conductive and convective modes of heat transfer gen-
where, ν is the Darcy velocity field, K is the saturated
erally govern the amount of heat energy transported
hydraulic conductivity of the porous medium, ρf is the
through a sensible porous media heat storage system.
fluid density, g is the gravitational acceleration, p is the
Heat conduction occurs in the porous cement based
pressure and ∇D is a unit vector in the direction of g.
storage material and the entrapped pore-water, Alu-
The pressure p is solved by inserting the Darcy’s law,
PE pipe wall and the heat carried fluid (water) inside
Equation 4, in to the equation of continuity, Equation 5,
the Alu-PE pipe. Heat convection on the other hand
and the Darcy velocity ν is obtained from Equation 4.
occurs in the fluid entrapped in the saturated heat stor-
age material and the heat carrier fluid in the Alu-PE
pipes.
The convective mode of heat transfer involves fluid
flow accompanied with conduction, or diffusion, and
is basically divided in to two processes. When the
motion of the fluid arises from an external cause the where, n is the saturated liquid fraction or porosity of
process is termed as forced convection. Whereas, if medium and Qs is the strength of fluid source. For
no such externally induced flow exists and the flow an incompressible fluid such as water the term ρf
arises naturally from the effect of a density difference, moves outside the divergence operator, and the con-
resulting from a concentration or a temperature dif- tinuity equation is expressed in terms of the storage
ference in body forces such as gravity, the process is coefficient S as:
known as natural convection. In this study, only natural
convection is considered.
Under isotropic medium conditions, where local
thermal equilibrium is ensured (Ts = Tf = T , where
Ts and Tf are the temperatures of the solid and
fluid phases respectively), heat transfer Equations 1–3 Richards equation is commonly used for modeling
for the porous storage medium can be obtained by variably saturated porous media. The model solves
278
for the pressure p and the fluid velocity ν using the Table 1. Material properties of the filling material and the
following equations: heat exchanger.
Heat exchanger Porous material
Materials Alu-PE composite* Fü. L ** Fü. M**
Density (kg m−3 ) 1825 1583 1609

Porosity (–) – 0.543 0.518
Thermal conduc. 0.400 0.960 0.965
where, C is the specific capacity describing the (W m−1 K−1 )
changes in liquid volume fraction of porous medium Specific heat 1261 2083.4 1957.1
for a variably saturated condition, Se is the effective (J kg−1 K−1 )
saturation and Kr is the relative permeability. The ana- Hydraulic conduc. – 11.95 7.13
lytical formulas of van Genuchten (1980) are typically (x10−8 m s−1 )
used to model the retention characteristics of variably
saturated porous media. * Carrier fluid density = 1000 kg m−3 , thermal conduc-
tivity = 0.58 W m−1 K−1 , specific heat = 4190 J kg−1 K−1 .
Darcy’s linear relationship between flow velocity ** Fully saturated condition.
and pressure gradient is not applicable for porous
media flow with high Reynolds number and high
particle size. In such instances, the Brinkman model Table 2. Chemical and mineralogical characteristics of the
(which is an extension of Darcy’s model) can be used. thermal storage materials (obtained by X-ray fluorescence
Brinkman modified the Darcy’s law by adding a vis- and X-ray diffraction analysis).
cous term to account for the presence of a solid bound-
ary. The unsteady Brinkman equation for viscous flow Porous material
in a porous medium is given as:
Materials Fü. L Fü. M
Lime + Limestone (wt.%) 75 65

Portl. Cement 42.5 (wt.%) 25 35
SiO2 (%) 17.28 16.43
Al2 O3 (%) 4.90 5.17
TiO2 (%) 0.27 0.29
MgO (%) 0.92 0.51
Fe2 O3 (%) 2.76 2.79
CaO (%) 48.75 51.62
P2 O5 (%) 0.12 0.16
where, µ is the fluid’s dynamic viscosity, K is the sat- Na2 O (%) 0.14 0.12
urated hydraulic conductivity of the porous medium, K2 O (%) 0.98 0.78
µdv is the dilatational viscosity, F is a force term used MnO (%) 0.04 0.04
to represent directed forces such as compressibility
effects and gravity and Qbr is an optional mass source
term used to model condensation in porous media.
3 EXPERIMENTAL PROGRAM & RESULTS
2.2 Boundary & initial conditions
3.1 Tested materials
The following boundary and initial conditions are
assumed to solve the governing equations: In Tables 1 & 2 material properties and physiochemical
analysis results of the two investigated cement-based
– An initial temperature of 20◦ C representing ground thermal energy storage materials, Füllbinders L and
temperature is assigned to the porous medium, Alu- M, is presented. The samples were prepared with a
PE and carrier fluid giving the following condition: water to solids ratio of 0.8 and were then stored under
T = To = 20◦ C. water for 28 days. Storage in water ensures full satura-
– A constant inlet temperature of 90◦ C is assigned to tion and prevents possible cracking of samples which
the inlet of the Alu-PE pipe representing the temper- may happen soon after sample preparation due to the
ature of the hot carrier fluid (water) during loading hydration of cement.
of the system.
– A thermal insulation specifying a zero flux is
applied to the outside boundary of the heat storage
system as: n · (λm ∇T ) = 0, where n is the vector 3.2 Thermal properties
normal to the boundary. The thermal conductivity and specific heat of the
– Fluid flow across impervious boundaries is pre- samples were measured with a Decagon KD2 Pro ther-
vented with the following condition: n · ρf v = 0, mal needle probe transient line source measurement
where n is the vector normal to the boundary. technique in accordance to ASTM D5334-08 (ASTM
279
Figure 2. Thin sections of Füllbinder M (top-left), Füll-
binder L (top-right) and 3-D view of Fü. L specimen
(bottom).
2008) and IEEE 442 standards (IEEE 1992). Ther-

mal needle probes SH-1, dual needle (with a length
of 30 mm, dia. 1.3 mm and 6 mm spacing between
needles) and TR-1, single needle (with a length of
100 mm and dia. 2.4 mm) were used to measure the
thermal conductivity and specific heat of the samples
at room temperature and atmospheric pressure, respec-
tively. The sufficient length to diameter ratio of the Figure 3. One dimensional confined compression test
probes ensure that conditions for an infinitely long and results at different curing times of Fü. L (top) and Fü. M
infinitely thin heating source are met. The recorded (bottom).
error for all measurements was kept well within the
0.015% limit. The KD2 Pro contains a linear heat
source and a temperature measuring element with a
resolution of 0.001◦ C, and calculates the thermal con-
ductivity and thermal diffusivity of the medium using
the following relations:
where, λ (W m−1 K−1 ) is the thermal conductivity of

the sample, Q (W m−1 ) is a constant rate of applica-
tion of heat, T (0 K) is the temperature response with Figure 4. Unconfined compression (uni-axial) test results.
time of the source, Ei is the exponent integral, r (m)
is the distance between the heater and the tempera-
ture sensor, t (s) is the amount of time that has passed along the solid skeleton providing a uniform thermal
since the start of heating and D (m2 s−1 ) is the thermal conductivity (mainly through the solid-solid cementi-
diffusivity of the sample. tious contact points within the skeleton) and a uniform
heat storage capability (mainly through the even dis-
tribution of entrapped pore water) across the storage
3.3 Thin section & hydro-mechanical properties
materials.
Figure 2 shows thin section images of Füllbinders L Results of one-dimensional confined compression
and M taken with a microscope to study the porous tests, unconfined compression or uni-axial tests and
material structure. For both materials, it can be noted drained tri-axial tests are presented in Figures 3–5
that the pore structure is homogenously distributed respectively.
280
Figure 6. Results of hydraulic conductivity tests at different
hydraulic gradients with a confining stress of 150 kPa.
Both Füllbinders L and M show significant increase

in mechanical strength with curing time, with Füll-
binder M exhibiting a comparatively higher stiffness
due to its higher cement binder content. The difference
in the cement binder content in the two energy storage
materials also plays a significant role in their thermal
behaviour (Table 1). The higher cement binder con-
tent of Füllbinder M when compared with Füllbinder
L, produces a medium with a relatively lower poros-
ity and hence a lower amount of entrapped pore water
when fully saturated. Consequently Füllbinder M has
a higher thermal conductivity but a lower heat storage
capacity when compared to Füllbinder L (Table 1).
Figure 5 (bottom) shows failure mechanism of Füll-
binder L samples at different confining stress σ3 levels
in a drained tri-axial test. At lower confining stress
levels the sample tends to fail by cracking diagonally.
However, at higher confining stress levels the sample
rather fails by bulging with minor or no visible crack-
ing. This provides a valuable insight on the design load
considerations for such storage systems when they
are used as part of the sub-structure of buildings, as
formation of major cracks along the storage material
significantly lowers the effective thermal conductivity
and hence seriously undermining the operation of the
system.
In Figure 6 results of hydraulic conductivity tests
performed with a specialised hydraulic conductivity
meter under controlled hydraulic gradient i and con-
fining stress σ conditions are presented. As expected
the measured hydraulic conductivity of both materials
decreases with an increase in the applied confining
stress due to a reduction in the effective poros-
ity upon application of stress. Moreover, Füllbinder
M with its comparatively lower porosity exhibits a
lower hydraulic conductivity as compared to that of
Füllbinder L.
Figure 5. Drained tri-axial test results on Fü. L speci- A good energy storage material should possess a
men at different confining stress σ3 levels with peak shear moderate to good thermal conductivity, a moderate
strength parameters of c = 196.77 kPa and ϕ = 19.900 (speci- strength to provide load bearing capability but not too
men size: dia. = 10 cm and height = 20 cm) [p = (σ1 + 2σ3 )/3 high so as to enable easy removal of energy storage
and q = σ1 − σ3 ]. material for future maintenance and repair operations
as well as a homogeneous and highly porous structure
281
Figure 7. Sensible heat storage system configurations with
a helical (left) and vertical (right) Alu-PE fluid carrier pipes.
to provide a high heat storage capacity when satu-

rated. Considering the experimental analysis results
presented in Section 3, Füllbinder L is better qualified
to be used as an energy storage material when com-
pared to Füllbinder M and hence it has been selected
for the majority of the numerical analysis presented in
this study.
4 NUMERICAL ANALYSIS RESULTS &

DISCUSSION
Two typical sensible heat storage system configura-

tions are shown in Figure 7. In this research, the con-
figuration with vertically oriented loading/unloading
Alu-PE pipes (Fig. 7 right) is analysed. The system
consists of an insulated cylindrical heat storage tank
with a dia. of 1.1 m and a height of 1.04 m. The heat-
ing/cooling operation of the system is facilitated via
embedded Alu-PE pipes with an inside dia. of 2 cm
and a wall thickness of 2.5 mm. Water with up to a
maximum temperature of 90◦ C is used as a carrier
fluid.
In Figure 8 the temporal temperature variation at the Figure 8. Temporal variation of temperature for: Fü. L &
center of the sensible heat storage system during load- M with Re = 458.2 (26 lph) (top), Fü. L with Re = 458.2 (26
ing with a hot carrier fluid (water) at 900 C for different lph) at different hydraulic conductivities (middle) and Fü.
heat storage material types and operation conditions is L with K = 1 × 10−3 ms−1 at different carrier fluid veloci-
presented. The variation in temperature with progress ties (bottom) [Re = Reynolds number; lph = liters per hour;
in time is analysed for: the two material types investi- K = hydraulic conductivity].
gated in this study (Füllbinders L and M) (Fig. 8 top),
the effect of convection considerations with hydraulic
conductivities K = 0 – 1 × 10−2 ms−1 (Fig. 8 middle)
and the effect of carrier fluid velocity or Reynolds
number Re between 119.5–916.3 (inlet fluid velocities affected by changes in the hydraulic conductivity of
between 6.8–52.0 liters per hour respectively) (Fig. 8 the porous media due to convection effects.
bottom). The system temperature variations for the porous
The temporal variation in temperature for both Füll- medium with no convection (pure conduction only)
binders L and M are closely matched due to similarities consideration and with convection K = 0 ms−1 consid-
in their thermo-hydro properties, however, the varia- eration are almost the same. However, with an increase
tion in the temperature changes of the system is highly in the hydraulic conductivity of the porous medium,
282
Figure 9. (from top to bottom) FE-mesh and system outline, spatial temperature distribution after 5 hours, isothermal contours
after 60 hours and spatial temperature distribution after 120 hours of loading via a carrier fluid at 90◦ C of the sensible heat
storage system with no convection or pure conduction only (left) and with convection K = 1 × 10−2 ms−1 consideration (right)
[heat storage material is Fü. L with Re = 458.2 (26 lph)].
the influence of convection in the temperature vari- medium with lower hydraulic conductivity, the effect
ation of the system increases significantly. Typically, of convection on the temperature variation can be
the system with no convection considerations reaches neglected by considering only pure conduction to sim-
near storage capacity (above 80◦ C) after around plify system analysis. Whereas, for a porous medium
100 hours of loading time, whereas, the system with with a higher hydraulic conductivity, consideration
K = 1 × 10−2 ms−1 reaches storage capacity within of convection in the analysis of temperature changes
24 hours of loading duration. Hence, for a porous upon loading/unloading enables an efficient system
283
design with accurate estimation of loading/unloading compared to the outlet while loading and vice versa.
rates, thus saving time and avoiding overdesign of the In such instances, consideration of natural convection
system. provides accurate estimation of the loading/unloading
Natural convection, in addition to its effects on the rates and avoids the overdesign of the system.
temporal variation of temperature of the heat storage
system, also plays a vital role in the spatial variation
(temperature distribution) in the sensible heat stor- ACKNOWLEDGEMENTS
age system (Fig. 9). Assumption of a pure conduction
yields in an even and a conservative distribution of The authors gratefully acknowledge the funding pro-
temperature as shown in Figure 9 (left). This repre- vided by the German Federal Ministry for Economic
sentation is far from reality, where the actual spatial Affairs and Energy (BMWi) under Grant numbers
temperature distribution of heat storage systems with 0325547B and KF3067302HF3, as well as the support
higher hydraulic conductivities when natural convec- of Project Management Jülich (PTJ).
tion is considered is as shown in Fig. 9 (right), with
a higher temperature field across the inlet as com-
pared to areas near the outlet of the carrier fluid during REFERENCES
loading and vice versa.
For the system considered in this research (Fig. 7 Braun, J. E., Klein, S.A., Mitchell, J.W. 1981. Seasonal
right) the effect of carrier fluid velocity or Reynolds storage of energy in solar heating. Solar Energy 26(5):
number Re on the temporal and spatial variation of 403–411.
temperature across the sensible heat storage system Hesaraki, A., Holmberg, S., Haghighat, F. 2015. Seasonal
is minor (Fig 8 bottom). However, for sensible heat thermal energy storage with heat pumps and low temper-
storage systems with complex shapes of embedded atures in building projects—a comparative review. Renew-
able and Sustainable Energy Reviews 43: 1199–1213.
fluid carrier pipes consisting of several twists or turns,
Dincer, I., Dost, S., Li, X. 1997. Performance analyses of sen-
changes in the carrier fluid velocity can significantly sible heat storage systems for thermal applications. Inter-
affect their operation. national Journal of Energy Research 21(12): 1157–1171.
Laing, D., Steinmann, W-D., Tamme, R., Richter, C. 2006.
Solid media thermal storage for parabolic trough power
5 CONCLUSIONS plants. Solar Energy 80(10): 1283–1289.
Laing, D., Bahl, C., Bauer, T., Fiss, M., Breidenbach, N.,
The coupled thermo-hydro behaviour of a saturated Hempel, M. 2012. High-temperature solid-media thermal
sensible heat storage system for domestic heat storage energy storage for solar thermal power plants. Proceed-
ings of the IEEE 100(2): 516–524.
applications was analysed numerically by considering
Bejan, A. 2004. Convection Heat Transfer (3rd ed.),
the effect of buoyancy-driven convective heat flow. New York: Wiley.
Results of extensive experimental analysis for the Bejan, A., Lorente, S. 2004. The constructal law and the ther-
determination of parameters for numerical analysis modynamics of flow systems with configuration, Inter-
are also presented. For sensible porous media heat national Journal of Heat and Mass Transfer 47(14–16):
storage systems with very low hydraulic conductivi- 3203–3214.
ties, the effect of natural convection on the temporal van Genuchten, M.Th. 1980. A closed-form equation for pre-
and spatial variations of temperature during load- dicting the hydraulic conductivity of unsaturated soils.
ing/unloading operations of the system is minimal and Soil Science Society of America Journal 44(5): 892–898.
ASTM. 2008. ASTM 5334-08: Standard test method for
can be neglected for simplicity of analysis. However,
determination of thermal conductivity of soil and soft rock
for systems with considerable hydraulic conductivity, by thermal needle probe procedure.
convective heat transfer plays a significant role with IEEE. 1992. Guide for soil thermal resistivity measure-
a resultant faster loading/unloading rates and a higher ments, Inc. New York, Inst. of Electrical and Electronics
temperature field near the carrier fluid pipe inlet as Engineers.
284
Experimental and constitutive study of the thermo-mechanical

behavior of an oil sand
M. Mohamadi, E. Piotrowska & R.G. Wan

Department of Civil Engineering, University of Calgary, Canada
ABSTRACT: The understanding of the thermo-mechanical behavior of oil sand deposits is of prime importance
in the safe and economic design of thermal oil recovery processes in Alberta, Canada. The present study focuses
on both the experimental and constitutive characterizations of the thermo-mechanical behavior of “Firebag”
oil sand from the Athabasca area in Alberta, Canada. Isothermal triaxial compression tests at different levels
of confining pressure and temperature were performed on samples obtained from a depth of approximately
270 m. Experimental results at ambient temperature show a quite brittle behavior with relatively large peak and
residual strengths due to the initially dense and interlocked structure of the granular material. High temperatures
were found to strongly increase the ductility and decrease the stiffness and strength of the samples. Based on
the experimental results, a simple non-associated plasticity model is proposed to realistically describe the non-
isothermal mechanical behavior of Firebag oil sands. The proposed model is built on the Mohr-Coulomb plastic
limit enriched with constitutive features that enable it to capture important behavioral aspects of oil sands such
as high peak strength and dilatancy. The adequacy of the model is verified through the simulation of laboratory
triaxial test results at different temperatures. The model provides a consistent and yet simple framework for easy
implementation into a coupled reservoir-geomechanics computer algorithm for thermal oil recovery processes.
1 INTRODUCTION elevated temperatures was reported in the same study.

Kosar (1989) reported that the initial stiffness, peak
The thermo-mechanical behavior of oil sands is a topic strength and dilatancy of high fines content (51% of
of particular interest in western Canada due to the role particles smaller than 0.074 mm) McMurray Forma-
it plays in the safety and economical aspects of steam tion oil sands increase when the testing temperature
injection oil recovery projects, among others. Gen- increases from 20 to 225◦ C. Based on multistage heat-
erally speaking, the injection of high-pressure steam ing from 21 to 102 and 200◦ C followed by drained
creates fractures that extend following both tensile and triaxial shearing at the end of each stage, Wong et al.
shearing modes in an oil sand reservoir.Although these (1993) reported that the peak strength of Cold Lake oil
failure processes are seen to be favorable for increasing sands remains unchanged up to 102◦ C and increases
the reservoir permeability and hence, the efficiency pronouncedly at 200◦ C. In the same study, no major
of oil production, fractures have to be avoided near change in the volumetric strain response was observed
the shale cap-rock which has to remain mechanically at elevated temperatures.
intact throughout the entire oil extraction process. Fur- It is obvious that no consistent temperature-induced
thermore, coupled thermo-mechanical deformations phenomenological mechanism can be identified from
induced by the propagating steam chamber within the the results of the aforementioned studies. This, among
reservoir produce surface heave that has to remain others, has resulted into approximating the non-
within the tolerable limits. As such, the need for the isothermal behavior of oil sands with isothermal elas-
accurate description of the thermo-mechanical behav- tic or elastoplastic models. For instance, Agar et al.
ior of oil sands for evaluating in situ stress changes (1987) adopted the very basic nonlinear elastic hyper-
and deformations is indispensable. bolic model developed by Duncan and Chang (1970) to
Reports of experimental results on the thermo- describe the pre-peak behavior of Athabasca oil sand.
mechanical behavior of oil sands are quite scarce in Vaziri (1989) combined the same nonlinear elasticity
the literature. Performing a series of isothermal triax- model as used in Agar et al. (1987), with critical state
ial compression tests at temperatures between 20 and soil mechanics concepts (Schofield and Wroth 1968)
200◦ C, Agar et al. (1987) reported that the strength of to develop a conceptual framework for describing the
Athabasca oil sands remains unchanged when the tem- behavior of oil sands. However, no comparison with
perature is increased from 20 to 125◦ C and decreases experimental or field data on oil sand behavior was
upon further heating to 200◦ C. A small but measurable made. Using Rowe’s stress-dilatancy theory along with
increase in the stiffness and dilatancy of the material at the Mohr-Coulomb-type yield function, Wan et al.
285
(1991) proposed an elastoplastic model to describe
the behavior of Cold Lake oil sand at two different
temperatures. Subsequently, Samieh and Wong (1998)
developed a constitutive model to capture the behavior
of Athabasca oil sand at low confining pressures (50–
750 kPa). The central advantage of their model is the
description of post-peak softening and shear dilation
using the disturbed state concept, originally proposed
by Desai et al. (1986). Li and Chalaturnyk (2005).
used the strain-softening Mohr-Coulomb model that
is available in the FLAC geomechanical simulator
(Itasca Consulting Group 2011) to reproduce results
of laboratory experiments on oil sands at ambient
temperature.
All of the above-mentioned constitutive models
are based on isothermal elasticity/elastoplasticiy, and
therefore are not suitable for describing the complex Figure 1. Grain size distribution of the tested oil sands.
behavior of oil sands under non-isothermal conditions.
It is worth mentioning that Wan et al. (1991) used varia- Table 1. Mass percentage of bitumen and water in three oil
tions of friction angle with temperature to incorporate sand samples (total mass of each sample is 130 gr).
temperature-dependence in their model. However, a
Sample No. Bitumen content (%) Water content (%)
consistent implementation of the temperature effects
to describe diverse behavioral aspects of oil sands such 1 14.4 3.0
as yielding, hardening and volume contraction was not 2 4.9 9.4
performed. 3 12.5 4.3
Given the lack of consensus among the exper-
imental results and the oversimplified modeling
approaches, as reviewed in the aforementioned previ-
ous studies, the focus of this paper is twofold. Firstly, the results of grain size distribution and Dean-Strak
the thermo-mechanical behavior of Firebag oil sand, tests, respectively. A typical CT-image of one of the
obtained from Athabasca area in Alberta, Canada, samples at its mid-height prior to testing is shown in
is investigated experimentally in a series of high- Figure 2, wherein dark spots correspond to regions of
pressure and high-temperature triaxial tests. Secondly, low density. The existence of localized micro-cracks
based on phenomenological mechanisms observed in and diffuse zones of low density in this image reveals
the above-mentioned lab experiments, a simple non- the initially disturbed structure of the sample, which
isothermal constitutive model is proposed to describe is typical of all tested samples as described later in the
the effects of temperature on the mechanical behavior paper.
of Firebag oil sand. The model formulation is generic All samples of the present study were tested in the
enough for describing the non-isothermal behavior of Rock Mechanics Laboratory at the University of Cal-
other oil sands or frictional granular material, and yet gary using a high-pressure/high-temperature triaxial
simple enough to be easily implemented in any com- apparatus. Technical specifications of the apparatus
mercial finite element packages. The latter computer are given in Mohamadi and Wan (2016), to which the
implementation is currently under progress and will interested reader is referred.
be described elsewhere.
2.2 Testing procedure

Consolidated drained triaxial compression tests were
2 EXPERIMENTAL STUDY
performed on oil sand samples (89 × 180 mm), at
both ambient (T0 = 25◦ C) and elevated temperature
2.1 Tested material
(T = 130◦ C). The testing procedure at ambient tem-
The oil sand samples tested in this study were obtained perature included an initial isotropic consolidation
from the Athabasca oil sands deposit at the Firebag stage followed by shearing along a conventional tri-
site, located 120 km northeast of Fort McMurray in axial compression stress path at a rate of 1 mm/hr. At
Alberta, Canada. The samples were chosen from the elevated temperatures, testing included one more stage
middle McMurray formation at a depth interval of in which the specimen was heated prior to the shear-
268.5–274.5 m where the in-situ porosity as per the ing stage, while keeping the applied isotropic stress
density logs, ranges from 30 to 34%. The initial con- constant. The heating stage was performed in several
ditions of the samples were characterized via X-ray sub-stages, each including an initial heating at a rate
computed tomography (CT) scanning and common of 0.5◦ C/min followed by a stabilization sub-stage.
geotechnical index tests related to grain size distri- Sample deformations were monitored during the lat-
bution and Dean-Stark. Figure 1 and Table 1 present ter sub-stage until a negligible change in the sample
286
Figure 2. CT-image of an oil sand specimen at its mid-height
showing sample disturbance.
dimensions was measured, at which point the testing

was pursued.
2.3 Test results and discussions

Figure 3 presents a typical set of responses for the
samples tested under a confining pressure of 1 MPa,
at both 25 and 130◦ C. The two samples (T12 S2 and
T12 S5) tested at 25◦ C show identical initial stiffnesses
and post-peak strengths, while their peak strengths are
different. The latter is believed to be mainly due to vari-
ability in physical characteristics of the field samples.
Furthermore, the peak strength of oil sands is known
to be quite sensitive to sample disturbance (Agar et al.
1987). Figure 3. Effects of temperature on shearing behavior of
As for volume changes of the above-mentioned two Firebag oil sands at the confining pressure of 1 MPa.
samples it is observed that their initial compactions
and subsequent constant-rate dilations in the pre-peak to be associated with the initial adjustment of the
regime are quite similar. Turning to sample T12 S5, it is sample typically observed in the testing of rocks. In
observed that the volume change is anomalously sup- fact, such concave-upward segments of the stress-
pressed at an axial strain of 0.64%. This is attributed strain curve are attributed to pore and fissure closure
to difficulties associated with the local measurements in the Rock Mechanics literature (Goodman 1989). A
of the deformations in the post-peak regime where more detailed experimental study regarding the source
the occurrence of non-homogeneous deformations is of these concave-upward features in the behavior of
inevitable. tested oil sands has been performed by the authors.
Comparing results at 25 and 130◦ C, it is observed The results, which are beyond the scope of the cur-
that heating causes a considerable decrease in the rent paper, confirm that crack closure is indeed a
initial stiffness and peak strength, and a moderate major contributing factor to the formation of these
decrease in the post-peak strength. Furthermore, sam- early-stage non-linearities.
ple T13 S2 reaches its peak strength at a larger axial In view of correcting for the crack closures, the
strain (1.4%) as compared to those tested at ambient initial section of the stress-strain and volume change
temperature, implying its enhanced ductility. Finally, curves of the tested oil sands have to be adjusted.
heating seems to increase the maximum contractancy, Plots of deviatoric stress versus radial strain (not
as suggested by the volume change curves in Figure 3. shown here) strongly suggest that the cracks are mostly
Another noteworthy feature in Figure 3 is the ini- horizontally oriented. As a result, the initial upward
tial upward concavity of the deviatoric stress-strain concavity is mainly due to axial deformations, and
curves. The value of the deviatoric stress at the end can be easily corrected following the conventional
point of these sections (see Figure 3(a)), are too large approach as per ASTM-D1883 (2007). The corrected
287
stress-strain and volume change curves following this decreases at elevated temperatures. The temperature-
approach are shown later in Figure 5. dependence of the initial stiffness entails pressure-
Triaxial test results at the confining pressures of dependence of the thermal properties that has to be
3 and 5 MPa are not presented herein, due to limita- accounted for using a coupled thermo-elastic rule
tions in the length of the paper. However, the general (Graham et al. 2001, Mohamadi and Wan 2015).
temperature-induced phenomenological mechanisms, Assuming isotropic thermal strains, the coupling is
i.e. decrease of the initial stiffness, peak and post-peak invoked by additively decomposing the rate of elas-
strength as well as the enhanced contractancy, are quite tic volumetric strain ε̇ev into mechanical and thermal
similar to those described earlier with regard to Figure components as
3. The proper description of these mechanisms entails
recourse to a robust thermo-elastoplastic constitutive
model, as described in the next section.
3 CONSTITUTIVE STUDY
This section presents a non-isothermal constitutive for-

mulation for describing the behavior of Firebag oil
sands.The data obtained from lab experiments are used
to extend a simple Mohr-Coulomb-based constitutive
model, henceforth referred to as the reference model,
whose elaborate description is given in Pietruszczak where K0 is the bulk modulus at reference temperature
(2010). This extension is accomplished by incorporat- T0 and reference mean effective stress p0 . Other mate-
ing temperature effects in the various components of rial parameters such as β1 = 1 and α1 describe vari-
the reference model such as temperature-dependence ations of Young’s modulus with mean effective stress
of the elastic and hardening rule, yield function and and temperature, respectively, while αT is the coef-
plastic flow rule, among others. ficient of volumetric thermal expansion of the solid
Developed for describing the behavior of cohe- skeleton. Eq. (1) preserves the path-independence of
sionless soils, the reference model involves a few the thermo-elastic volumetric strains, that is the elastic
simplifying assumptions as: (i) no dependency of the volumetric strains do not depend on the order of the
yield locus or plastic potential on the Lode angle, application of the incremental temperature and mean
(ii) exclusive strain-hardening behavior for both loose effective stress.
and dense samples, and (iii) no cap-type yield surface The rate of elastic shear strain ε̇eq is assumed to be
for the occurrence of the irrecoverable strains during purely mechanical, i.e. it is being merely induced by
isotropic loading. incremental deviatoric stress q̇ as
Adopting the above-mentioned assumption (i),
the model formulation is herein presented in the
triaxial stress/strain space where p = (σ1 + 2σ3 )/3,
q = σ1 − σ3 , εv = ε1 + 2ε3 and εq = 2(ε1 − ε3 )/3 with
σi and εi (i = 1, 3) being the major and minor principal
components of the stress and strain tensor, respec-
tively. With regard to assumption (ii), a simple form
of the cohesion intercept is introduced in subsection where ν is Poisson’s ratio. It is worth mentioning
3.3 to adapt the model formulation to experimental that the pressure-dependent shear modulus necessi-
observations made on the tested oil sands. Such a mod- tates the dependence of bulk modulus on deviatoric
ification is useful for describing a strain-softening type stress (Hueckel et al. 1992), in order for the thermoe-
of behavior, and shall be differentiated from the unified lastic rule to be mechanically conservative. However,
pressure- and density-dependent modeling approach, this type of coupling is trivial under monotonic load-
see e.g. Wan and Guo (1999). As for the assumption ing (Zytynski et al. 1978), and has been neglected in
(iii), temperature-induced plastic mechanisms are also the development of the present model.
assumed to be only associated with the cone-type yield
surface, i.e. Mohr-Coulomb. 3.2 Thermo-mechanical strength
Finally, the components of the stress σ and the strain
ε tensor are assumed positive in compression. The Results of triaxial compression tests show that both
overdot associated with a variable denotes its rate, and peak and post-peak strengths of Firebag oil sands
the superscripts ‘e’, ‘p’designate the elastic and plastic decrease at elevated temperatures. Other researchers
components, respectively. have reported that the peak strength of oil sand sam-
ples may remain unchanged, increase or decrease upon
heating, e.g. see (Agar et al. 1987, Kosar 1989, Wong
3.1 Thermo-elasticity
et al. 1993). The lack of consensus among these stud-
As mentioned earlier, the initial stiffness of Firebag ies entails the use of a generic temperature-dependent
oil sands determined from triaxial compression tests mathematical expression to uniformly describe the
288
Eq. (4) is employed to define the family of yield loci
f , i.e.
where MmT and ξmT represent suitable functions of the

p
plastic shear strain εq and temperature T , for prescrib-
ing hardening or softening in the behavior of oil sands,
that is:
Figure 4. Non-isothermal MC criterion used to describe the

peak strength of Firebag oil sands. where M0 is the stress ratio at incipient yielding at
ambient temperature T0 , a is the evolution parameter
controlling the pace of mobilization of friction angle
complex thermo-mechanical strength of oil sands. For
ϕ0 , ξ0 and b are the material parameters for prescribing
simplicity sake, the well-known Mohr-Coulomb (MC)
values of peak stress and strain, respectively, and α3
criterion is adapted here to serve as a non-isothermal
is a material parameter for controlling the degradation
strength limit.
of ξ0 with temperature.
The MC criterion in terms of deviatoric stress (q)
The proposed expression for ξmT , Eq. (6b), is
and mean effective stress (p) is given as:
intended to capture the rapidly strain-softening behav-
ior of oil sands as observed in the experiments. Such
a description of the strain-softening phenomenon is
linked to the loss of interlocked structure and conceiv-
ably interparticle bonding, and cannot be captured in
conventional models including those based on merely
mechanical state variables, i.e. pressure and void ratio.
where c0 and ϕ0 are cohesion and friction angle, It is noteworthy that the strain-softening phenomenon
both measured at the ambient temperature. An easy in the aforementioned sense has to be complemented
extension of Eq. (3) to non-isothermal conditions with the so-called strain localization analysis to thor-
is obtained by assuming that the p-intercept, i.e. oughly describe the observed post-peak behavior of the
c0 / tan ϕ0 , remains unchanged at elevated temper- tested samples. The latter, however, requires a numeri-
atures (Mohamadi and Wan 2016). As such, the cal analysis considering the triaxial test as a boundary
non-isothermal MC criterion is given by value problem, and is beyond the scope of this study.
3.4 Flow rule

Experimental observations on Firebag oil sand indi-
cate that the initial contractile volume changes
where α2 is a material parameter that can be smoothly transform into dilatancy in a close proxim-
used to describe temperature-insensitivity (α2 = 0), ity of the peak stress. Furthermore, it is observed that
temperature-induced decrease (α2 > 0) or temperature- the maximum contractancy increases at elevated tem-
induced increase (α2 < 0) of the strength. It is cau- peratures. These features can be adequately modeled
tioned here that Eq. (4) is only a simple expression by assuming a non-associated temperature-dependent
obtained from the curve fitting of experimental data flow rule as
and cannot be used to capture non-monotonic varia-
tions of strength with temperature. Figure 4 shows that
the peak strength of Firebag oil sands are described
quite satisfactorily as per Eq. (4).
3.3 Yield locus and hardening

A key feature of the non-isothermal model is the exis- where ε˙p and σ are the rate of plastic strain and stress
tence of a thermoelastic limit whose evolution is used tensor, respectively, λ̇ is a positive plastic multiplier,
to describe the progressive deformation of oil sands. McvT0
is slope of the critical state line in the q − p plane,
Herein, a similar mathematical form as that shown in i.e. McvT0
= 6 sin ϕ0cv /(3 − sin ϕ0cv ) with ϕ0cv being the
289
Table 2. Model parameters for Firebag oil sands.
K0 β1 ν ϕ0 (deg) ϕ0cv (deg) ξ0 d0 a b α1 α2 α3 αT
3100 0.33 0.11 41 38 4000 0.4 0.00003 500 0.981 0.1587 0.0761 –∗
∗
Not needed in the present study.
critical friction angle at ambient temperature T0 , d0 is performed at two temperatures and consolidation pres-
a material parameter controlling the post-peak dilation sures, are needed to determine all model parameters.
rate, and pc is a constant representing the intercept of For a more accurate evaluation of these parameters, it
the plastic potential g with the p-axis at the current is recommended that a testing program includes three
state of stress. different temperatures and confining pressures.
The values of K0 , β1 and ν are easily determined
from the initial slope of q − ε1 and εv − ε1 curves,
3.5 Thermo-elastoplastic constitutive equation as commonly done in geotechnical engineering. Vari-
ations of the initial slope of q − ε1 with temperature
Applying the consistency condition (f˙ = 0) to Eq. (5) can be used to determine α1 . The values of ϕ0 , ϕ0cv and
and making use of Eq. (6) and (7), the positive plastic α2 can be determined using the procedure described
multiplier λ̇ can be obtained as: in subsection 3.2. Assuming that elastic strain com-
ponents are negligible compared to plastic ones, the
hardening parameter a is determined by plotting MmT at
ambient temperature versus εq in the pre-peak regime.
Having determined ϕ0 and α2 , the values of ξ0 , b and
α3 can be evaluated such that the peak strain and post-
peak softening are precisely captured at both ambient
and elevated temperatures. The value of d0 is chosen
to describe the post-peak dilation rate. Finally, αT can
be determined based on the heating test under constant
isotropic stress prior to the shearing stage. All material
parameters are reported in Table 2 for Firebag oil sand.
where are the Macaulay brackets such that x = x

if x ≥ 0 and x = 0 if x < 0. Invoking the additive
decomposition of the rate of total strain into elastic 4.2 Simulation of drained triaxial compression tests
and plastic components, the thermo-elastoplastic con- Model simulations as per Eq. (9) were employed to
stitutive equation in a stress-temperature-controlled reproduce results of drained triaxial compression tests
program can be readily formulated as: at different temperatures, on Firebag oil sands. Figures
5 and 6 present the comparison of model simulations
with experimental results at confining pressures of
1 and 5 MPa, respectively. With regards to the stress-
strain curves, it is observed that the initial stiffness as
well as the peak and post-peak strengths are predicted
reasonably well, both at the ambient and elevated tem-
perature. It is only the peak strength corresponding to
the test at the confining pressure of 1 MPa and temper-
It is noteworthy that by setting αT = α1 = α2 = α3 = 0, ature of 130◦ C that is underestimated. This, however,
and doing simple manipulations, Eq. (9) reduces to its is due to the use of an average peak strength enve-
isothermal counterpart as described in Pietruszczak lope, including experimental data from three different
(2010). confining pressures, which has to be compatible with
the special form of cohesion intercept as described in
Eq. (6b).
Turning to the volume change behavior in Figures
4 MODEL ASSESSMENT
5 and 6 it is observed that both calculated and mea-
sured volumetric strains are in good agreement in the
4.1 Determination of model parameters
pre-peak regime, while the quality of the match in the
The proposed model requires the evaluation of three post-peak regime deteriorates. The latter, as mentioned
elastic parameters, i.e. K0 , β1 and ν, six elastoplas- earlier in the paper, is attributed to non-objectivity of
tic parameters, i.e. ϕ0 , ξ0 , d0 , ϕ0cv , a and b, and four the post-peak response of the tested oil sands due to
thermal parameters, i.e. αT , α1 , α2 and α3 . The results the occurrence of localized deformations. A numeri-
of four drained isothermal triaxial compression tests, cal analysis considering the triaxial test as a boundary
290
Figure 6. Comparison of model simulations (solid lines)
Figure 5. Comparison of model simulations (solid lines)
and experimental measurements (symbols) at p0 = 5 MPa:
and experimental measurements (symbols) at p0 = 1 MPa:
(a) deviatoric stress vs axial strain and (b) volumetric strain
(a) deviatoric stress vs axial strain and (b) volumetric strain
vs axial strain.
vs axial strain.
value problem should be carried out in order to fully

examine the localization phenomenon whereas model Athabasca area in Alberta, Canada. The results of
simulations in the present work were obtained at the lab experiments indicate that the initial stiffness as
element level. well as the peak and post-peak strength of the sam-
ples decrease at elevated temperatures, while their
maximum contractancy increases.
5 CONCLUSIONS Based on these observations a simple Mohr-
Coulomb-type constitutive model was extended to
The constitutive study of the non-isothermal behavior non-isothermal conditions. While maintaining sim-
of geomaterials in the context of continuum the- plicity, the temperature-dependency of elastic moduli,
ory of plasticity is usually performed based on the hardening rule, yield function and plastic flow rule
observed phenomenological mechanisms in the lab were addressed. The new model was applied to suc-
experiments. Such mechanisms cannot be consis- cessfully simulate the behavior of Firebag oil sand at
tently derived from previous studies on oil sands, different temperatures.
given the presence of diverse sources of complex- Although being developed based on only one series
ity governing their behavior. As a result, a series of of lab experiments, the proposed constitutive model is
high-pressure/-temperature triaxial compression tests believed to be versatile enough for describing the non-
were performed on Firebag oil sands, obtained from isothermal behavior of other oil sand types.As ongoing
291
research, the proposed model is currently being imple- Jamiolkowski, M., R. Lancellotta, & D. Lo Presti (1995).
mented in a commercial finite element package to test Remarks on the stiffness at small strains of six italian clays.
its efficiency and accuracy in simulating boundary Pre-failure Deformation of Geomaterials 1, 817–836.
value problems. Menq, F. Y. (2003). Dynamic Properties of Sandy and
Gravelly Soils. Ph.D. thesis, The University of Texas at
Austin.
Rahman, M. M. (2009). Modelling the influence of fines on
ACKNOWLEDGEMENTS liquefaction behaviour. Ph.D. thesis, The University of
New South Wales at Australian Defence Force Academy.
The authors would like to thank the Natural Sci- Rahman, M. M., S. R. Lo, & C. T. Gnanendran (2008). On
ences and Engineering Research Council of Canada equivalent granular void ratio and steady state behaviour
(NSERC) and the CMG Foundation for providing of loose sand with fines. Canadian Geotechnical Jour-
financial support of this research through a CRD grant. nal 45, 1439–14561.
Acknowledgment is also made to Suncor Energy for Salgado, R., P. Bandini, & A. Karim (2000). Shear strength
and stiffness of silty sand. Journal of Geotechnical and
providing the oil sand cores from the Firebag in-situ Geoenviromental Engineering 126, 451–462.
project. Santamarina, J. C. & m. Aloufi (1999). small strain stiffness:
A micromechanical experimental study. In Proceeding of
pre-failure deformation characteristics of geomaterials.
REFERENCES Tao, M., J. Figueroa, & A. Saada (2004). Influence of non-
plastic fines content on the liquefaction resistence of soils
Carraro, J. A. H., M. Prezzi, & R. Salgado (2009). Shear in terms of the unit energy. In Cyclic behavior of soils and
strength and stiffness of sands containing plastic of non- liquifaction phenomena.
plastic fines. Journals of Geotechnical and Geoenviro- Tatsuoka, F., T. Iwasaki, &Y. Takagi (1978). Hysteretic damp-
mental Engineering 135, 1167–1178. ing of sands under cyclic loading and its relation to shear
Chien, L. K. & Y. N. Oh (2002). Influence of fines content modulus. Soils and foundations 18, 26–39.
and initial shear stress on dynamic properties of hydraulic Thevanayagam, S. (1998). Effect of fines and confining
reclaimed soil. Canadian Geotechnical Journal 39, stress on undrained shear strength of silty sands. Journal
242–253. of Geotechnical and Geoenviromental Engineering 124,
Darendeli, M. B. (2001). Development of a new family of nor- 479–491.
malized modulus reduction and material damping curves. Wichtmann, T. & T. Triantafyllidis (2009). On the influence
Ph.D. thesis, University of Texas at Austin. of the grain size distribution curve of quartz sand on the
Drnevich, V. P. (1978). Resonant-column testing: Problems small strain shear modulus. Journal of Gerotechnical and
and solutions. Dynamic Geotechnical testing, 384–398. Geoenviromental Engineering 135, 1404–1418.
Hardin, B. O. & V. P. Drnevich (1972). Shear modulus and Yanagisawa, E. (1983). Influence of void ratio and stress
damping in soils measurment and parameter effects. Soils condition on the dynamic shear modulus of granular
Mechanics and Foundations Division 98, 603–624. media. Adv. in the mechanics and the flow of Granular
Hardin, B. O. & M. E. Kalinski (2005). Estimating the shear Materials 2, 947–960.
modulus of gravelly soils. Journal of Geotechnical and Yimsiri, S. & K. Soga (2002). Application of micromechanics
Geoenvironmental Engineering 131, 867–875. model to study anisotropy of soils at small strains. Soils
Iwasaki, T. an Tatsuoka, F. (1977). Effects of grain size and and Foundations 42, 15–26.
grading on dynamic shear moduli of sands. Soils and
Foundations 17, 19–35.
292
Effect of temperature and pressure on mechanical behavior of rock

salt in underground storage caverns
K. Khaledi, E. Mahmoudi, D. König & T. Schanz

Chair of Foundation Engineering, Rock and Soil Mechanics, Ruhr-Universität Bochum, Germany
ABSTRACT: Assessing the material response of rock salt subjected to the cyclic thermo-mechanical loading
conditions is essential for engineering design of underground storage caverns. In this paper, a coupled thermo-
mechanical model is employed to describe the effect of temperature and pressure variations on the stress-strain
behavior of rock salt around the underground cavities. In the numerical section, the operating condition of a
typical compressed air storage cavern is first simulated. Then, the effects of cyclic pressure and temperature on
the thermo-mechanical response of rock salt medium are investigated for this numerical problem and, finally,
some important design aspects such as stability of the cavern and the long term serviceability will be evaluated.
1 INTRODUCTION et al., 2012). The main objective of this paper is to

model the thermo–mechanical cyclic response of rock
Underground caverns excavated in rock salt are recog- salt around the underground energy storage caverns.
nized as the appropriate places to store energy in the Furthermore, the stability and long-term serviceabil-
form of compressed air or hydrogen. Rock salt is one ity of the system under the influence of cyclic loading
of the prime alternatives for constructing the under- condition are discussed.
ground cavities because of its favorable properties.
Low permeability, adequate thermo–mechanical prop-
erties and solubility in water are some of the unique 2 THERMO-MECHANICAL CONSTITUTIVE
features that make the rock salt distinguished from MODELING OF ROCK SALT
other rock materials. On the other hand, storing energy
in the form of compressed air inside the underground To describe the material response of rock salt, an
salt caverns is recognized as a promising technique elasto-viscoplastic-creep model has been employed in
to meet the energy demand fluctuations in electricity this paper. Under the small strain and strain additivity
power grids. This type of caverns are constructed using assumptions, the total strain rate is defined using the
the solution mining technique and operate with daily equation below.
cyclic periods. That means, the pressure and tempera-
ture applied to the inner boundary of salt cavern have
rapid changes in comparison to the seasonal storage
caverns. For this reason, it is essential to investigate
the effect of thermo–mechanical cyclic loading on the where ε̇el
ij is the elastic strain rate obtained by the gen-
vp
stability and serviceability of storage systems. Obvi- eralized Hooke’s law. ε̇ij and ε̇cr ij are the viscoplastic
ously, accurate design of these underground cavities strain rate and the creep strain rate, respectively. ε̇thij
requires adequate numerical simulations which take denotes the thermal induced strain rate.
into account the most important processes that could
affect the performance of the system. However, rare
investigations have been made to model the behav-
ior of rock salt in storage caverns considering both 2.1 Viscoplastic deformation
cyclic mechanical and cyclic thermal loading. There-
fore, design of salt caverns that work under cyclic The viscoplastic behavior of the rock salt has been
loading conditions is still a challenging task. During described by (Desai and Varadarajan, 1987) model
the recent years, a limited number of experimental which is based on Perzyna’s viscoplasticity (Perzyna,
studies have been performed to assess the effect of 1966). In this model, the viscoplastic strain rate is
cyclic loading on the mechanical behavior of rock salt defined through a non-associated flow rule as follows:
with main focus on the fatigue failure and cyclic dam-
age progress, e.g. see (Fuenkajorn and Phueakphum,
2010), (Ma et al., 2013), (Liu et al., 2014) and (Guo
293
This quantity is related to the linear thermal expan-
sion coefficient αs and the temperature changing rate
Ṫ :
2.4 Heat transfer in rock salt

The temperature fluctuation at the cavern boundary
during the charge and discharge processes changes the
temperature of the surrounding rock. To calculate the
temperature distribution around the cavern, the heat
transfer equation has to be solved. Assuming a con-
stant thermal conductivity coefficient in Fourier’s law
of heat conduction, the following heat transfer equa-
Figure 1. The viscoplastic and creep potential functions as tion is solved to obtain the rock salt temperature over
well as the failure and dilatancy boundaries of rock salt. the time and space. In this paper, the finite element
code Code–Bright is used (Code-Bright user’s guide,
where x = (x + |x|) /2 is the Macauley bracket. The 2010) to solve the equation below:
fluidity coefficient µ1 and the stress exponent N1 are
material parameters. F vp and Qvp are the proposed
viscoplastic yield and potential functions by (Desai
and Varadarajan, 1987), respectively. F0 is a reference
value with the same unit as F vp . This model accounts where cs , ρs and ks are the specific heat, density
for the compressibility and the dilatancy of rock salt in and thermal conductivity of rock salt, respectively. In
the stress space as well as the short-term failure. The addition, Ts represents the rock salt temperature.
compression behavior of rock salt in the stress space is
separated from the dilative behavior through the dila-
tancy boundary. Fig. 1 shows the applied viscoplastic 3 FINITE ELEMENT MODELING OF
potential surface, dilatancy boundary and the failure COMPRESSED AIR STORAGE CAVERN
boundary of rock salt in this paper.
To investigate the effect of thermo-mechanical cyclic
loading on the behavior of salt cavern, a finite ele-
2.2 Creep deformation ment model has been built using GiD software. GiD
The creep deformation is described using the following has been developed by the International Center for
equation: Numerical Methods in Engineering (CIMNE) and
is used as the pre-processor and post-processor for
the Code–Bright finite element solver (Code-Bright
user’s guide, 2010). In this section, the practical data
of Huntorf CAES plant is adopted to define the dimen-
sions and working conditions of the model. In Huntorf
The stress exponent N2 is a material parameter plant, the compressed air is stored in two relatively
and µ0 denotes the value of fluidity coefficient at identical storage caverns (i.e. in terms of dimension,
a reference temperature. R is the universal constant depth and storage capacity). The total storage capacity
of perfect gas and Ts is the absolute temperature of of the plant is about 300000 m3 . However, to efficiently
rock salt. According to this equation, for any J2 > 0 model the geometry of the cavern, some simplifi-
and Ts > 0, a steady-state creep deformation occurs. cations are required. In this study, only one of the
Fig. 1 shows the creep potential surface Qcr applied caverns is selected for the simulation. The geometry
in this study. In the compressibility domain(below the and boundary conditions of the selected cavern are
dilatancy boundary), the creep strain has only the devi- shown in Fig. 2a. As seen, the cavern has been ideal-
atoric part and the volumetric creep deformation is ized as a half cylinder with a radius of 18 m and a height
assumed to be zero. However, when the stress state is of 150 m. The top and bottom of the cavern have the
in the dilatancy domain (above the dilatancy bound- semi-spherical form. The axisymmetrical model with
ary), the irreversible time-dependent dilation occurs the height and width of 500 m is shown in Fig. 2a as
and volumetric creep strain increases. well. With this assumptions, the storage volume of the
simulated cavern is about 150000 m3 . On the upper
boundary of the model, a load of 10 MPa is applied
2.3 Temperature induced deformation which represents the overburden load at the top of the
The term ε̇th
ij in Eq. 1 is the volumetric strain rate caused
model. The vertical displacement at the model bottom
by the temperature change, which is considered as is restrained. The finite element discretization of the
an additional strain tensor in the mechanical model. model is shown in Fig. 2b.
294
beginning of simulation, a load which is equal to
the geostatic pressure is applied to the boundary of
the cavern. The initial temperature of rock salt is
assumed to be 50◦ C everywhere.
Phase II : in order to model the excavation process,
the applied mechanical loads to the inner bound-
ary of cavern are gradually reduced to a minimum
air pressure (in this example: 4.7 MPa). Assum-
ing a leaching rate of 35 m3 /h and considering the
time needed for debrining phase (Tryller and Musso,
2006), the whole construction phase lasts approxi-
mately 230 days. Additionally, the temperature of
the inner boundary is reduced to 35◦ C in this phase.
This temperature reduction is due to the heat trans-
fer between brine and the rock medium during the
leaching process (see Fig. 3).
Phase III : in this phase, the thermo-mechanical
cyclic loads resulted from the charge and discharge
processes are simulated. It is assumed that, during
the first 100 cycles, the cavern works under the nor-
mal operating conditions reported in (Cortogino et
al., 2001). According to (Cortogino et al., 2001),
the charge and discharge time of the cavern are 8
and 2 hours, respectively. Therefore, this simula-
tion phase takes about 41 days (i.e. days: 230–271).
The air pressure fluctuates between 4.7 MPa and
7.2 MPa, while, the temperature ranges between
30–70 ◦ C (see Fig. 3).
Phase IV : after the first 100 cycle, an extreme load-
ing scenario is defined and applied to the boundary
of the cavern for another 100 cycles. The time dura-
tion of this simulation step is about 41 days (i.e.
days: 271–312). The range of cyclic pressure drops
down to 2.2–4.7 MPa in this phase. With this pres-
sure reduction, the temperature cycles are obtained
in the range 10–80◦ C (see Fig. 3).
Figure 2. (a) Geometry and applied boundary conditions of

the model (b) the finite element discretization of the model. 4 RESULTS AND DISCUSSION
The main objective in this section is to assess the

performance of the simulated salt cavern under the
influence of defined loading conditions. In the fol-
lowing, the temperature distribution around the cavern
as well as the changes in mechanical stability and
serviceability of the cavern are discussed.
4.1 Temperature around the cavern

Because of the temperature difference between the
cavern’s inner boundary and the surroundings, the ther-
mal energy is transferred through the rock salt medium
by conduction. Because of this reason, the tempera-
Figure 3. The pressure and temperature applied to the cav- ture distribution changes in the vicinity of the cavern.
ern boundary during excavation and cyclic loading phases. During the injection process the temperature of air
The following steps are performed to simulate the inside the cavern increases. When the air temperature
construction process and cyclic loading operation of is higher than the surrounding rock, heat is transferred
the salt cavern: from the air to the rock. Therefore, the temperature
of the surrounding rock salt rises up. Over the dis-
Phase I : initial stress state is considered to be charge time, because of the reverse heat transfer from
isotropic (i.e. σxx = σyy = σzz ). Therefore, at the the rock to the air, the rock salt temperature reduces.
295
In the following, the above-mentioned stability
criteria are evaluated for the numerical example
explained in the previous section.
4.2.1 “No-dilatancy” criterion

The loading conditions of cavern have to be defined
in a way that the stresses around the cavity remain
in the compressibility zone. To determine the state of
stress with respect to dilatancy boundary, the following
safety factor is defined:

where J dil is the equation of dilatancy boundary in
Figure 4. Variation of rock salt temperature at selected √2
distances from the cavern boundary during cyclic loading the I1 − J2 plane shown in Fig. 1 and J2 is the sec-
phase. ond invariant of deviatoric stress tensor. Fig. 5 shows
the schematic representation of “No-dilatancy” crite-
rion. When FS < 1, the current stress state is below the
dilatancy boundary. In this case, the opening of micro-
cracks does not occur and subsequently, damage does
not progress. In contrary, when FS > 1, the stress state
locates beyond the dilatancy bounday. Thus, the oper-
ating condition of the cavern is not safe and cavern
may experience long-time failure due to the damage
progress. Fig. 6 compares the maximum value of FS
around the cavern obtained for the normal operating
condition and the low pressure cyclic loading. As it
is observed, during the normal cyclic operation (i.e.
phase III), the FS value is less than 0.9. That means, the
internal pressure of the cavern in this loading scenario
is high enough to keep the stresses below the dilatancy
boundary. While, during the phase IV, where the inter-
nal pressure reduces drastically, the FS factor becomes
more than one for the points located around the cavern.
Figure 5. Schematic definition of the stability criterion FS. Thus, for this loading scenario, the stresses are in the
dilatancy zone and the “No-dilatancy” criterion is not
Figs. 4 shows the variation of rock salt temperature at fulfilled.
three selected distances from the cavern boundary dur-
ing cyclic loading phases. As depicted in this figure, 4.2.2 “No-tensile stress” criterion
during the cyclic loading operation, the average tem- Rock salt has a poor tensile strength (i.e. around
perature of each cycle remains approximately 50◦ C 1.8 MPa). The thermo–mechanical loading conditions
(i.e. equal to the ground temperature). In addition, the have to be defined in a way that no tensile stress is
temperature fluctuation in the rock only takes place in experienced around the cavern. As shown in Fig. 3, for
a narrow zone less than one meter thick. Inside this the point which are very close to the cavern wall, the air
zone, the amplitude of temperature cycles reduces as temperature during the low pressure working condition
the distance from the cavern’s boundary increases and reduces to 10◦ C over the cyclic loading. This tempera-
finally, it approaches the ground temperature. ture reduction induces high thermal stresses in an area
less than one meter thick around the cavern. Thus, due
to the fast cooling and extra thermal contraction, the
tangential component of stress becomes positive at the
4.2 Stability of the cavern points which are very close to the boundary (in par-
In this section, the mechanical stability of the simu- ticular at the cavern roof). Fig. 7 shows the variation
lated cavern is investigated considering two criteria: of principal stresses at cavern roof during excavation
(days: 0–230), normal cyclic loading (days: 230–271)
1. “No-dilatancy” criterion: this criterion indicates and extreme working condition (days: 271–312). As it
whether the stress state around the cavern is in the is seen from this figure, the minimum principal (i.e.
dilatancy zone or not. in this case the tangential stress) becomes positive
2. “No-tensile stress” criterion: rock salt exhibits poor during the last phase of simulation. Therefore, “No-
tensile strength. Therefore, no tensile stress should tensile stress” criterion is not satisfied for this loading
be experienced around the cavern. condition.
296
Figure 6. Contour plot of factor of safety FS at the end of (a) phase III (b) phase IV.
4.3 Serviceability of the cavern changing of VL values for the defined loading scenar-
ios in this paper. As it is observed, in phase IV, when the
The serviceability of the system is strongly related to
internal pressure drops down, the rate of volume con-
the storage capacity of the cavern. If, for any reason,
vergence increases significantly in comparison to the
the volume of the cavern reduces significantly, the effi-
normal working condition. In this case, the increased
ciency and the serviceability of the system will be in
creep strain rate resulted from the higher deviatoric
danger. For this reason, it is important to control the
stresses accelerates the cavern closure. Thus, the rate
factors which may increase the rate of cavern closure.
of volume loss increases in this case.
In this section, the volume loss of the cavern (VL) dur-
ing the cyclic loading phase is evaluated for the defined
loading scenarios. The volume loss of the cavern is
calculated as: 5 CONCLUSION
In this paper, an elasto-viscoplastic-creep model has

been employed to describe the stress-strain relation of
rock salt under the influence of thermo-mechanical
cyclic loading conditions. In order to investigate the
where, V0 and Vt are the initial defined volume and effect of internal temperature and pressure on the sta-
the volume after time t, respectively. Fig. 8 shows the bility and serviceability of cavern, its operation under
297
serviceability of the system are affected by the inter-
nal pressure and temperature. The numerical example
presented in this work demonstrates that the stability
factors of the system are strongly governed by the inter-
nal pressure and temperature of the cavern. For this
reason, these quantities should be set in a way that the
stresses around the cavern satisfy the “No-dilatancy”
and “No-tensile stress” criteria.
ACKNOWLEDGMENT
This work was performed in the frame of the

project ANGUS+ funded by the Federal Ministry of
Education and Research (BMBF) under Grant no.
03EK3022C. The authors are grateful for this support.
Figure 7. The minimum principal stress at cavern roof dur-
ing excavation (days: 0-230), normal cyclic loading (days:
230–271) and extreme working condition (days: 271–312). REFERENCES
Code-Bright user’s guide (2010). Department of the Geotech-
nical Engineering and Geosciences of the Technical
University of Catalonia (UPC).
Cortogino, F., K.-U. Mohmeyer, & R. Scharf (2001). Huntorf
caes: More than 20 years of successful operation. In SMRI
Spring Meeting, Orlando, 23-24 April, pp. 351–357.
Desai, C. & A. Varadarajan (1987). A constitutive model
for quasi-static behavior of rock salt. J. Geophys. Res 92,
445–456.
Fuenkajorn, K. & D. Phueakphum (2010). Effects of cyclic
loading on mechanical properties of Maha Sarakham salt.
Eng. Geol. 112(1–4), 43–52.
Guo, Y., C. Yang, & H. Mao (2012). Mechanical properties of
Jintan rock salt under complex stress paths. Int. J. Rock.
Mech. Min. Sci 56, 54–61.
Liu, J., H. Xie, Z. Hou, C. Yang, & L. Chen (2014). Damage
evolution of rock salt under cyclic loading in unixial tests.
Figure 8. The volume loss of cavern during normal opera- Acta Geotechnica 9, 153–160.
tion (days: 230–271) in comparison to extreme cyclic loading Ma, L. J., X.Y. Liu, Q. Fang, H. F. Xu, H. M. Xia, E. B. Li, S. G.
(days: 271–312). Yang, & W. P. Li (2013). A new elasto-viscoplastic damage
model combined with the generalized HoekBrown failure
criterion for bedded rock salt and its application. Rock.
the normal and extreme loading conditions has been
Mech. Rock. Eng. 46, 53–66.
analyzed numerically. Obtained results show that dur- Perzyna, P. (1966). Fundamental problems in viscoplasticity.
ing the normal operating condition, the applied loads Rec. Adv. Appl. Mech 9, 243–377.
to the rock salt medium generate only thermoelastic Tryller, H. & L. Musso (2006). Controlled cavern leaching in
and steady-state creep deformations without any dila- bedded salt without blanket in Timpa del Salto. In SMRI
tion. However, when the loading pattern of the cavern Spring Meeting, Brussels, Belgium, 30 April-3 May.
is changed to the extreme condition, the stability and
298
CTE analysis of saturated cement-based sensible heat storage materials

ABSTRACT: The linear coefficient of thermal expansion (CTE) of three cement based energy storage materials
in saturated condition is determined experimentally. The CTE is a material property that indicates the extent to
which a material expands or contracts upon heating or cooling respectively and plays a vital role in the thermo-
hydro-mechanical analysis of sensible heat energy storage systems. Throughout their lifetime, sensible heat
energy storage systems are subject to frequent temperature changes due to heating and cooling effects of loading
and unloading of heat energy respectively. The different components of sensible heat energy storage system such
as heating/cooling element, energy storage material, insulators and others, are constructed of different materials
which possess un-equal CTE values, and hence produce extra stress in the system when heated or cooled. When
unaccounted for, the extra stresses generated due to the heating/cooling processes may cause damage or failure
of the thermo-hydro-mechanical system.
1 INTRODUCTION utmost importance for the design and safe operation of

such systems. However, the CTE is a much neglected
With few exceptions, all solids expand or con- material property and the level of material science is
tract when heated or cooled respectively undergoing much more primitive. Furthermore, on top of the lack
reversible dimensional changes on a macroscopic level of experimental data of CTE analysis of porous media
due to a change in thermal vibrations that are always sensible heat storage materials, traditional methods of
present in the crystal – which is known as thermal CTE determination assume a uniform temperature dis-
expansion or dilation. However the extent of expansion tribution within materials while heating or cooling,
or contraction per degree of change in temperature is which is far from reality. The actual heat flow within
different for different materials. Analysis of the coeffi- sensible heat storage materials to-and-fro the sample
cient of thermal expansion (CTE) of porous materials via the heating/cooling element creates an inhomoge-
is different from that of solids such as metals and neous linearly varying temperature distribution within
ceramics due to the presence of pore water and/or pore the sample.
air. In a porous multiphase material, thermal expan- In accordance with this phenomenon, a thermal
sion of in situ pore water contributes to the volumetric expansion and conductivity meter is used in this
change of the material due to a change in the ground research to obtain the CTE of three commercial
surface temperature, thus influencing the stress field heat storage samples - namely, Füllbinders L and M
of the medium. (SCHWENK Zement KG1 ) and Dyckerhoff Dämmer
In the past, several studies have been conducted on S (Deuna Zement2 ). The samples were prepared with
the thermal expansion of porous media and soils in a sufficient diameter to length ratio to ensure a homo-
particular. Agar et al. (1986) analysed both the drained geneous linear variation of temperature within the
and undrained thermal expansion coefficients for oil materials while heating. The accuracy of the system
sands by means of heating experiments in the range of was checked by testing standard materials of alu-
temperature between 20 and 300◦ C. Cui et al. (2000) minum, stainless steel and brass with known CTE
described two phenomena produced during heating of values, and a good agreement between the measured
saturated clays - namely, expansion of soil constituents and the standard values was obtained.
(solid and water) and mechanical weakening of the
contacts between soil aggregates. Saix et al. (2000)
observed a contraction during heating of clayey silty 2 THEORETICAL ANALYSIS
sand under constant stress at 42, 160 and 800 kPa
in a one dimensional confined compression test. On The temperature dependent linear coefficient of ther-
compacted FEBEX bentonite and Boom clay, Romero mal expansion α is calculated as per unit change in
et al. (2005) observed a thermal expansion under low
stresses.
Analysis of CTE of the several components of a 1
http://www.schwenk-zement.de/
porous media based sensible heat storage system is of 2
www.deuna-zement.de
299
length with per unit change in temperature and is Table 1. Properties of the heat storage materials.
normally expressed as follows:
Porous material
Properties Fü. L* Fü. M* Dä. S*
where, Lo is the initial sample height (length) at room Density (kg m−3 ) 1583 1609 1560
Porosity (–) 0.543 0.518 0.562
temperature, L is the change in height (length) of Thermal conduc. (W m−1 K−1 ) 0.960 0.965 0.957
sample due to a temperature change of T . This is an Specific heat (J kg−1 K−1 ) 2083.4 1957.1 2162.8
average coefficient of thermal expansion, which is the Modulus of elasticity (MPa) 1911.1 2751.5 1987.9
one normally provided in the literature, and is not the
same as the differential form (dl/dt) of α. For practical * Fully saturated condition.
applications specifying an average value of α is often
adequate. Moreover, α is a function of temperature and
hence when referring to α of a specific material, the 3 EXPERIMENTAL PROGRAM
temperature range over which the measurements were
made is usually specified. 3.1 Tested materials
Heating or cooling affects all the dimensions of a
material, with a resultant volume change. The volume In Table 1, obtained material properties of the three
coefficient of thermal expansion αv represents the vol- investigated cement-based thermal energy storage
ume change of a material due to heating/cooling and materials, Füllbinders L & M and Dämmer S, are pre-
is obtained as: sented. The samples were prepared with a water to
solids ratio of 0.8 and were stored under water for 28
days. Storage in water ensures full saturation and pre-
vents possible cracking of samples which may happen
particularly on the specimen surface soon after sample
where, where Vo and V are the original volume and preparation due to the hydration of cementing material.
the volume change due to a temperature change of T
respectively. Many materials have anisotropic values 3.2 Experimental set-up
of α depending on the crystallographic direction along
which CTE is measured. For materials in which the Determination of the thermal expansion coefficient of
thermal expansion is isotropic, αv can be approximated materials undergoing a thermal cycle requires the mea-
as 3α. surement of two physical quantities – namely, displace-
The mean CTE of the heat storage materials investi- ment and temperature. The three main technologies
gated in this research is calculated according to ASTM used for the measurement of CTE are interferometry,
E831 (2006) guidelines for the determination of linear dilatometry and thermomechanical analysis. Interfer-
thermal expansion of materials by thermomechanical ometry works on the principle of measuring displace-
analysis as follows: ment of the specimen ends with optical interference
techniques in terms of the number of wavelengths of
monochromatic light. It is the most precise technique
of the three methods and is capable of measuring CTE
values lower than 5 × 10−6 /◦ C. Mechanical dilatom-
etry techniques involve the application of heat to the
specimen in a furnace and the measurement of the
displacement of the specimen by means of an assem-
where, α is the mean coefficient of linear thermal bly of a sensor and push rods. The test precision
expansion of the sample (×10−6 /0 C), αref is the mean is lower than that of interferometry and is generally
coefficient of linear thermal expansion of reference suitable for materials with CTE above 5 × 10−6 /◦ C.
material (x10−6 /0 C), k is calibration coefficient, Lo is Thermomechanical analysis measurements are made
the initial specimen length at room temperature (m), with a thermomechanical expansion meter consisting
Lref is the change of reference material length due of a specimen holder, a transducer that measures the
to heating (µm), Lref is the reference material initial change in length of the specimen upon heating, a fur-
length at room temperature (m), L is the change of nace or heating plate for uniform heating of specimen,
specimen length (µm), T ref is the temperature dif- a thermocouple or temperature sensing element and a
ference over which the change in reference material means of data recording apparatus. The lower limit for
length is measured (◦ C), T is the temperature dif- CTE determination with this technique is 5 × 10−6 /◦ C.
ference over which the change in specimen length However, the method may also be used at lower or
is measured (◦ C). In this study, Polyoxymethylene negative expansion levels with decreased precision &
(POM) thermoplastic is selected as a reference mate- accuracy.
rial due to its high stiffness, low friction, excellent In this research, a thermal expansion and conductiv-
dimensional stability and known coefficient of linear ity meter which works on the principle of thermome-
thermal expansion. chanical analysis is used (Fig. 1, top). The apparatus is
300
a diameter of 50 mm and sufficient length of 40 mm
to produce a linear variation of temperature within the
samples while heating. In addition, the higher speci-
men length results in a greater length change signal
upon heating or cooling and hence provides a higher
CTE accuracy. Care was taken to ensure that the top
and bottom faces of the samples are perfectly flat
so that the minimum contact force applied over the
samples is spread out uniformly over a wide enough
surface area avoiding minor sample deformations. The
accuracy of the system was checked by testing sam-
ples of aluminum, stainless steel and brass with known
CTE values.
For the porous material specimen shown in Figure
1 (middle), under isotropic medium conditions, where
local thermal equilibrium is ensured (Ts = Tf = T ,
where Ts and Tf are the temperatures of the solid and
fluid phases respectively), heat transfer Equations 5–8
can be obtained by combining heat transfer equations
in solid and fluid phases (Bejan 2004, Bejan & Lorente
2004).
where, the subscripts m, s and f refer to the porous

medium, solid and ?uid phases, respectively, n is poros-
ity, ρ is density, C is the specific heat of the solid, Cp
is the speci?c heat at constant pressure of the fluid, λ
is the thermal conductivity, q (W m−3 ) is the heat
production per unit volume, T is temperature, t is
Figure 1. Thermal expansion measurement set-up (top), time and ν is the flow velocity field. For the system
schematic representation (middle) and dimensional analysis shown in Figures 1 (bottom), Equation 9 applies pro-
(bottom) of the thermal expansion and conductivity cell. vided that T1 > T2 > T3 , ensuring that heat flow occurs
from the top plate via the sample and reference disc to
modified to produce a homogenously linear tempera- the bottom plate with a linear temperature gradient as
ture distribution within the specimen to represent the the specimen is homogeneous, and T1 > Tambient > T3 ,
actual heat flow and distribution within sensible heat ensuring that heat flow from the environment does not
storage systems. penetrate to the specimen (Stegner et al. 2011, Sass &
The apparatus consists of a top heating plate, a bot- Stegner 2012).
tom cooling plate and a reference disc with known
thermal conductivity (Fig. 1, middle). The specimen
is sandwiched between the top heating and reference
plates. The plates consist of extremely thin PT 100
temperature sensors with an accuracy of 0.05◦ C. The
system measures sample expansion or contraction with The average sample temperature (at the center of the
a TRS-0050 linear transducer with an independent specimen) which corresponds to T1 , T2 and T3 can thus
linearity of 0.15% and repeatability of 2µm. be obtained as:
The sample expansion tests were conducted under
fully saturated condition by heating the top plate at a
constant rate of 2◦ C/min within the temperature range
of 22–75◦ C, as the materials are to be used for the
purpose of a sensible heat storage system with a heat
carrier fluid (water) for the loading/unloading opera- where, T1 is the temperature of the top (heating) plate
tions of the system. The samples were prepared with (◦ C), T2 is the temperature of the reference disc (◦ C),
301
Table 2. Comparison of the measured and standard CTE Table 3. Measured CTE values of the three heat storage
(22–75◦ C) values of the reference materials. materials.
Material Measured CTE* CTE, Agil. Tech.* Material Measured CTE*
Aluminium 23.77 23.60 Füllbinder L 10.69

Stainless steel 14.54 14.90 Füllbinder M 16.13
Brass 19.87 19.10 Dämmer S 11.03
* ×10−6 /◦ C. * ×10−6 /◦ C.
The thermal expansion test results of the reference
materials matched closely with the standard CTE val-
ues proving the accuracy of the system. Similarly,
a close match between experimental and numerical
results is obtained for the heat flow and temperature
distribution across the specimen and reference plate
for all the materials.
4.2 Testing of heat storage materials

In Table 3 and Figure 3, results of thermal expansion
tests on the three cement-based sensible heat storage
materials obtained by heating the top plate at a constant
rate of 2◦ C/min within the temperature range of 22–
75◦ C are presented. Since all the tests are conducted
under drained condition, volume change caused by
pore water expansion is negligible. Both Füllbinder L
Figure 2. Analysis of reference plate temperature T2 of the
and Dämmer S exhibit a higher rate of expansion upon
studied invar alloy and the reference materials.
heating for average sample temperature range of up to
30◦ C followed by a gradual decrease afterwards due
T3 is the temperature of the bottom (cooling) plate to material softening. However, Füllbinder M shows
(◦ C), T is the average sample temperature (◦ C), λp a consistent rate of expansion upon heating for the
and λv are the thermal conductivities of the sample considered temperature range. Most solids exhibit a
and the reference disc respectively (W m−1 K−1 ), Sp , Sv constant rate of expansion upon heating at lower tem-
andS23 are the distances (m) as shown in Figure 1 (bot- perature ranges, usually followed by an increase in the
tom). The average sample temperature obtained using rate of expansion at very high temperature ranges.
Equation 10 is used in the temperature calculations of The results are also consistent with the proper-
Equations 3 & 4. ties and composition of the investigated heat storage
materials. Both Füllbinder L and Dämmer S have a
comparatively lower measured strength and modu-
4 RESULTS AND DISCUSSION lus of elasticity, a lower cement binder content and
a higher porosity when compared with Füllbinder M,
4.1 Testing of reference materials and hence show a higher rate of softening or mechan-
Table 2 shows the observed and standard (Agilent ical weakening of cementitious bonds and a decrease
Technologies 2002) CTE values of the three reference in the rate of expansion upon heating.
materials used in this study to verify the accuracy of Figure 4 shows the temperature distribution across
the experimental results. The baseline correction due Aluminium (reference material) and Füllbinder L (heat
to expansion of the components of the system while storage material) samples heated up to a tempera-
heating was obtained by testing a specimen of an inert ture of 80◦ C. The results, obtained using COMSOL
or low expansivity invar alloy. Multiphysics, show a homogenous and linearly vary-
In Figure 2, the measured reference plate tempera- ing temperature distribution across the samples. As
ture T2 values of the three reference materials and invar expected, a higher temperature across the reference
alloy are compared with numerically obtained (using plate is recorded for the Aluminium specimen as com-
COMSOL Multiphysics3 ) reference plate temperature pared with the Füllbinder L specimen due to its higher
of a specimen with thermal property of an average of thermal conductivity.
the invar alloy and the three reference materials used.
5 CONCLUSIONS
3
COMSOL Multiphysics is a finite element analysis, solver
and simulation software package for various coupled phe- A thermal expansion and conductivity meter which
nomena in physics and engineering applications works on the principle of thermomechanical analysis
302
Figure 4. FE-mesh and system outline (top), temperature
distribution of the Aluminium reference material (middle)
and temperature distribution across the Füllbinder L spec-
Figure 3. Plot of specimen expansion vs temperature at the imen and reference plate (bottom) [All temperature values
top of specimen (top), specimen expansion vs temperature shown are in ◦ C].
at the center of specimen (middle) and average CTE values
vs the upper range of temperature used for calculating the
average CTE (bottom) of the three heat storage materials [For
the calculations of the average CTE values in the bottom
figure, a bottom temperature range of 24◦ C is used for all heating or cooling, the linear temperature distribu-
points]. tion across the specimen represents the actual heat
flow and temperature distribution within the porous
was modified to determine the linear coefficient of medium of sensible heat storage systems upon load-
thermal expansion (CTE) of three cement based sen- ing/unloading operations via fluid carrier pipes. The
sible heat storage materials in saturated condition findings of this research provide an accurate way
by maintaining a homogeneously linear temperature of estimating the CTE of sensible heat energy stor-
distribution across the investigated materials. Unlike age materials, which plays a fundamental role in the
traditional CTE determination methods which assume thermo-hydro-mechanical analysis and design of such
a uniform temperature distribution in a material upon systems.
303
ACKNOWLEDGEMENTS ASTM. 2006. ASTM E831-06: Standard test method for
linear thermal expansion of solid materials by thermo-
The authors would like to gratefully acknowledge the mechanical analysis.
financial support provided by the German Federal Bejan, A. 2004. Convection Heat Transfer (3rd ed.), New
York: Wiley.
Ministry for Economic Affairs and Energy (BMWi)
Bejan, A., Lorente, S. 2004. The constructal law and the ther-
under Grant numbers 0325547B and KF3067302HF3, modynamics of flow systems with configuration, Inter-
as well as the support of Project Management national Journal of Heat and Mass Transfer 47(14–16):
Jülich (PTJ). 3203–3214.
Stegner, J., Nguyen, D., Seehaus, R., Sass, I. 2011. Develop-
ment of a thermal conductivity and diffusivity meter for
REFERENCES unconsolidated rocks. Proceedings of the 18. Tagung für
Ingeniergeologie, Berlin (original in German).
Agar, J.G., Morgenstern, N.R., Scott, J.D. 1986. Thermal Sass, I., Stegner, J. 2012. Coupled measurements of ther-
expansion and pore pressure generation in oil sands. mophysical and hydraulical properties of unsaturated and
Canadian Geotechnical Journal 23: 327–333. unconsolidated rocks. Proceedings of the 37th Work-
Cui, Y.J., Sultan, N., Delage, P. 2000. A thermo-mechanical shop on Geothermal Reservoir Engineering, Stanford
model for saturated clays. Canadian Geotechnical Journal University, Stanford, California, 30 Jan.–01 Feb., 2012.
37(3): 607–620. Agilent Technologies. 2002. Laser and Optics User’s Manual,
Saix, C., Devillers, P., El Youssoufi, M.S. 2000. Elément Chap. 17, Material expansion coefficients.
de couplage thermomécanique dans la consolidation de
sols non saturés. Canadian Geotechnical Journal 37:
308–317.
Romero, E., Villar, M.V., Lloret, A. 2005. Thermo-hydro-
mechanical behaviour of heavily overconsolidated clays.
Engineering Geology 81: 255–268.
304
A laboratory investigation on the thermo-mechanical

behaviour of thermal piles in sand
A. Rafiei & C. Arya

University College London, London, UK
S.W. James
Cranfield University, Cranfield, UK
R.G. Correia
University of Nottingham, Nottingham, UK
R. Fuentes
University of Leeds, Leeds, UK
ABSTRACT: This paper reports on an investigation on the behaviour of base and shaft resistant piles embedded
in sand subject to heating and cooling cycles. It describes the 1 g laboratory model developed at University
College London. The test pile consists of a hollow stainless steel tube which is instrumented using Fibre Bragg
Grating (FBG) sensors that allow measurement of strain and temperature profiles along the pile length and the
surrounding soil. The pile is subjected to two heating-cooling cycles. Water held at 50◦ C is circulated inside
the pile for 24 hours; this is followed by a 24-hour cooling period during which the temperature of the water
is allowed to fall back to ambient conditions. The results show that small irreversible settlements occurred
during this period. Additionally, it is found that the degree of freedom varies between 0.9 and 1.0 which indicate
low levels of restraint are present. Nevertheless, considerable axial load and axial stress is induced in the pile.
Moreover, significant shaft friction is mobilized during both heating and cooling periods.
1 INTRODUCTION may compromise the safety of piles. Normally, to avoid

the risk of failure, higher safety factors are used by
Thermal piles are a newly developed technology which designers.
not only provide structural support but also allow heat A significant amount of the research on thermal
at shallow depths in the soil to be extracted from the piles has been carried out in the field by a num-
ground during cold periods and excess heat from the ber of authors, most notably Laloui et al. (2006) and
overlying structure to be deposited in the ground dur- Bourne-Webb et al. (2009). However, in-situ tests are
ing warm spells. This is achieved by circulating fluid expensive, time consuming and also it is difficult to
through pipes embedded in the pile. If the temper- ascertain boundary conditions. Therefore, laboratory
ature of the circulated fluid is lower than the soil, tests were used in this work. A 1 g small scale model
the fluid will absorb heat from the soil which can be was designed and developed to perform a series of
subsequently recovered via a heat pump and used for monotonic and cyclic mechanical only, thermal only
heating. Alternatively, if the temperature of the fluid and thermo-mechanical loading tests. Due to space
is higher than the surrounding soil, the ground can limitations, this paper only presents the results of the
be used as a heat sink. Thermal piles are increas- thermal only tests. Fibre Bragg Gratings (FBG), a type
ingly being used in major projects worldwide for of point based fibre optic sensor, are used to monitor
heating and cooling purposes. However, despite the strain and temperature changes along the pile surface
increasing number of installations, evidence of their and inside the soil bed. Further details of the apparatus
thermo-mechanical performance is still rather limited. and test procedure are presented below.
The research on thermal piles is mainly divided into
2 subjects: heat transfer within the pile and the sur-
rounding soil and thermo-mechanical behaviour under 2 APPARATUS
combined loading conditions. The work reported in
this paper addresses the later subject. Thermal loading An apparatus capable of assessing the behaviour
induces additional strains and hence increases axial of both shaft resistant and base resistant piles is
loads and stresses on the pile. These additional stresses developed. The tank is made from a 3 mm-thick steel
305
sheet with an outer diameter and height of 500 mm. A
polished round tube made of grade 304 stainless steel
with an external diameter of 28 mm and height and
wall thickness of 535 mm and 1.5 mm respectively is
used to model the pile. The bottom of the model pile is
capped by welding a base plate also of grade 304 stain-
less steel in order to allows water to be circulated inside
the pile. For the shaft resistant test, the toe of the pile
passes through a hole in the container base. Container
and pile dimensions are chosen to make sure that the
results obtained are not being affected by the boundary
conditions. In the literature, the ratio between the pile
diameter and the container diameter is chosen as a fac-
tor to minimize these boundary effects. As such, the
ratio between the diameter of the container and model
pile is 17.85 which lies near the limits specified by
Parkin & Lunne (1982). Another boundary condition
is introduced by Al-Mhaidib & Edil (1998), namely
zone of influence, and determined that this is between
3D-8D for a pile installed in sand: this limit is also
comfortably met in this test. Figure 1. Shaft resistant pile set-up.
The soil used in this study is a poorly graded uni-
form fine sand with a uniformity coefficient of 2.32
and a median grain size of D50 = 0.18 mm. The aver- sensors an accurate monitoring tool. The main rea-
age specific gravity is 2.64 gr/cm3 and the maximum son for using FBG in this study is the multiplexing
and minimum dry densities are 1.658 and 1.349 gr/cm3 potential of FBG sensors where several sensors can
respectively. A relative density of 57.6% is measured be fabricated on a single fibre which reduces required
in all tests which classifies it as medium dense. The space for the installation of fibres. In this study, two
size of the sand grains in the reduced scale model is set of fibres each with 5 FBG sensors were fabri-
not scaled down with the rest of the system. Several cated at Cranfield University to monitor temperature,
recommendations are found in the literature based on TP1-TP5, and strain along the pile surface, SP1-SP5.
the relationship between the pile diameter and median Moreover, 5 set of fibres each consisting of 4 FBGs
grain size. In this study, this ratio is equal to 155 which are used to monitor the temperature in the soil bed at
is above the recommended limits specified in the lit- different levels and distances, T1-1 to T5-4 (Figure 1).
erature, namely 35 to 100 (King et al. 1984, Weinstein Data collected for temperature variations in the soil is
2008). In order to prevent heat loss, an encapsulated not presented in this paper though as the focus is on the
fibreglass insulation jacket is used to cover the top pile itself. Pile head displacement is measured using a
and bottom surfaces as well as the sides of container. Linear Variable Differential Transformer (LVDT) with
The insulation jacket has 8 mm thickness and a ther- measurement range of ±7.5 mm and data collection
mal conductivity value of 0.04 (W/m.K). A Techne frequency of 3 seconds.
compact water circulator model C-400 with tempera-
ture range of −20◦ C and +80◦ C is used to recirculate
thermostatically-controlled water through tubes con- 3 EXPERIMENTAL PROCEDURE
nected to the pile head. The circulator is not able to
cool the water and this is achieved by simply allowing A repeatable sample preparation method is used for
the water to fall back to room temperature. both tests to reach a relative density of approximately
The model pile is instrumented using Fibre Bragg 57%. Driving the pile in the sand bed was not an option
Gratings (FBG) which is one of the most commonly as it would damage the fragile FBGs surface sensors
used point-based optical fibres. Application of Fibre despite using protective coatings. In order to perform
Optic Sensors (FOS) for health monitoring of geotech- the shaft resistant test, the pile is held in position at the
nical structures including foundations and tunnels has centre of the container using a support at the pile head.
increased significantly over the past decade. Applica- Sand is poured into the container in 16 layers and each
tion of FOS for in-situ installation of Thermal Piles layer is levelled using a wooden tamper made partic-
is reported by Bourne-Webb et al. (2009). Using ularly for this purpose. Temperature FBG sensors are
FOS instead of conventional monitoring instruments placed at 5 levels. One end of the FBG sensor is placed
have several advantages including small size, high next to the pile surface and it is laid on the sand until
sensitivity, large bandwidth, automated and fast data it is sealed to the container wall and the other end is
acquisition, being immune to water and not using elec- attached to the laser box. Once the soil and sensors
trical signals (Iten, 2011). Typical strain resolution for are in place, the inlet and outlet pipes are connected
distributed sensors is around 20µε while for point- to the circulator and pipes are insulated to reduce the
based sensors it is 1µε which has made point-based heat loss during water circulation. Before starting the
306
Figure 2. Temperature changes along the pile (a) Shaft Figure 3. Observed strain distribution (a) Shaft resistant pile
resistant pile (b) Base resistant pile. (b) Base resistant pile.
test, the pile base support is removed which results in

0.05 mm (0.17% of the pile diameter) pile head move- used as a more reliable tool (See Figure 2). For the
ment due its self-weight. Water circulator is turned on shaft resistant pile, the maximum changes in temper-
and the temperature is raised to 50◦ C in less than 10 ature during heating and cooling periods are 31.09◦ C
minutes. During each heating-cooling cycle, the tem- and −29.79◦ C from each particular period. These are
perature is kept constant for 24 hours at 50◦ C and then recorded at TP5 and TP4 levels respectively which are
left to recover to room temperature for 24 hours and approximately 8.5% higher than initial assumed value
the same procedure repeated in the second cycle – see of T = 50 − 22 = 28◦ C. For the base resistant pile,
Figure 2. slightly higher temperatures changes are recorded but
For the base resistant pile, the same procedure is the same trend is seen in both tests where the pile is
applied with only one difference: Initially, the sand is heated up to 50◦ C within 10 minutes and kept at 50◦ C
poured in the container layer by layer until it reaches for 24 hours and then left to cool for 24 hours and same
the height of 90 mm and then the pile base is placed procedure is repeated in the 2nd cycle.
on it after which a similar procedure as that of shaft Three type of strain values are used in the thermo-
resistant pile is repeated. The results are presented mechanical analysis of thermal pile systems. The first
below. is the observed strain collected in the laboratory, shown
in Figure 3. A residual strain of approximately 10 µε is
seen in both tests. Lower strain values are obtained in
4 RESULTS AND DISCUSSION the second cycle compared to the first one. Observed
strain is compared with the free-state strain which is
As previously noted, this paper only discusses the the product of temperature changes and thermal expan-
results of shaft and base resistant piles under thermal sion coefficient of the pile material. In this study,
only loading. Both types of piles are not restrained a separate test is performed on free-state condition
either of the top or bottom and hence the only restraint where the pile is heated up and cooled down with-
is the surrounding sand. Pile temperature is increased out any soil restraint or end restraint and strain values
from 22◦ C to 50◦ C and data has shown that varia- collected are used as the free-state strain values. The
tions in the temperature is seen at different depths difference between observed and free state strain is
during heating and cooling periods. Hence, instead called Restrained Strain which represents the strain
of using absolute temperature values, constant room caused by the friction between pile and sand at the
temperature at 22◦ C is considered as the reference interface. It is this restrained strain that causes induced
and temperature variations from room temperature is loads (Bourne-Webb et al. 2009).
307
Figure 4. Degree of freedom variations along the depth (a) Figure 5. Induced axial force distribution along the pile (a)
Shaft resistant pile (b) Base resistant pile. Shaft resistant pile (b) Base resistant pile.
A term called ‘degree of freedom’ is used as the

ratio of observed strain over free strain to quantify
the amount of restraint. The value of degree of free-
dom varies between 0 and 1, where 0 means maximum
restraint. Existence of restraint will induce axial load
and consequently additional induced stresses in the
pile. Figure 4 shows the values of degree of freedom
along the pile length achieved in both tests. For the
shaft resistant pile, the maximum pile restraint is seen
at the bottom of the pile. The maximum and minimum
degrees of restraints are experienced at the end of first
cooling and heating periods respectively. For the base
resistant pile, similar trend as shaft resistant pile is
seen for the heating and cooling cycles although there
are less differences in the degree of freedom along
its depth. In both tests, the value of degree of free-
dom varies between 0.9 and 1.0 which shows that a
relatively small restraint is seen from the surrounding
sand.
Despite these small values of restrained strain and
high degree of freedom, a significant axial load is
induced in the pile from thermal loading. It is mainly
due to the high thermal expansion coefficient of stain-
less steel and the magnitude of temperature changes in
the pile. For the shaft resistant pile, the highest induced
load during heating and cooling periods are 860N and
1050N respectively. During the cooling period, ten-
sion load is also observed – see Figures 5 and 6. This
would be very significant for concrete piles. For the
base resistant pile, lower induced axial load is applied Figure 6. Induced axial stress distribution along the pile (a)
Shaft resistant pile (b) Base resistant pile.
to the pile due to the higher degree of freedom.
308
Figure 7. Pile head displacement (a) Shaft resistant pile (b)
Base resistant pile.
Induced stress along the pile surface for both shaft

and base resistant piles are shown in Figure 6. For
the shaft resistant pile, the maximum induced stress
in seen in the first cooling period at the bottom of the
pile where it has reached up to −8.4 MPa. For the base
resistant pile, the maximum induced stress is recorded
at 95 mm depth as −5.57 MPa.
Pile head displacements under thermal cycles for
both shaft and base resistant piles are shown in Fig-
ure 7. For the shaft resistant pile, heave is seen during
heating and settlement during cooling of the pile. Sig-
nificant drop is seen during first cooling period where
it has gone below the original starting point and again
heaving happens due to heating and larger settlement
up to 0.22 mm is seen in the second cooling period that
shows the degradation of skin friction during heat-
ing and cooling cycles. This also explains why the
induced load is lower in the second cooling cycle than
in the first. Moreover, irreversible settlement is seen
and hence, the assumption of perfect thermo-elastic
behaviour does not hold. The base resistant pile shows
a very similar trend where the maximum settlement is
seen during the 2nd cooling period.
In response to heave and settlement of the pile in Figure 8. Mobilised shaft friction at 4 zones along the pile
the heating and cooling periods, shaft friction on the for shaft resistant pile.
pile is mobilized to resist the movement. In order to
determine the value of mobilised shaft resistance, the
formula by Laloui et al. (2006) is used in this study. Mobilized shaft friction for the shaft resistant pile
Four zones are defined for the mobilized shaft resis- is shown in Figure 8. During 2 heating periods, higher
tance from bottom of the pile between SP1 and SP2 magnitude of mobilized friction is seen in zones 1
called zone 1 to top of the pile between SP4 and SP5, and 3 during 1st heating period and in the 2nd heat-
zone 4. ing period, higher values are seen in zones 2 and 4.
309
At the end of cooling periods, higher magnitude of different cycles. As part of the on-going research, the
mobilized friction is seen during 2nd cooling periods results presented in this study for thermal only load-
in all zones except zone 3. This could be due to the ing will be used to assess the behaviour of piles under
higher magnitude of pile head displacements during thermo-mechanical loading.
the 2nd cooling period (Figure 7). Results presented
in Figure 8 for shaft resistant pile is compared with
the findings by Yavari et al. (2014) for test E2 for a REFERENCES
shaft and base resistant pile. In terms of absolute mag-
nitudes, lower magnitude of mobilized friction is seen Al-Mhaidib, A.I. & Edil, T. B. 1998. Model tests for uplift
byYavari et al. (2014), possibly due to the lower magni- resistance of piles in sand. Geotechnical testing journal,
tude of temperature increase, reaching 35◦ C compared 21(3), pp. 213–221.
Bourne-Webb, P., J., Amatya, B., Soga, K., Amis, T., David-
to 50◦ C in this study. Moreover, shaft resistant pile is son, C., Payne, P. 2009. Energy pile test at Lambeth Col-
expected to have higher mobilized friction and use the lege, London: Geotechnical and thermodynamic aspects
entire capacity of shaft friction due to the lack of sup- of pile response to heat cycles. Géotechnique, 59(3),
port at the bottom compared to shaft and base resistant pp. 237–248.
model. Iten, M. 2011. Novel applications of distributed fibre optic
sensing in geotechnical engineering. PhD thesis. ETH
Zurich.
5 CONCLUSIONS King, G.J.W., Dickin, E.A., and Lyndon, A.1984. The devel-
opment of a medium size centrifugal testing facilities.
Proceedings of The Application of Centrifuge Modelling
A laboratory test for 2 types of shaft resistant and base
to Geotechnical Design, Manchester, England, 24–46.
resistant piles embedded in sand under thermal load- Laloui, L., Nuth, M., Vulliet, L., 2006. Experimental and
ing is presented. A novel monitoring technique using numerical investigations of the behaviour of a heat
Fibre Bragg Grating optical fibre sensors is used to exchanger pile. International Journal for Numerical and
monitor temperature and strain on the pile surface Analytical Methods in Geomechanics, 30(8), pp. 763–781.
and in the soil bed. It is found that small restraint is Parkin, A. K., & Lunne, T. 1982. Boundary effects in the
applied from the surrounding sand but due to the phys- laboratory calibration of a cone penetrometer for sand.
ical properties of the model pile and also due to the Proceeding of 2nd European Symposium on Penetration
high temperature changes, significant axial loads and Testing, Netherlands National Society for Soil Mechanics
and Foundation Engineering, May, pp. 761–768.
stresses are induced in the pile. In addition to compres-
Weinstein, G.M. 2008. Long-term behaviour of micropiles
sive forces, considerable tensile force also develops in subject to cyclic axial loading. Ph.D. thesis, Polytechnic
the pile. Irreversible pile head settlement is also seen University, Brooklyn, New York. 373 pages.
for both shaft resistant and base resistant piles. Higher Yavari, N., Tang, A., M., Jean-Michel Pereira, J. M.,
settlements are seen in the second cooling period com- Hassen, G. 2014. Experimental study on the mechani-
pared to the first cooling period showing the signs of cal behaviour of a heat exchanger pile using physical
skin degradation under 2-way cyclic loading that are modelling. Acta Geotechnica, 9(3), pp. 385–398.
also evident from the distribution of induced loads at
310
Temperature dependence of elastic P- and S-wave properties of rocks:

Applications to geothermal reservoir evaluation
H.B. Motra & F. Wuttke

Institute of Geosciences, Marine and Land Geomechanics and Geotechnics,
Christian-Albrechts-University, Kiel, Germany
ABSTRACT: The rising cost of energy every year necessitates the development of alternative energy sources,
and each development involves disturbance of a natural system, which must be analysed and predicted. Geother-
mal energy, energy stored within the Earth as heat, is a topic which has garnered attention in the recent past
because of its potential as a powerful energy resource.
Various models are used to predict the effect of development, but there is an extreme lack of quantitative infor-
mation in the field of rock mechanics that is essential for this purpose. The thermal effect on rock that limit
energy-resource recovery and development can be described briefly; each area need new research studies. In
geothermal-energy exploration and production, questions arise concerning temperature and pressure effect on
mechanical properties of reservoir rock.
The increasing sophistication of seismic wave measurement, processing and recent experimental work on factors
governing wave propagation in rocks has stimulated increased interest in the use of active seismic techniques
to determine the in-situ physical state of crustal rocks for engineering applications. In this paper, we measured
the elastic P- and S-wave velocities and the velocity anisotropy of geothermal core sample considering thermal
gradients. The measurements were done at temperatures of up to 600◦ C and confining effective pressures of
100 MPa, corresponding to depths of ∼3000–4000 m. The measurements were carried out in a cubic multi-anvil
pressure apparatus, using the pulse transmission technique. The effect of temperature and pressure act in oppo-
site direction on most of rock properties. For example, P (Vp ) and S (Vs ) wave velocities, density, deformation
modulus, such as Young’s modulus and bulk modulus, and shear modulus decrease with temperature. In addi-
tion, the correlated porosity and permeability with elastic seismic wave show that at increase with temperature.
In this paper, we review experimental work showing how wave velocities in rocks are sensitive to parameters
of interest to geothermal exploration; effective pressure, temperature, Young’s, bulk, shear moduli . From the
results of temperature dependence P and S wave velocities and mechanical properties of rock possible approach
to characterisation of geothermal reservoir systems.
1 INTRODUCTION Temperature-dependent properties which may be

important include: flow characteristics such as abso-
Reliable estimates of the physical properties of rock lute and relative permeabilities; fluid-storage capacity,
are required for exploration of hydrocarbon, geother- expressed in terms of porosity, involving such char-
mal, and mineral reservoirs, as well as for subsurface acteristics as bulk and pore compressibility and ther-
storage of nuclear waste or CO2 . Temperature is on mal expansions; properties of importance in well-log
of the factors that has a significant impact on the interpretation, for example,formation electrical resis-
change in the mechanical properties of rock. Numer- tivity factor and elastic-wave velocities; mechanical
ous laboratory measurements, mostly are sedimentary properties such as strength and deformation moduli.
rocks, have been performed to determine the tem- In addition, knowledge of the mechanical proper-
perature dependence mechanical properties of rocks ties or rocks in high-temperature environments is
(R. and Homand 1979, Zhang et al. 2009, Hong useful for application of properties of geothermal
et al. 2012, Tian et al. 2016). All these work tested reservoir characterization (Somerton 1992). There
a number of rocks deformation modulus, Poissons are number of important parameters that influence
ratio, the tensile strength, compressive strength, cohe- the mechanical properties of the rock mass like
sion and internal friction angle, viscosity, and other grain size and shape, density, porosity, anisotropy,
parameters and discussed the dependence of lateral pore water, confining pressure, temperature, weath-
pressure and high temperature on thermal expan- ering and alteration zones, bedding planes, and joint
sions as well as the creep characteristics of the rock. properties.
311
The ultrasonic wave velocities is highly efficient, the plane of magmatic layering (fluid flow), X is paral-
handy and reflects the mechanical properties of the lel to a shape preferred crystal orientation, if present,
tested object, which is widely used in rock sample and Z is normal to the layering.
testing. Experiments focussing on the temperature Densities were calculated from mass and dimen-
dependence of ultrasonic wave velocities in meta- sions of the cubic specimen, the porosities being
morphic rocks are rare. Laboratory measurement of derived by mass/volume analysis. Two representa-
temperature-dependent ultrasonic wave velocities of tive samples selected Puntone-4B (3289.6–3292.6 m
rocks at high temperatures have been performed by depth core), of amphibolite: composed by hornblende,
(Kern 1978, Punturo et al. 2005, Scheu et al. 2006). biotite, quartz, plagioclase (opaques, apatite) (Fig-
However, there experiments were not discussed the ure 1) and Carboli-11A (3594–3596 m depth core),
effect of temperature on the mechanical properties of schist with quartz prevailing on muscovite: an opaque
rocks. Many researchers have found that seismic wave granulation is pervasively dispersed in the rock also
velocities (compressional P (Vp ) and shear wave veloc- enhancing the schistosity. The bulk densities of these
ities S (Vs )) is closely related to mechanical properties cylindrical samples are in good agreement with the
of rocks (Lama and Vutukuri 1978, Gaviglio 1989, densities of the cubic samples. The determined den-
Yasar and Erdogan 2004). sity values range of 2.75–2.91 g/cm3 corresponding
In the present work Vp and Vs wave velocities of two to a open porosity range of 28.15–36.96 vol.%. This
core sample taken from area of the active geothermal covers the density range from the most abundant to
field of Larderellol, Italy have been measured at ele- the maximum density observed for the block-and-ash
vated temperature and pressure. Results are reported flow deposits.
for porosity, density, deformation, deformation mod-
uli, permeability have been calculated in situ condition
from Vp and Vs elastic wave velocities. Further, an 3 METHODOLOGY
attempt has been made in the present study is to
correlate porosity, permeability, deformation moduli, The elastic VP - and VS -wave velocities and veloc-
density with Vp and Vs wave velocities. ity anisotropies of the samples were determined
experimentally. The measurements were conducted
on cube-shaped specimens in a true triaxial multi-
2 SAMPLE DESCRIPTION anvil press using the ultrasonic pulse-transmission
technique (Figures 2 and 3). A state of nearly hydro-
Two samples from Larderellol, Italy, were investigated. static stress was achieved by pressing six pyramidal
They were collected from the Carboli-11A and Pun- pistons onto the sample cubes. The special arrange-
tone 4B core. The clasts for this analyses were chosen ment of the sample-piston-transducer assembly allows
to be as representative as possible, in terms of den- simultaneous measurements of compressional (P) and
sity and homogeneity/heterogeneity of the Larderellol orthogonally polarized shear wave velocities (S1, S2).
core. For the experiments cube-shaped specimens with The end of each piston next to the specimen is sur-
43 mm edge lengths were cut from these clasts.The ori- rounded by a furnace and heat is transmitted from the
entation of the cube axes followed visible mesoscopic pistons to the specimen, allowing homogenous heating
fabric coordinates where possible: X and Y are within and temperature distribution within the large volume
specimens, as has been confirmed by temperature
measurements at different places within a test sam-
ple (Kern et al. 1997, Punturo et al. 2005, Scheu et al.
2006). Temperature is measured using thermocouples
Figure 1. Thin section: Puntone-4B, amphibolite: com-

posed by hornblende, biotite, quartz, plagioclase (opaques, Figure 2. Schematic draw of the triaxial multi-anvil press
apatite). Hbl-prevailing levels, 3–4 mm thick, regularly alter- used to measure the elastic P- and S-wave velocities. The
nate to Bt+Qtz-prevailing levels of similar thickness. Grain measurements were performed up to 100 MPa pressure and
size is small, planar schistosity. 600◦ C temperature (Scheu et al. 2006).
312
placed in a cavity at the end of each piston very close to 100 MPa, followed by measurements at the con-
(∼1 mm) to the specimen. The transducers (2 MHz and stant confining effective pressure of 100 MPa over
1 MHz) are placed on the low-temperature side of each the temperature range 20–600 ◦ C. The intrinsic effect
piston, and the travel time of the pulses through the of temperature on velocities is hard to determine,
specimen is obtained by subtracting the calibrated time due to thermal expansion and the consequent loos-
needed for the pulse to travel to and from the specimen ening of structure. Figure 4a shows the change of
through the pistons from the total time measured by the the P-wave velocities and Vp anisotropy of the sam-
transducers. Length and resulting volume changes of ple Carboli-11A as a function of temperature at
the sample cubes, due to changes of principal stress and 100 MPa. In addition, P-wave velocities measured in
temperature, are obtained by the piston displacement. three direction decrease with increasing temperature
The cumulative error in VP - and VS -wave is estimated from 6.32–5.62 km/s at room temperature to 5.03–
to be around 1% (Kern et al. 1997, Punturo et al. 2005, 5.88 km/s at 600◦ C, it was similar to (Kern et al.
Scheu et al. 2006, Wenk et al. 2008). 1997, Punturo et al. 2005, Scheu et al. 2006). Fur-
thermore, The velocity anisotropy (A-Vp%) shows an
inverse behavior; it increases with higher tempera-
4 RESULTS ture from 11.63% at room temperature to 15.56% at
600◦ C. In general, the P-wave velocity decrease and
VP - and VS -wave velocities were measured simulta- velocity anisotropy increase at increasing the temper-
neously in the three structural directions. Measure- ature for both samples. The mean P-wave velocities of
ments were performed first at room temperature up both samples increase from 3,87–4.60 km/s at 20◦ C to
4.21–5.35 km/s at ◦ C. A similar behavior at increasing
temperature was observed on the other samples. The S-
wave velocities show a similar trend. Figures 4b shows
the Vs velocities of the sample Carboli-11A as a func-
tion of temperature. All three directional Vs velocities
decreased from 3.80–3.60 km/s at room temperature
to 3.63–3.49 km/s at 600 ◦ C, it similar to (Kern et al.
1997, Punturo et al. 2005, Scheu et al. 2006).
Figure 5 shows average curve for theYoung’s E, bulk
K and shear µ, moduli of the rock sample Carboli-
11A as a function of temperature from 20 to 600◦ C.
Deviations from the moduli occur with changes in the
Poisson’s ratio. This is because of the fundamental
linkage between temperature and because of the elas-
Figure 3. True triaxial multianvil pressure apparatus at Kiel tic equations depends on VP - and VS with the elastic,
University to investigate mechanical and seismic rock proper-
bulk, and shear moduli. Concerning moduli parameter,
ties at elevated stress and temperature conditions. a) Triaxial
(multi-anvil) pressure apparatus for direct measurements of P high Poisson’s ratios give values below the Figure 5,
and S-wave travel times. b) Reference system of a clay sam- and low Poisson’s ratios give higher values of moduli.
ple with regard to the polarisation directions of the various Additionally, it is found that the moduli of metamor-
shear waves (see e.g. Kern et al. 1997, Wenk et al. 2008). phic tested rock samples decreases with increasing
Figure 4. Elastic wave velocities as a function of temperature for the sample Carboli-11A. The measurements were carried out
at 100 MPa pressure. (a) The compressional wave velocities (Vp) decrease with increasing temperature, whereas the anisotropy
(A) increase. (b) The averaged shear wave velocities (Vs) decrease with increasing temperature.
313
temperature from 20 to 600◦ C. It is apparent that tem- permeability of Carboli-11A is increase at the increas-
perature has a significant effect in decreasing these ing temperature. In contrast, the mean permeability
moduli properties. Furthermore, the mean bulk den- of sample Puntone-4B is approximately constant at
sity, volumetric strain of both samples increase with increasing the temperature (Figure 6). It is found that,
the increasing the temperature. The mean porosity and an average 10% increase in pore volume compress-
ibility of Carboli-11A sample, upon increasing the
temperature from room to 600◦ C, which is very close
to to Puntone-4B sample observed change in bulk vol-
ume compressibility for the same metamorphic rock
over this temperature range. Thus, if porosity did not
change significantly over this temperature range. For
the determination of density, volumetric strain, poros-
ity and permeability, used th correlations function of
these parameters wit the VP - and VS -wave veloci-
ties, which were measured in the laboratory with the
dependence of temperature. The correlation function
of porosity and permeability choose from literature of
similar type of rock and minerals.
In order to investigate the deformation behaviour
here also evaluated the axail strain in the three direc-
tion and calculated the volumetric strain (see Figure 7).
Figure 5. Average curves of Carboli-11A, for the field-scale However, comparing the lateral strain parallel to the
Young’s E, bulk K, and shear µ moduli, in relation to bedding with the axail strain it is obvious that the
temperature. expansion behaviour of the sample is dominated by
Figure 6. Average curves for the bulk density, volumetric strain, porosity and permeability, in relation to temperature.
314
elastic, bulk and shear moduli, behavior with respect
to temperature changes. The mechanical properties of
rocks calculation based on the analysis and using elas-
tic wave data, is had measured at high temperature.
In general, elastic, bulk shear moduli of the metamor-
phic rock samples shows decrease with the increasing
temperature. It is concluded that moduli of reservoir
rocks decreased substantially when pore-fluid pressure
is decrease in high temperature reservoir. However, all
the laboratory test be made at reservoir in-situ temper-
ature for the propose of determine physical properties
of rocks.
The presented data set of density, volumetric strain
porosity and permeability show that there is a clear
trend of increasing porosity with increased for the
selected area (Figure 6). The dependence of perme-
ability on porosity, density and volumetric strain is
Figure 7. Temperature-strain curves for Carboli-11A sam-
generally explained by the assumption that a more con-
ples in axail and lateral direction. nected pore space of rock matrix (pores and cracks)
provides more efficient pathways for the fluid move-
the axail expansion in direction perpendicular to the ment. The samples Carboli-11A show higher than
bedding plane resulting in an overall expansion. average permeability for the these porosity and den-
sity, which supports the conclusion of that possibilities
of the geothermal fluid migration.
The ratio of axial stress and resulting axial strain
5 DISCUSSION
gives a deformation modulus characterizing the defor-
mation behaviour. Samples show an axial and a lateral
In this paper, the effect of temperature of rocks on their
deformation; relative change of dimension divided
physical properties were discussed in details. These
by relative change of axial length in stress direc-
are the irreversible changes that occur upon heating
tion is Poissons ratio; it represents a second property
rocks to their room temperature. The effects of tem-
characterizing the deformation behaviour.
perature are considered to be the reversible changes
The relationship between the Young’s, bulk and
that occur at increasing temperatures. In metamor-
shear moduli as well as density, porosity, volumetric
phic rocks, increasing temperature generally leads to
strain a nd permeability with elastic P- and S- wave
decreasing velocities. The main causes of the velocity
velocities are valuable for understanding a geothermal
change are:
reservoir evaluation. The data indicates strong corre-
• temperature dependence of the elastic properties lation between these parameters. This paper consider
of the rock-forming minerals and phase change of that presented dataset can help reduce the technical and
minerals; financial risk of drilling and exploiting deep geother-
• temperature dependence of the elastic properties mal drilling by in improving knowledge of the physical
of the pore constituents and change of pore con- parameter in geothermal reservoir characterisation.
stituents (e.g., pore water) from a liquid to a gaseous Therefore, a deeper understanding of how mechani-
state; cal parameters may be influence the migration fo fluid
• changes in the contact conditions at grains, crack and change in mechanical parameter with respect to
boundaries, etc., resulting from variations of the pressure and temperature is important. Knowledge of
interface effects and/or from different thermal the mechanical properties at temperature of rock has
expansion properties of rock-forming minerals. become insincerely important with the wide-spread
interest in physical process in underground geother-
The decrease in elastic wave velocities with
mal reservoir. For meaningful analyses of mechanical
increased temperature could result in predicting higher
properties, in addition to rock properties and their
porosities in using the time-average relationship. This
behaviour in high-temperature environment must be
is contradictory to some evidence that porosity tends
known. Some of these mechanical properties requir-
to decrease with higher temperatures.
ing knowledge of high-temperature behaviour of rock
The velocity anisotropy can be linked to the texture
system include underground storage heat, disposal
of the samples. Those with a high anisotropy show
nuclear wast and geothermal reservoir evaluation.
a pronounced shape preferred orientation and micro-
crystal sometimes in addition to layering within the
groundmass of the sample.
The effect of temperature on elastic and compress- 6 CONCLUSIONS
ibility is substantially higher on the reservoir meta-
morphic rocks. Differences in mineral composition Based on the experimental results, it is apparent
and lower porosity probably explain the differences in that the interpretation techniques for the geothermal
315
reservoir must take into account the effect of tem- Hong, T., T. Kempka, X. Neng-Xiong, & M. Ziegler (2012).
perature on rock properties. The properties affected Physical properties of sandstones after high temperature
by temperature to be discussed include bulk and pore treatment. Rock Mech. Rock Eng. 45, 1113–1117.
compressibility, P- and S-wave velocities. An example Kern, H. (1978). The effect of high temperature and high
confining pressure on compressional wave velocities in
of the potential application of these mechanical prop- quartz-bearing and quartz-free igneous and metamorphic
erties would be the understanding of the response of a rocks. Tectonophysics. 44, 185–203.
in-situ temperature effect at depth. At the 3000-4000 Kern, H., B. Liu, & T. Popp (1997). Relationship between
m depth samples from the potential geothermal reser- anisotropy of p and s wave velocities and anisotropy of
voir of Larderellol, Italy. The aim of this study is to attenuation in serpentinite and amphibolite. Journal of
evaluate the mechanical behavior of rock based on the Geophysical Research-Solid Earth. 102, 3051–3065.
ultrasonic wave velocities as a function of tempera- Lama, R. & V. Vutukuri (1978). Handbook on Mechanical
ture to serve as host materials to geothermal reservoir Properties of Rocks. Clausthal: Trans. Tech. Publications.
investigation. Punturo, R., H. Kern, R. Cirrincione, P. Mazzoleni, &
A. Pezzino (2005). P- and s-wave velocities and densities
In order to determine mechanical properties of in silicate and calcite rocks from the peloritani moun-
the reservoir rocks tested two core sections from the tains, sicily (italy): The effect of pressure, temperature and
well at a safe depth, as well as using existing cores the direction of wave propagation. Tectonophysics. 409,
from metamorphic basement. This offered the oppor- 55–72.
tunity to perform petrophysical laboratory analyses R., H. & E. Homand (1979). Influence of temperature on the
on core samples so as to define important mech- mechanical behavior of rocks. Proc. 4th Int. Cong. Rock
nical characteristics of the reservoir host rocks, in Mech., Montreux. 110, 115–122.
particular seismic P- and S-wave velocities, phys- Scheu, B., H. Kern, S. O., & D. Dingwell (2006). Temperature
ical and geomechanical properties. Rock samples dependence of elastic p- and s-wave velocities in porous
mt. unzen dacite. Journal of Volcanology and Geothermal
were carried under simulated in-situ conditions at the Research. 153, 136–147.
high-pressure, high-temperature by in a cubic multi- Somerton, W. (1992). Thermal properties and temperature-
anvil pressure apparatus, using the pulse transmission related behavior of rock/fluid systems.Amsterdam: ELSE-
technique. VIER.
Based on the results of this study, further inves- Tian, H., T. Kempa, S. Yu, & M. Ziegler (2016). Mechani-
tigation on laboratory measurement of permeability cal properties of sandstones exposed to high temperature.
and porosity under in-situ temperature and pressure Rock Mech Rock Eng. 49, 321–327.
is necessary. The effect of the temperature on elastic Wenk, H., M. Voltolini, H. Kern, T. Popp, & M. Mazurek
wave velocities is one of the major issues to be inves- (2008). Anisotropy in shale from mont terri. Leading
Edge. 1, 742–748.
tigated. Furthermore, the influence of temperature on Yasar, E. & Y. Erdogan (2004). Correlating sound velocity
the mechanical and chemical properties of rocks need with density, compressive strength and youngs modulus
to be further analysed and will be the topic of future of carbonate rocks. Int. J. of Rock Mechanics and Mining
work. Sciences. 41, 871–875.
Zhang, L., X. Mao, & A. Lu (2009). Experimental study on
the mechanical properties of rocks at high temperature.
REFERENCES Science in China, Series E: Tech. Sci. 1, 979–985.
Gaviglio, P. (1989). Longitudinal wave propagation in a lime-

stone: the relationship between velocity and density. Rock
Mech Rock Eng. 22, 299–306.
316
Inverse-volume pore pressure projection technique for the coupling

of reservoir flow and geomechanics simulators
P.E. Vargas & N.A. González

Repsol Technology Centre, Móstoles, Madrid, Spain
C.N. Mena Paz, M. de Bayser & P.W. Bryant

IBM-Research, Río de Janeiro, Brazil
J.M. Segura & M.R. Lakshmikantha

Repsol Technology Hub, The Woodlands, Texas, USA
ABSTRACT: Coupling geomechanical reservoir models with fluid flow models improves accuracy when
simulating phenomena relevant to production forecast, drilling and well integrity, and environmental impact
(e.g. fault reactivation). A geomechanical model provides to a hydrodynamical model updated values of pore
volume and permeability, and a hydrodynamical model in turn provides to a geomechanical model updates to
the pore pressure, which then alters effective stress. The main challenge in coupling is the exchange of data.
This work presents three different pore pressure projection schemes used for data exchange, that are based on
weighted averages (inverse distance, volume, and by inverse of the volume). To validate the three methods, the
pressure projected was compared with a hydrostatic pressure distribution while accounting for changes in the
size of the surrounding cells of the hydrodynamical model. The weighting by the inverse of the volume shows
the better agreement with hydrostatic pore pressure distribution of the three techniques presented. As a further
demonstration, the three methods were compared within a small reservoir model with varying cell sizes, and the
inverse volume method performs favorably. Finally an extension to the inverse volume method was developed,
in which permeability is included in the weighting factors. Its effect is demonstrated in a benchmark system
including both a reservoir and a nonpay region.
1 INTRODUCTION using iterative algorithms that require the resolution of

fluid flow within a ‘reservoir simulation’ and also of a
Diverse scenarios require incorporating geomechanics geomechanical ‘stress simulation’.
effects into petroleum engineering analyses, leading The iterative algorithm is based on data exchange
to coupled geomechanics-reservoir modeling of the between the two simulators (see Fig 1). First the reser-
stress and pressure changes within one framework. voir simulator provides the pressure changes to the
The main challenge in coupling is implementation of stress simulator. The pressure change is transformed
data exchange (stresses & strains, and pore pressure into loads in the stress simulator which computes in
data for both matrix and fractures) between the geome- turn the resulting stress and strain changes. Stability
chanics and the flow simulator. Coupling of fluid flow of coupled formulations is extensively discussed in the
and geomechanics models provides an accurate rep- literature by Hughes et al. (1986) and in the specific
resentation both of the dynamic model based on a context of geomechanics by Haga (2011), Kim et al.
geomechanics-guided update of pore volume and per- (2011), Murad et al. (1994) and Wan et al. (2003).
meability, and of the geomechanical analysis based on A scheme is required to transfer the pore pressure
an effective stress update. (from the flow simulator), calculated at the center of
The geomechanical and the flow systems are cou- the cell, to the nodes (of the geomechanical simulator)
pled via the pore pressure term in the mechanical of the finite element mesh. This transfer is known as
equilibrium equations and the time derivative of the Pore Pressure Projection.
volumetric strain in the flow equations, and also by the This work presents three different methods for pore
possible dependence of the material parameters like pressure projection: inverse distance, volume average
permeability and/or porosity with the stress state in the and inverse volume method. These names describe
porous medium. The coupled problem can be solved the weighting factor used to project pressure values
319
Figure 1. Data exchange between flow simulator and
geomechanical simulator for coupled models.
from cell centers to the nodes. For the inverse distance

method the weighting factor is the distance between
block-centered data and the nodal data. For the volume
average and the inverse volume the weighting factor is
determined by the sub-volumes that surround a node.
The inverse volume method divides the finite element Figure 2. Pressure projection scheme: Relation between the
into sub-volumes according to the number of nodes of finite element mesh and the finite volume grid. Cell centered
the element, associates each node to its correspond- values indicated with the gray circles. Nodal values indicated
ing sub-volume and pore pressure values, and finally with black circles.
computes the pore pressure by clustering around each The discrete space Wh is made up of continuous
element node weighted by the inverse volume. linear functions on τh , such that the pressure field is
given by
2 TECHNIQUES TO PROJECT PORE

PRESSURE
∼
Fluid flow and geomechanical simulations represent From Equation 1 and the expression of ph in terms
the domain as a grid of discrete cells. However, the of the nodal basis functions N j , a linear system of equa-
Finite Volume Method (FVM) employed by fluid sim- tions in the nodal pressure coefficients and involving
ulators calculates properties such as pressure at the a mass matrix, is obtained, i.e.,
centers of cells whereas the Finite Element Method
(FEM) used by geomechanical simulators calculates
forces and displacements at the corners (also called
nodes). Therefore even if both grids coincide spatially where each component of the mass matrix and the right
the pressures that the FVM calculates at cell centers hand side read
must be interpolated, or projected, to cell corners. In
a hexahedral grid each corner is shared by up to eight
cells and the pressure at this point is related to the
pressures of these cells.
Figure 2 shows the relation between the finite ele-
ment mesh in which the mechanical problem is solved Possible variations that avoid the cost of solving
and the finite volume mesh in which the flow problem a full linear system are to use a technique known as
is solved. For simplicity we show the two dimen- lumping of the mass matrix M (Bathe et al. 1976), in
sional (2D) case considering quadrilateral elements. which the mass matrix is approximated by a diagonal
The dashed arrows indicate where the cell-centered matrix, or to use more simplified weighted averaging
value would be included in the computation of the techniques, such as volume average, inverse volume
nodal values when a projection scheme is applied. and inverse distance.
Some schemes of pore pressure projection require The projection of pore pressure from cell centers to
the resolution of a linear system. An example is the the nodes with a weighted technique is calculated for
L2 () projection, which is a natural option if a nodal known values (fi ) as:
value for the pressure is needed for accuracy or stabil-
ity reasons. In this case the problem to be solved reads:
∼
Given ph ∈ P0 (τh ), find ph ∈ Wh such that ∀qh ∈ Wh
where n = number of points with known values;

fi = known value at point i; and wi = weight assigned
to each point and defined in Sections 2.1, 2.2 and 2.3.
320
Figure 4. Subdivision of cell with known center-cell pres-
sure Pi in sub-volumes for Vertexk .
Figure 3. Distance between point P and location of known
values (P1, P2, P3 and P4). The weighting factor wi (see Eq 5) is defined as
In the following subsections are described the

weighting factor for inverse distance, volume weighted
average and inverse of volume techniques.
where V = associated sub-volume to the known value

2.1 Inverse distance Pi (see Fig 4).
In the inverse distance technique the pore pressure at
vertex P (see Fig 3) is calculated using a weighted
average of the available values defined at the centers
of the cells (P1, P2, P3 and P4). 2.3 Proposed Inverse volume weighted average
Shepard (1968) and Lam et al. (1983) define the The inverse volume technique uses the same definition
weighting factor wi (see Eq 5) as of the sub-volume as in Section 2.2 (see Fig 4), and
the weighting factor wi (see Eq 5) is defined as:
where d = distance between interpolation point and

location of known value (see Fig 3). Explicitly, the
distance in three dimensions is
3 VALIDATION OF TECHNIQUES
3.1 Projected value, a matter of space location

Figure 5 shows two different meshes with the same
cell-center points P1, P2, P3 and P4. Pore pressure val-
where (x, y, z) = interpolation point coordinates and
ues at these points are known from theoretical variation
(xi , yi , zi ) = data point coordinate.
or from a flow simulator.
Figure 5 also exhibits points Pa and Pb with
unknown values.These points correspond to the upper-
right vertex of cells with center located at P1.
2.2 Volume weighted average Depending on the mesh, the location of the upper-
In the volume weighted average technique, the pore right vertex of the shaded element (Pa or Pb) may vary,
pressure at each vertex is computed using sub-volumes and therefore the pressure projected to this point (Pa or
(see Fig 4) for each concurrent element to the vertex Pb) through the projection of the known cell-centered
in the model. pressures may vary as well.
321
Figure 5. Four center cell points P1, P2, P3 and P4 with
known pressure in two different meshes. Hatched/shaded area
corresponds to a cell with known pressure at center (P1) but
unknown pressure at respective vertex Pa or Pb.
Figure 7. Theoretical values for hydrostatic pore pressure

distribution in normalized space (rx , rz ) and pore pressure
projected values based on theoretical values at corners, com-
puted with three different techniques (inversed distance,
inverse volume and volume).
(see Eq 10). Also shown are the theoretical values

Figure 6. Definition of normalized coordinates ri based on for the hydrostatic pore pressure distribution in the
center-cell location of cells (P1, P2, P3 and P4) and loca- normalized space (rx , rz ).
tion of upper-right vertex of shaded cell (P) in 2D system of
The pore pressure distribution computed with the
coordinates.
inverse volume technique show good agreement with
theoretical values. The pore pressure distribution com-
Normalized coordinates (rx , ry , rz ) are used in the puted with inverse distance technique show agreement
following sections to localize a projected value at given in the middle, close to rz = 0.5. The pore pressure dis-
point P (see Fig 6). tribution for the direct volume technique also shows
good agreement only in the rz = 0.5 region.
When P is centered vertically between line P1-P2 3.3 Error analysis of projected values
and line P4-P3 (ry = 0.5) and centered horizontally The values computed in section 3.2 were compared
between line P1-P4 and line P2-P3 (rx = 0.5), the with hydrostatic pore pressure in order to evaluate
projected value will be at the coordinate (rx = 0.5, the range of usability of each technique described in
ry = 0.5) inside the normalized space. section 2.
The normalized axis ri is 0 when the length in i-axis Figure 8, shows the error of the projection tech-
(li ) is very small with respect to a neighbor cell length, niques (see Fig 7) when compared with the hydrostatic
and 0.5 when the length in i-axis is the same as that of pore pressure distribution. The normalized axes rx and
the neighbor cell (Li = 2li ). rz correspond to Equation 10.
Figure 8 exhibits very good agreement (error less
than 5%) between the pore pressure distribution result-
3.2 Projected values with different techniques
ing from the inverse volume projection technique and
The techniques defined in Sections 2.1, 2.2 and 2.3 hydrostatic pore pressure distribution for all rz values.
were applied in order to obtain projected values based The inverse distance projection technique and the vol-
on eight cell-centered pore pressure known values. The ume projection technique show good agreement only
known pore pressure values follow from a hydrostatic for rz values close to 0.5. In other words, when the
pore pressure distribution. heights of two neighboring cells are very similar the
Values computed with each technique are shown results are in good agreement with a hydrostatic dis-
in Figure 7 using normalized coordinates (rx , rz ) tribution. When the heights are not similar, the error
322
Figure 9. Reservoir geometry and position of vertical line
of cells.
Figure 8. Error computed in normalized space (rx , rz ) for

projected values using three different techniques (inverse
volume, inverse distance and volume) with respect to the
theoretical hydrostatic pore pressure distribution.
in the projected value may reach between 5% and 40%

of a hydrostatic pore pressure distribution. Figure 10. Pressure at nodes and cells of vertical line.
4 SYNTHETIC EXAMPLE
4.2 Comparison of results
Pore pressure projection techniques defined in Section 4.2.1 Vertical pressure distribution
2 are applied to a small reservoir model that has diverse Figure 9 also shows the vertical column of cells (in red)
cell sizes to illustrate the geometrical weighting. where the pore pressure is extracted for a comparative
analysis.
Figure 10 shows pore pressure projected values at
4.1 Description of the model vertices and the known values at the centers of the
The reservoir contains a single vertical well (wellbore surrounding columns of cells. Projected values were
radius 0.5 ft) in the center, completed in all reservoir computed with the techniques described in Section 2.
layers with production of 50,000 STB/D for 4,000 Figure 10 shows that the inverse volume technique
days. The 20 first and the last 18 time steps have a has the smoothest curve in between the curves of
duration of 20 and 200 days each, respectively. the cell centers, while the other two techniques show
The grid is 11 cells long in the X and Y directions considerable variation. A divergence is observed at
and 5 cells high, as shown in Figure 9. In the X and Y boundary cells due the lack of information to project
directions the cells close to the well are smaller giving values at the boundary of the model. Projected values
a higher resolution at a point where the simulation with the inverse of distance or the direct volume tech-
results in a more dynamic pressure state. In Z the cells niques suffer a shift because of the irregularity of the
are also smaller in the middle layers. mesh in z-axis.
The three different pressure projection procedures
were tested on the 20th simulation step where there 4.2.2 Horizontal pressure distribution
are already marked differences in the pressure distri- Figure 11 shows the horizontal row of cells (in
bution. The results are evaluated along a vertical and red) where the known pore pressure is extracted in
a horizontal line of cells in the reservoirs. order to compare with the projected values using the
The different heights in the model (see Fig 9) allows techniques defined in Section 2.
to capture the variation of results for each pore pressure Figure 12 shows pore pressure projected values at
projection technique applied. vertices and the known values at the centers of cells.
323
Figure 13. Variation of projected pore pressure in Vertex
Figure 11. Location of horizontal line of cells. k, based on known cell-centered values Pi and Pj and their
corresponding permeability (κi and κj).
5.1 Proposed petrophysical weighting method

Equation 11 shows the proposed petrophysical weight-
ing scheme.
where pj is the pressure at node j. Each node j has

associate a set of adjacent elements τj . Each adjacent
element k in τj , has its own pressure Pk at center of
Figure 12. Pressure at nodes and cells of vertical line. the cell, sub-volume Vk (see Fig 4), and permeability
norm ||κk || computed as
Projected values were computed with the techniques

described in Section 2.
Weighting by permeability introduces petrophysical
Figure 12 exhibits that the inverse volume tech-
parameters into the otherwise geometrical scheme.
nique results in the best agreement with values at
By choosing weights proportional to permeability,
vertices regarding cell-centered known values. The
pressure cells within regions of high permeability
inverse distance projection method and direct volume
are more significant than those within regions of
projection method give values different from those at
low permeability. In regions where the permeability
the midpoints.
field is homogeneous or nearly homogeneous, projec-
tion results are identical or nearly identical to purely
geometrical methods (see Fig 13), respectively.
5 INFLUENCE OF PETROPHYSICAL
PROPERTIES IN PORE PRESSURE 5.2 Application of proposed method
PROJECTION
Pore pressure projection techniques defined in Sec-
The numerical schemes used to solve the hydrody- tion 2 together with the petrophysical weighting
namical equations tend to be relatively less accurate scheme (Eq 11) are applied to a system comprised
in regions of low permeability than in regions of high of a reservoir embedded within a nonpay region, and
permeability. Also it is important to take into account the projected values computed with these techniques
that different values of permeability between neighbor are analysed.
cells (see Fig 13) could generate variation in projected
pressure as well. 5.2.1 Description of the model
Because of these shortcomings, a heuristic adjust- The reservoir contains a single vertical well (wellbore
ment to the cell-to-node pressure projection method is radius 0.25 ft) in the center, completed in all reser-
proposed (Repsol/IBM, 2015), in which the norm of voir layers with production of 50,000 STB/D for 4,000
the permeability field is used as a weighting factor in days. The grid geometry for the model is shown in
the computation of nodal pressure. Figure 14. The reservoir is highlighted in red.
324
Figure 16. Comparison of Methods for Pore Pressure Pro-
jection: Subsidence at surface of the model; Dual Porosity
And Multilaminate Model.
Figure 14. Grid geometry of the synthetic example (Dean

et al. 2006).
Figure 17. Comparison of Methods for Pore Pressure Pro-

jection: Subsidence at top of the reservoir; Dual Porosity And
Multilaminate Model.
In the geomechanical model, the material of non-

Figure 15. Comparison of Methods for Pore Pressure Pro- production areas is stiff and not fractured. Rock
jection: Average Reservoir Pressure; Dual Porosity And mechanics properties are as follows: Poisson ratio is
Multilaminate Model. 0.25, and the initial in-situ density of the solid material
is 2.7 g/cm3 . Within the reservoir the elastic modulus
The grid geometry is equivalent to one published is 1e4 psi and outside the reservoir it is 1e6 psi, and
by Dean et al. (2006) excepting for the dual porosity the compressibility inside the reservoir is 1.5e4 psi−1 .
model used in this work. As a consequence, to treat Each cell has exactly one horizontal fracture family
both the fractures and the matrix in the hydrodynamic consisting of one fracture with an initial aperture equal
model the number of grid-blocks in the Z direction is to 0.0217 inches which results in an permeability of
doubled from 12 in the original grid to 24. The well is 0 mD in Z and 100 mD in X and Y according to
also as published in Dean et al. (2006) but is connected Poiseuille flow assumption (Snow, 1965), shown in
only to the fracture cells. equation (13). The geomechanical mesh coincides spa-
Horizontal and vertical permeabilities of the frac- tially with the fluid simulation grid. The permeability
tures within the reservoir are 100 mD and 0 mD, of a fracture family is related to the permeability of a
respectively. Horizontal and vertical matrix permeabil- single fracture by
ity values are 0.001 mD and 0.0001 mD, respectively.
The dual porosity matrix-fracture coupling factor, σ,
is 0.0076 ft-2 (Kazemy et al. 1976 & Schlumberger
2016). Outside the reservoir the permeability is zero.
This benchmark case was run coupled with a where a is the aperture size, s is the fracture spacing,
fractured geomechanical, multi-laminate constitutive and f is the fracture frequency.
model. The reason for this approach is that in frac-
tured models the permeability has a great impact on 5.2.2 Results
fluid flow and, in turn, pressure changes greatly influ- In figures 15, 16 and 17 we can see the average pressure
ence the closure of fractures, which translates directly and the measured subsidence at the surface and at the
to a reduction in permeability in the fluid model. top of the reservoir, with respect to simulation time,
325
respectively. From the results we can see that although Hughes T.J.R., Franca L., Balestra and M. A new finite ele-
the three geometrical weighting methods resulted in ment formulation for computational fluid dynamics: V.
different pressure profiles locally, at the global scale Circumventing the Babuka Brezzi condition. A stable
they behave similarly. However, to assess the impact Petrov Galerkin formulation of the Stokes problem accom-
modating equal-order interpolations. Comput. Methods
of permeability weighting, we also ran the test with Appl. Mech. Engrg., 59:85–99, 1986.
the inverse volume technique but with permeability Kazemi, H., Merrill Jr., L.S., Porterfield, K.L., and P.R.
weighting excluded. We can see that this resulted in Zeman. 1976. Numerical simulation of water-oil flow in
different pressure and subsidence curves. naturally fractured reservoirs, SPE-AIME Fourth Sympo-
sium on Numerical Simulation of Reservoir Performance,
Los Angeles, CA, USA, SPE-5719-PA.
6 CONCLUDING REMARKS Kim J., Tchelepi H.A., and Juanes R. Stability and con-
vergence of sequential methods for coupled flow and
Pore pressure projection is a fundamental task for geomechanics: Fixed-stress and fixed-strain splits. Com-
puter Methods in Applied Mechanics and Engineering,
petroleum engineering analysis when passing cou- 200:1591–1606, 2011. 20
pled information between the cell-centered values of Lam, N.S-N. 1983. Spatial Interpolation Methods: A Review.
flow simulators and the nodal values of geomechanical Cartography and Geographic Information Science 10 (2):
simulators. 129–150.
Three pore pressure projection schemes based on Murad M.A. and Loula A.F.D. On stability and convergence
weighted averages were described in this work, inverse of finite element approximations of Biots consolida-
distance, volume, and inverse of the volume. tion problem. Comput. Methods Appl. Mech. Engrg., 37:
Validation of these schemes showed that inverse vol- 645–667, 1994.
ume weighted average has the best agreement when Wan J., Durlofsky L.J., Hughes T.J.R., and Aziz K. Stabilized
finite element methods for coupled geomechanics reser-
a hydrostatic pressure distribution is applied while voir flow simulations. SPE Res Simul Sym (SPE 79694),
accounting for changes in the size of the surrounding Houston, 3–5 Feb., pages 85–99, 2003.
cells of the hydrodynamical model. Inverse distance Repsol/IBM, 2015. Method of managing petro-chemical
and volume projection schemes shown good agree- reservoir production and program product there-
ment only when the size of surrounding cells are very fore. Patent filed in Europe and US. IBM Ref.:
similar, when the heights in neighbor cells are not YOR920140126US1. REPSOL Ref.: PA00179
similar, the error in the projected value may reach Schlumberger. 2016. Eclipse Black Oil Simulator Reference
between 5% and 40% of a hydrostatic pore pressure Manual.
distribution. Shepard, Donald (1968). “A two-dimensional interpolation
function for irregularly-spaced data”. Proceedings of the
Evaluation of three pressure projection schemes 1968 ACM National Conference. pp. 517–524.
accounting for both geometrical factors and petro- Snow, D. 1965. A parallel plate model of fractured permeable
physical factors were performed. Testing on various media. Ph.D. thesis, University of California, Berkeley,
systems indicates that the inverse volume weight- CA, USA
ing produces superior results and that weighting by
permeability is indeed significant.
REFERENCES
Bathe, K. J., & Wilson, E. L. 1976. Numerical methods in
finite element analysis.
Haga C.J.B. Numerical methods for basin-scale poroelastic
modelling. Ph.D. Thesis, Univ. of Oslo, 2011.
326
Axisymmetric transient modelling of a suction caisson in sand:

A numerical study
B. Cerfontaine, F. Collin & R. Charlier

Geomechanics and Engineering Geology, University of Liege, Belgium
ABSTRACT: This paper deals with the axisymmetric behaviour of a suction caisson installed in sand upon
vertical monotonic and cyclic loading. A steel caisson is numerically modelled using the finite element code
LAGAMINE. The Prevost model reproduces the cyclic behaviour of the soil, i.e. it captures the accumulation of
deformation and pore water pressure within the soil. Coupled interface finite elements allow the modelling of
the uplift behaviour of the caisson in both drained and partially drained conditions. Upon compression or traction
loading, the suction caisson presents different modes of resistance: friction along the shaft, bearing capacity under
the lid or the tip of the caisson, suction effect. The first part of this work describes the progressive mobilisation of
these modes of resistance during monotonic simulations. The partially drained effect is particularly interesting
since it drastically increases the resistance to transient loading. It proceeds from the transient consolidation
process induced by the caissons loading. The second part describes the evolution of the settlement of the suction
caisson upon different kind of cyclic loading signals.
1 INTRODUCTION
Offshore wind energy is a keystone in the struggle

against global warming and the total offshore capacity
is strongly increasing. Foundation design is a crucial
issue to ensure economic viability of offshore projects
and may represent up to one third of their total cost
(Byrne and Houlsby 2003). Therefore there is a real
need of innovative foundation techniques and design
procedures.
Among different types of foundations, suction
caissons, also termed bucket foundations or suction
anchors should be highlighted (Iskander et al. 2002,
Houlsby et al. 2005). They consist of a hollow cylin-
der open towards the bottom. Their top (the lid) can
Figure 1. Sketch of the installation process.
be a stiffer plate or a dome. Installation of suction
caissons is straightforward and does not require heavy
equipment. Initially the caisson penetrates the seabed
under its own weight. Water trapped inside is allowed The cyclic nature of the loading complicates the
to escape through an opening. It is pumped out after- problem since the behaviour of sand in this case is quite
wards, creating a differential of fluid pressure between complex. It is widely studied in the fields of earthquake
inside and outside as shown in Figure 1. This differen- (Ishihara 1996) and offshore geotechnics (Rahman
tial of pressure digs the caisson into the soil (Senders et al. 1977). The main outcome is the pore water
and Randolph 2009). pressure (PWP) and settlement accumulations with
The foundation type may be monopod or multi- increasing number of cycles. This general behaviour
pod. In the first case, there is a single foundation is also observed for suction caissons experimentally
per wind turbine subjected to a combination of hori- (Kelly et al. 2006) and numerically (Cerfontaine et al.
zontal and overturning moment (Achmus et al. 2013). 2015). Therefore the modelling of this sand behaviour
In the second, the superstructure lies on a system of requires specific constitutive laws.
foundations. The large overturning moment is mainly This paper presents numerical drained and partially
transformed into a push-pull loading of the suction drained simulations of monotonic and cyclic loadings
caissons (Senders and Randolph 2009). Therefore the of a suction caisson embedded in dense sand upon
behaviour of the caisson under large extraction load is vertical loading. The first objective is the deep under-
one of the main issues (Houlsby et al. 2005). standing of the different mechanisms of resistance of
327
the caisson upon monotonic loading and their inter-
actions. The second objective is to understand the
cyclic behaviour of suction caissons in the light of their
monotonic response.
2 NUMERICAL MODEL
2.1 Geometry
A sketch of the investigated suction caisson is provided
in Figure 2. The studied behaviour is purely vertical
then the mesh is axisymmetric. The cross section of
the caisson is assumed circular. The diameter D of the
caisson is equal to 7.8 m and its length L to 4 m. The
caisson is made of a stiff lid (0.4 m thick), closing its
upper aperture, and a more flexible skirt (0.1 m thick).
The behaviour of sands is inherently non-linear and
involves plasticity effects such as contractancy and
dilatancy. Therefore elastic models are not sufficient.
Classical elasto-plastic models are able to reproduce
the monotonic behaviour of sands but not the cyclic
one, involving plasticity during loading and unload- Figure 2. Zoom on the mesh adopted around the caisson.
ing. The Prevost model is adopted in the following
(Yang et al. 2003). This model is made of a yield and
an arbitrary number of hardening surfaces discretising
the field of hardening moduli. It takes into account the
phase transition described in (Ishihara 1996). The evo-
lution of the stiffness parameters with confinement (p )
is introduced.
A full description of the model implementation into
the finite element code LAGAMINE and calibration
of parameters can be found in (Cerfontaine 2014).
These parameters correspond to a very dense Lund
sand (Ibsen and Jacobsen 1996). The phase transition
line slope η̄ is equal to 1.15 and a cohesion shift pc to
5 kPa.
The superficial sand layer outside the caisson is
prone to liquefaction due to its low confinement. How-
ever modelling its post-liquefaction is meaningless
in the scope of this study since it does not con-
tribute significantly to the resistance of the foundation. Figure 3. Sketch of the loading applied to the caisson and
It is modelled by a linear elastic soil layer (E, ν) = initial stresses.
(10 MPa, 0.15). The depth of this layer is limited to
0.8 m. It includes the first two rows of elements. Sim- set up between the soil and the caisson. The installation
ilarly an elastic toe is also set up under the tip of the phase of the suction caisson is not considered.
caisson as shown in Figure 2. It compensates the over-
estimated width of the skirt. A detailed justification of
this approximation can be found in (Cerfontaine et al. 2.2 Boundary conditions and initial stresses
2015). The lower limit of the mesh is deemed impervious, i.e.
The soil is assumed to be a very dense sand (relative it corresponds to a layer of consolidated clay under
density of 90%). The specific weight of the solid grains the sand layer for example. The right and upper sides
is equal to 26.5 kN/m3 , the porosity of the soil to 0.36 of the mesh are considered drained. They respectively
and its permeability to 5 · 10−12 m2 (corresponding to correspond to the continuity of the sand layer and to
5 · 10−5 m/s) (Andersen et al. 2008). The caisson is the transition between the sand layer and the sea.
made of steel and assumed to remain elastic-linear. Its The sea level is considered to be 10 m over the sand
parameters are equal to (E, ν) = (200 GPa, 0.3). layer. It is taken into account by a vertical pressure
The 26 m × 24 m mesh is composed of 2364 hydro- of 100 kPa applied at the top of the soil, as repre-
mechanical coupled finite elements and 7085 nodes. A sented in Figure 3. The corresponding initial pore
description of these elements can be found in (Gerard water pressures (PWP) are set up accordingly in the
et al. 2008). Hydro-mechanical interface elements are whole domain. Effective stresses are initialised within
328
the soil (and the interface), due to its self weight. The
coefficient of earth pressure at rest K0 is assumed equal
to 1.
2.3 Loading of the caisson

The loading of the caisson consists of a stress-
controlled signal applied at the top of the lid, as shown
in Figure 3. Monotonic loading is simply a positive
(compression) or negative (traction) pressure applied
uniformly at a constant rate. The cyclic loading is a Figure 4. Comparison of reaction components: Compres-
more complex pressure signal and is described in the sion load Ftot .
following.
2.4 Interface elements

The accurate modelling of suction caissons neces-
sitates the use of hydro-mechanical coupled finite
elements of interface. From a mechanical point of
view, the soil-caisson shearing along the skirt is impor-
tant especially in traction. In compression, the top of
the caisson is in contact with the soil but this contact
may be lost during the simulation and soil-caisson trac-
tions can not exist. From the hydraulic point of view,
the evolution of the PWP along the walls of the caisson
modifies the maximum shearing available. Moreover
the opening of a gap along the skirt introduces prefer-
ential paths for water flows. The interface element is
extensively described in (Cerfontaine et al. 2015). Figure 5. Drained compression simulation: Variations of
The mechanical contact is enforced through a global reaction components.
penalty method such that
diverge from their initial value. However the simula-
tion is not undrained and they are able to dissipate
progressively.
where ṗN is variation of effective contact pressure, KN A positive or negative variation of vertical load
a penalty coefficient (equal to 1010 N/m3 ) and ġN the applied to the caisson is balanced by different com-
variation of the distance between two sides of the inter- ponents of reaction, described in Figure 4 in case of
face. The effective contact pressure is obtained from compression. The variation of total load applied on
the following relation the upper part of the lid of the caisson is termed Ftot .
The integral of the variation of PWP distribution on
the lower part of the lid of the caisson is represented
by Fpw . The integral of the effective normal contact
where pN is the total pressure at the soil-caisson inter- stresses on this lower part are gathered into Flid . The
face and pw is the water pressure. The maximum integral of the variation of shear stresses along the skirt
shearing resistance is bounded by of the caisson is denoted Fin inside and Fout out-
side. In the following they are depicted with respect to
the displacement of the top center of the caisson y.
Stress-controlled simulations are carried out up to
the local failure of a material point or to the global
where µ is the friction coefficient, equal to 0.5. When failure of the soil-caisson system. The post-failure
this maximum shear stress is reached, there is a free behaviour of the system of foundations is not repre-
relative tangential displacement between the two sides sented. Thence the maximum displacement remains
of the interface. limited.
3 MONOTONIC LOADING
3.1 Compression simulations
Two configurations of monotonic loading are consid- 3.1.1 Drained configuration
ered : drained and partially drained. In the former, the Drained results upon compression load are provided
loading rate is assumed very slow with respect to the in Figure 5. The simulation underlines the sequential
PWP dissipation rate within the soil. Therefore the mobilisation of the reaction components. Up to 4mm
PWP are constant. In the second case, PWP generated of settlement, the main part of the total load Ftot
329
Figure 7. Partially drained compression simulation,
Figure 6. Comparison of the settlement of the caisson and k = 5 · 10−12 m2 , rate of loading 0.4 MN/s: Variations of
the surrounding soil for Ftot =4.3MN: Drained simulation. global reaction components.
applied to the caisson is sustained by the shearing that

develops along the sides of the caisson. However shear
stresses within the interface progressively reach their
maximum value.
Friction is not mobilised simultaneously on both
sides of the skirt. Indeed, the soil-caisson relative dis-
placement is lower inside since the soil settles with
the caisson as depicted in Figure 6. This is due to
the effect of the load transferred by the lid of the
caisson and the shear mobilised depends on the soil-
caisson relative displacement. Another consequence is
the increasing confinement of the soil inside. There-
fore the Fout component firstly reaches its maximum
value but this latter increases due to the increasing
confinement. Finally, further increments of total load
applied are reported mainly under the lid and the tip
of the caisson.
Figure 8. Comparison of the settlement of the caisson and
3.1.2 Partially drained configuration the surrounding soil for Ftot = 4.3 MN: Partially drained
A partially drained compression simulation is illus- simulation.
trated in Figure 7. The total load is increased up
to 4.3 MN at a rate of 0.4 MN/s and kept constant nearly no effective stresses transferred by the lid to the
afterwards. soil and the relative displacement between the soil and
The soil-caisson system is much stiffer in partially the skirt is almost null inside as shown in Figure 8. The
drained than in drained conditions. The displacement increasing of tip Ftip , lid Flid and inside shearing
required to reach the same load is almost half in the Fin components is delayed when the early generated
partially drained simulation. This is due to the new PWP dissipate.
component of reaction induced by the PWP generation At the end of the pushing phase, the PWP have
Fpw . The loading compression of the suction caisson time to dissipate and the settlement increases while
is nothing but a classical consolidation process. Vari- the total load is kept constant. The load sustained by
ations of PWP are “trapped” inside the caisson due to Fpw is transferred to the other components of resis-
its skirt, limiting their dissipation. The variation of the tance. The normal effective stress within the interface
field of PWP generated within the soil upon a com- also increases du to the drainage process. There-
pression load is illustrated in Figure 9. The difference fore the maximum value of shearing components also
of pressure between inside the caisson and outside it increases. This illustrates that the increase in stiffness
is at the origin of Fpw . and resistance is a purely transient effect.
During the compression phase, the outer shearing
Fout and suction Fpw are the main contributors to 3.2 Traction simulations
the resistance of the caisson as depicted in Figure 7.
Indeed, the PWP inside are not yet dissipated and In Figure 10, the pull drained simulation illustrates that
the behaviour is almost undrained. Therefore there is only the two components of friction Fin and Fout ,
330
Figure 11. Partially drained traction simulation: Variations
Figure 9. Variation of the pore water pressure in the of global reaction components, rate of loading 0.4 MN/s.
soil surrounding the caisson upon compression, partially
drained case, Ftot = 4.3 MN, k = 5 · 10−12 m2 , rate of load-
ing 0.4 MN/s.
Figure 12. Influence of the permeability on partially drained

results, pull rate 0.4 MN/s.
Figure 10. Drained traction simulation: Variations of global by considering the fluid flow surrounding the cais-
reaction components. son. Indeed, if the loading rate of the caisson is equal
to 0.4 MN/s, the total load sustained for a displace-
ment of −1.5 mm is increased by almost 50%. This
actively contribute to the resistance to traction. The phenomenon is supported by experimental (Byrne
variation of Ftip is only due to the deconfinement and Houlsby 2002) and numerical (Thieken et al.
of initial stresses and does not play an active role in 2014) evidences. The negative variations of fluid pres-
the resistance. The contact is lost under the lid and sure increase the normal effective stress within the
effective traction stresses are not admissible. Therefore soil-caisson interface and then the maximum fric-
the lid component Flid is equal to zero. tion available. The absolute value of Fout is slightly
The difference of stiffness between inner and outer greater than in drained conditions.
friction components proceeds from the uplifting move-
ment of the soil inside the caisson. The relative
soil-caisson displacement is reduced and so is the 3.3 Influence of soil’s permeability in traction
shear stress mobilisation. The outer friction is fully
mobilised after an upward movement of 1.5 mm and a It was shown that the suction component can signif-
plateau is reached in Figure 10. Therefore the increas- icantly increase the total resistance to traction load.
ing load is sustained only by the mobilisation of shear Permeability of the soil and rate of loading both mod-
stress within the inner interface. Simulation stops ify the available suction. Traction simulations for three
when it is fully mobilised and no additional load can order of magnitudes of permeability are illustrated in
be sustained. Figure 12.
The partially drained simulation depicted in Fig- The stiffness of the soil-caisson system increases
ure 11 illustrates the increase of resistance obtained with decreasing permeability of the soil. Indeed, the
331
negative PWP inside the caisson are less rapidly dis-
sipated, maintaining a high differential of pressure.
However the corollary effect is the generation of high
pressure gradients leading to local liquefaction of the
soil.
For a similar applied traction load, the uplift dis-
placement of the caisson also increases with increasing
permeability. For the highest permeability, there are
clearly two distinct phases. During the first phase
(down to almost −1.5 mm), the stiffness of the caisson
is only slightly different from the other simulations.
This corresponds to the progressive mobilisation of
friction along the caisson’s walls (inside and outside).
After this point, the stiffness suddenly degrades. This
actually corresponds to the full mobilisation of fric-
tion on the skirt inside and outside the caisson. The
caisson slides upwards and suction Fpw is the only Figure 13. Half-cycle analysis of the load signal ptot .
component of resistance. In this case, there is also a
loss of contact between the lid and the soil, creating a
gap. This gap is filled with water and the caisson acts Table 1. Number of equivalent cycles, associated ampli-
like a piston. tudes and periods.
A1 A2 A3 A4
Number of cycles [−] 50 28 4 1

4 CYCLIC LOADING ptot [kPa] 4.5 13.5 22.5 40.5
T [s] 4.6 11 11.6 11.1
4.1 Loading
The cyclic load is applied in three phases.The first con-
sists in applying monotonically a mean load ptot,mean =
20 kPa, in a drained fashion. This load corresponds to
the weight of the wind turbine and the constant compo-
nent of the storm. The second phase is the application
of a cyclic loading around the mean load. Finally a
300s consolidation phase at constant initial mean load
is simulated in order to allow dissipation of PWP and
to compute a final settlement.
The cyclic loading of the caisson originates from
the effects of wind and waves acting on the offshore
superstructure. A typical output of the analysis of a
tripod superstructure to waves and wind is presented
in Figure 14 (top figure). It is termed pseudo-random
since it results from a numerical analysis by specialised
software. It corresponds to a storm sample including an
extreme event, i.e. the biggest load encountered by the
superstructure during the storm. Only the vertical load
applied by a leg of the superstructure is considered.
Within the pseudo-random signal, high amplitude
cycles alternate with the low amplitude ones and the
effect of each type of cycle is difficult to isolate. A
half-cycle analysis (Byrne and Houlsby 2002) is car- Figure 14. Pseudo-random and equivalent cyclic load
ried out to transform the pseudo-random signal into signals.
a sinusoidal equivalent one. A half-cycle is defined
as the part of a signal between two successive cross-
ings of its mean value, as shown in Figure 13. A half
period T and a peak value ptot are associated to Ni cycles are identified in the pseudo-random signal.
each half-cycle. Therefore an equivalent sinusoidal These equivalent cycles are ordered into equivalent
cycle of identical characteristics (T,ptot ) can be load signals, as depicted in Figure 14. These signals
reconstituted. mainly differ by the position of the extreme event, at
In this study, four types Ai of periods and ampli- the beginning, in the middle or at the end. A more
tudes are identified in the pseudo-random load signal. detailed description of the method could be found in
They are supplied in Table 1. For each category Ai , (Cerfontaine 2014).
332
Figure 16. Evolution of the permanent displacement for
four load signals (equivalent 1, equivalent 2, equivalent 3
and pseudo random).
Figure 15. Comparison of cyclic variation of total pressure general (Cuéllar et al. 2014). This results from the
applied on the lid ptot and variation of mean pore pressure plasticity of the soil, implying excess PWP in partially
inside the caisson pw for two load signals: Equivalent 1 (up)
and pseudo-random (down); total load applied to the caisson
drained conditions. In the upper graph of Figure 15, it
(solid black line), average PWP inside the caisson (dashed can be observed that a maximum accumulation arises
line). after the extreme event. It is progressively dissipated
afterwards during cycles of lower amplitudes.
The cyclic loading of suction caissons can be
4.2 Results
decomposed into two parallel consolidation processes.
Figure 15 presents a comparison between the first The first, named short-term, consists of the immedi-
equivalent and the pseudo-random load signals. The ate response of the soil to the variation of the applied
variation of the total load ptot around its mean value load at the scale of a cycle. Variations of PWP are large
ptot,mean is represented as well as the averaged PWP since the load reverses before all PWP are dissipated. It
under the lid of the caisson pw . The full response is the origin of the “suction effect”. The displacement
signal is difficult to analyse due to the large number of varies accordingly and is mainly recoverable.
cycles. Therefore the envelope curve, i.e. the locus of On the contrary the second consolidation process
local minima or maxima is represented for both ptot arises from the progressive dissipation of the accumu-
and pw . lated PWP and is termed long-term. It results from
The tendency curve describes the long-term evolu- the plastic contractancy of the soil and is responsi-
tion of the PWP. If the process was totally reversible, ble of the non-recoverable settlement. Accumulation
the PWP should be equal to zero each time the cyclic of deformation during cyclic loading is also a clas-
amplitude ptot is equal to zero. However pw is not sical results since it is linked to the accumulation of
equal to zero, denoting a non recoverable part. The PWP.
locus of all these non-recoverable parts describes the The trend of settlement accumulation is computed
tendency response in Figure 15. similarly to the trend of PWP. It is the locus of the
It can be observed that the variation of PWP insidde settlements measured each time the total load applied
the caisson pw is almost identical to the variation is equal to its mean value. Only this trend is repre-
of the total load applied ptot . This is a consequence sented since the full response signal is illegible due to
of the partially drained behaviour highlighted for the large number of cycles. The evolution of this per-
monotonic simulations. A large part of the loading is manent settlement under the top centre of the caisson
sustained by a PWP variation which is hardly dissi- is represented in Figure 16. The maximum transient
pated before the load reverses. Therefore the cyclic settlement encountered during the storm event (the
effective amplitude applied to the solid skeleton of the global maximum) is also represented since it could
soil surrounding the caisson is much lower than the affect serviceability.
total cyclic amplitude applied on the caisson. Conse- Results presented converge to a similar final set-
quently this partially drained behaviour induces less tlement, justifying the pertinence of the half-cycle
stiffness degradation and settlement than a drained analysis method for the elaboration of a load sig-
behaviour. nal. However there is a small divergence between
Both response signals present a tendency to PWP them since the stress paths of material points are not
accumulation. Such an observation is classical in identical for all load signals.
undrained laboratory experiments on soil samples One of the advantages of such a load signal is
(Seed and Lee 1966) or in offshore engineering in the clarification of the effect of each type of cycles
333
(A1,A2,A3 or A4). The low-amplitude cycles lead Andersen, K., H. Jostad, & R. Dyvik (2008). Penetration resis-
to almost no plastic deformation. This is quite clear tance of offshore skirted foundations and anchors in dense
in results corresponding to Equiv. 3 load signal but sand. Journal of Geotechnical and Geoenvironmental
totally impossible to observe in the pseudo-random Engineering 134(1), 106–116.
Byrne, B. & G. Houlsby (2002). Experimental investigations
response. The second batch of cycles (A2) exhibits of response of suction caissons to transient vertical load-
a clear tendency of settlement accumulation which ing. Journal of the Geotechnical and Geoenvironmental
could be extrapolated to a larger number of cycles. Engineering 128(11), 926–939.
The asymptotic non-linear evolution of the set- Byrne, B. & G. Houlsby (2003, December). Foundations
tlement is due to the progressive dissipation of the for offshore wind turbines. Philosophical transactions.
accumulated PWP, which is maximum during the Series A, Mathematical, physical, and engineering sci-
extreme event. Therefore the sooner this event occurs, ences 361(1813), 2909–30.
the sooner this asymptotic evolution starts. Cerfontaine, B. (2014, september). The cyclic behaviour of
sand, from the Prevost model to offshore geotechnics. Ph.
D. thesis, University of Liege.
Cerfontaine, B., F. Collin, & R. Charlier (2015). Numerical
5 CONCLUSIONS modelling of transient cyclic vertical loading of suction
caissons in sand. Géotechnique 65(12).
Suction caissons represent an interesting competitive Cerfontaine, B., A. Dieudonne, J. Radu, F. Collin, & R. Char-
alternative to other types of foundations for offshore lier (2015). 3d zero-thickness coupled interface finite
wind turbines. However their behaviour upon traction element: Formulation and application. Computers and
and cyclic loading is not entirely mastered and sim- Geotechnics 69, 124–140.
plified methods for design should still be elaborated. Cuéllar, P., P. Mira, M. Pastor, J. Fernández Merodo,
This paper presents the results of monotonic and cyclic M. Baeß ler, & W. Rücker (2014, June).A numerical model
for the transient analysis of offshore foundations under
loading of a suction caisson embedded in dense sand. cyclic loading. Computers and Geotechnics 59, 75–86.
Upon traction, the main mechanism of reaction is Gerard, P., R. Charlier, R. Chambon, & F. Collin (2008). Influ-
the friction progressively mobilised along the skirt of ence of evaporation and seepage on the convergence of a
the caisson. It is mobilised inside and outside the cais- ventilated cavity. Water Resources Research 44(5).
son in drained conditions (low pull rate). Upon high Houlsby, G., L. Ibsen, & B. Byrne (2005). Suction Caissons
rate of loading, a consolidation process takes place, for Wind Turbines. International Symposium on Frontiers
generating over- or under- pressures respectively with in Offshore Geotechnics 75(September), 94.
compression or traction loads. The transient differen- Houlsby, G., R. Kelly, & B. Byrne (2005). The tensile capac-
tial of pressure between inside and outside the caisson ity of suction caissons in sand under rapid loading. In
Frontiers in offshore geotechnics, pp. 405–410.
creates a suction effect, increasing the resistance of Ibsen, L. & F. Jacobsen (1996). Lund sand no. 0. Technical
the caisson in both traction and compression. It is report, Aalborg University.
of greater importance in traction than in compres- Ishihara, K. (1996, September). Soil Behaviour in Earth-
sion. Permeability and loading rate strongly influence quake Geotechnics. Oxford University Press, USA.
stiffness and resistance of the caisson. Iskander, M., S. El-gharbawy, & R. Olson (2002). Perfor-
The cyclic loading of suction caissons is mainly par- mance of suction caissons in sand and clay. Canadian
tially drained. Therefore, the major part of the load Geotechnical Journal 584(may), 576–584.
variation is sustained by positive or negative variations Kelly, R., G. Houlsby, & B. Byrne (2006). A comparison of
of PWP within the soil inside the caisson and around field and laboratory tests of caisson foundations in sand
and clay. Géotechnique 56(9), 617–626.
it. The principal consequence is a low loading of the Rahman, M., H. Seed, & J. Booker (1977). Pore pressure
solid skeleton of the soil. All simulations present an development under offshore gravity structures. Journal
accumulation of settlement during the cyclic loading of the Geotechnical Engineering Division 103(December
of the caisson. However this settlement is reduced with 1977).
respect to a purely drained behaviour. Seed, B. & K. Lee (1966). Liquefaction of saturated sands
The accurate description of the mechanisms of during cyclic loading. Journal of Soil Mechanics &
resistance could lead naturally to a spring-dashpot Foundations Div 92.
macro-element, describing the behaviour of the cais- Senders, M. (2008). Suction caissons in sand as tripod
son. This is already done in (Senders 2008) but the foundations for offshore wind turbines. Ph. D. thesis,
University of western Australia.
definition of the different parameters is still an issue. Senders, M. & M. Randolph (2009). CPT-Based Method
Moreover, the modelling of settlement accumulation for the Installation of Suction Caissons in Sand. Jour-
upon cyclic loading requires the elaboration of a more nal of Geotechnical and Geoenvironmental Engineer-
complex non-linear law for the springs. ing 135(January), 14–25.
Thieken, K., M. Achmus, & C. Schröder (2014, July). On the
behavior of suction buckets in sand under tensile loads.
REFERENCES Computers and Geotechnics 60, 88–100.
Yang, Z., A. Elgamal, & E. Parra (2003). Computational
Achmus, M., C.T.Akdag, & K.Thieken (2013). Load-bearing model for cyclic mobility and associated shear defor-
behavior of suction bucket foundations in sand. Applied mation. Journal of Geotechnical and Geoenvironmental
Ocean Research 43, 157–165. Engineering 129(12), 1119–1127.
334
Modelling the environmental impact of underground coal gasification
B.D. Roullier, P.A. Langston & Xiang Ling Li

Faculty of Engineering, University of Nottingham, Nottingham, UK
ABSTRACT: Underground coal gasification has the potential to access vast energy resources in a safe, eco-
nomical and efficient manner, with many environmental advantages over traditional methods. Previous trials
have led to local environmental issues concerning surface subsidence, groundwater contamination and water
table depletion. The risk of such issues can be greatly reduced however, through a greater understanding of the
coupled thermal, mechanical and hydraulic phenomena involved. Such an understanding allows operators to
design gasifiers to reduce or completely eliminate these risks. Numerical simulations provide a cost effective
means of investigating these issues. This paper uses the commercially available 2D discrete element code UDEC
to model these effects. Initial results show agreement with field trial experience and previous predictions, despite
a relative lack of available validation data. Future modelling aims to investigate the influence of a number of
design decisions on environmental performance to produce guidelines for future operations.
1 INTRODUCTION
1.1 UCG technology

Underground coal gasification (UCG) is an energy
extraction technology in which coal is converted into
a valuable synthesis gas (syngas) in-situ within an
unmined coal seam. The syngas typically consists of a
mixture of carbon monoxide, carbon dioxide, hydro-
gen, methane and water vapour. The syngas can be
used as a fuel for gas fired power generation or indus-
trial processes or can be converted into commodity
chemicals or synthetic transport fuels. UCG operations
involve drilling two wells down from the surface into
the coal seam and connecting these with a horizontal
channel to form a gas circuit. Oxygen is injected via
one of the wells and the coal seam is ignited with the
use of a burner inside the well casing. The flowrate of
oxygen is controlled in order to promote gasification
rather than combustion of the coal seam. Operating
pressure is maintained at a level slightly below the local Figure 1. Schematic of CRIP UCG configuration and
hydrostatic pressure. This both prevents the escape of cavity.
syngas into the surrounding rock and causes a small
inflow of groundwater into the cavity. The incoming
water helps regulate temperatures and increases the of one cavity ceases, the injection point is retracted
concentration of hydrogen and methane in the syngas. several metres and gasification begins again in a new
After gasification is complete, product gases flow cavity (Couch, 2009). A typical CRIP UCG setup is
along the channel and are extracted from the second shown in Figure 1.
well. As the reaction proceeds the channel expands
into a cavity, continually exposing fresh coal and pro-
1.2 UCG advantages and disadvantages
longing the process. Eventually, the cavity grows too
large for the coal seam, at which point gasification 1.2.1 Economic
is ceased. Modern operations use a technique known Coal is the most abundant fossil fuel on Earth, with
as the Controlled Retracting Injection Point (CRIP), economically recoverable reserves of over 800 billion
which allows several cavities to be formed from a tonnes and potential total resources of over 18 tril-
single pair of wells. In CRIP UCG, when gasification lion tonnes (Self et al, 2012). Proven reserves alone
335
account for approximately 7 billion GWh of thermal around environmental issues, a perceived lack of pro-
energy. In comparison, traditional oil and gas reserves cess control and general distrust in fossil fuel operating
represent approximately 2 billion GWh each, while companies (Shackley et al, 2006). UCG is also com-
unconventional oil and gas reserves (e.g. shale gas) are monly compared to fracking, with many of the same
estimated to be around 8 billion GWh (World Energy issues presented despite the differences in the two
Council, 2010). UCG allows for the exploitation of technologies.
coal seams which are considered uneconomical by
traditional methods due to their rank, depth or inac-
cessibility. As such, UCG could potentially access 1.3 UCG environmental concerns
reserves of stored energy even greater than those As with any fossil fuel extraction technology, there
afforded by unconventional oil and gas. are concerns around the local environmental impact of
Due to the more geographically ubiquitous nature UCG. Firstly because the gasification of coal removes
of coal, UCG could provide a viable source of energy solid material from underground, UCG can lead to
in many locations without access to indigenous oil and surface subsidence above the cavity. Secondly, heavy
gas sources, for example the UK, Western Europe and metal pollutants and coal pyrolysis products remaining
Africa. In addition, the capital costs of a UCG opera- in the cavity can leach into surrounding groundwa-
tion are much lower than that of a traditional coal mine, ter resources. Finally, sub-hydrostatic operation can
as considerably less surface equipment is required. cause the depletion of local groundwater resources,
Capital and operating costs of commercial UCG would leading to a lowering of the water table. All of these
be similar to those of shale gas, however the current issues can however be reduced, if not entirely elimi-
experimental state of the technology increases these nated, through ensuring UCG operations are designed
costs somewhat. and sited appropriately. Key design parameters include
coal seam depth, thickness and horizontal extent, cav-
1.2.2 Environmental ity operating pressure and local water table depth and
Because of the nature of the syngas product, UCG has quality. The effects of each parameter are explained
several environmental advantages over traditional coal below:
use. Firstly, due to their low volatility, many of the con-
taminant species present in coal (e.g. lead, mercury, 1.3.1 Effects of coal seam depth
cadmium etc.) remain trapped underground. Second, The depth of the coal seam to be gasified has impli-
the gaseous nature of the product allows for the use of cations for both the economics and environmental
highly efficient combined cycle gas turbines, which impacts of a UCG operation. Deeper coal seams can
greatly improve energy efficiency compared to tra- operate under higher pressures, increasing methane
ditional pulverised coal power plants. Finally, due to yields (and therefore syngas quality) and CO2 partial
the high carbon dioxide partial pressure of the syn- pressure (aiding CCS). Below approximately 800 m,
gas, UCG is ideal for connection to carbon capture and lithostatic pressure is high enough that CO2 exists as a
storage (CCS) facilities. In addition, the creation of the supercritical fluid, allowing for RZCS. Increased depth
highly permeable cavity allows some of the captured reduces the extent of surface subsidence by increas-
CO2 to be stored on site. This is referred to as reac- ing the amount of supporting overburden between
tor zone cavity storage (RZCS) (Burton et al 2006). the cavity and the surface. Finally, increased depth
Despite these advantages, it is important to state that reduces the effects of groundwater contamination as
UCG is still a fossil fuel based energy source and deeper aquifers tend to be unpotable. On the other
should be seen as a bridging technology towards an hand, drilling of wells is one of the largest capital
eventual renewable energy system (Roddy & Younger expenses of a UCG project, and increases in depth
2010). greatly exacerbate this cost.
1.3.2 Effects of coal seam thickness and extent

1.2.3 Social Increases in coal seam thickness and/or horizontal
The principal social advantage of UCG over tradi- extent are beneficial as they increase the amount of
tional mining is the greatly improved safety of miners. coal that can be gasified by a single pair of wells. While
Because no workers are sent underground, the risk to cavity height tends to be determined by the height of
life from cave-ins, flooding and explosions are elimi- the coal seam, width is usually limited by heat transfer
nated. Other social benefits include a much smaller effects. The length of a cavity is mainly determined by
surface footprint (and hence a reduced landscape the spacing between wells. Longer spacings increase
impact) as well as reductions in dust, noise and traffic. the amount of coal accessed, but also increase drilling
UCG also has the potential to create high tech jobs costs, pressure losses and the chance of a cave-in
and revitalise the economies of regions which were blocking the channel. Cavity width, length and height
previously dependent on coal mining. are all positively correlated with subsidence extent,
A particular issue with the social acceptance of however this effect is greatly reduced with increasing
UCG is the negative public perception of the technol- depth.
ogy. The key reasons for this are a lack of publically In coal seams with large horizontal dimensions,
available information on the technology, concerns multiple UCG cavities can operate in parallel to exploit
336
greater volumes of coal. A minimum separation dis- 1.5 Model aims
tance is required in order to prevent cavities from
The model presented in this paper aims to predict, and
merging, increasing the effective cavity width and
therefore prevent, the coupled effects of subsidence,
exacerbating subsidence. The use of multiple cavities
groundwater contamination and water table depletion
has a further benefit, in that produced gases can be
caused by CRIP UCG.The model operates on a general
blended to achieve a desired syngas composition.
basis rather than being site specific. The design aims to
produce simulations which can run in under a week on
1.3.3 Effects of operating pressure
a standard desktop PC. It is envisioned that the model
Increases in cavity operating pressure have a number
will be used as a tool for investigating the influence
of economic benefits as shown in section 1.3.1. Such
of site design parameters on environmental issues. In
increases also give rise a number of issues however.
addition, simulation results could be used as a ‘first
Firstly, excessive operating pressures can cause syn-
pass’ filtering tool for the selection of sites for future,
gas to leak into the overburden rock. This both reduces
commercial UCG developments.
the efficiency of the UCG process and provides a path-
The following section describes the design and
way for pollutants to enter groundwater. As mentioned
development of the model. Section 3 presents initial
previously, reduced pressures cause groundwater to
model results alongside validation and verification
flow into the cavity rather than gas to flow out. This
studies. Section 4 discusses the advantages, disad-
reduces the risk of pollution, however it can increase
vantages, challenges and limitations of the model
the chance of water table depletion if drawdown rates
methodology as well as potential future applications.
are greater than the rate of replenishment. In addition,
Section 5 concludes the paper and suggests further
excessive rates of drawdown can lead to a quenching
developments and applications of the model.
of the gasification reactions. In practice, cavity oper-
ating pressure is usually set to give a small net inflow
of water.
1.3.4 Water table effects 2 MODEL DESIGN

The depth of the water table below the ground’s sur-
face has a profound effect on UCG cavity design. As The thermal, hydraulic and mechanical behaviour of
operating pressure is usually set below local hydro- the rock mass around a UCG cavity is simulated using
static pressure, deeper water tables reduce the potential the commercially available 2D Universal Distinct Ele-
operating pressure of the cavity. In cases where the ment Code (UDEC v4.01) (Itasca Consulting Group
water table lies below the cavity, operating pressures 2004). UDEC is a combined finite difference – discrete
can be set higher as there is no water present to be element code which has been successfully applied to
contaminated. In this case pressure is only limited by geotechnical problems in a range of fields, including
the need to minimise the loss of valuable syngas to the tunneling (Solak, 2009), mining (Keilich et al, 2006),
rock mass. Similar considerations arise where the local underground construction (Barton et al, 1994) and
groundwater is unpotable. As deeper aquifers tend to nuclear waste storage (Blum et al, 2009).
be saline (and therefore unpotable), water table issues UDEC approximates a highly jointed rock mass as
are often intrinsically related to cavity depth. an assembly of intact rocks (blocks) separated by dis-
continuities (joints). Blocks are further subdivided into
triangular finite difference elements (zones) which
1.4 Environmental modelling of UCG
allow the blocks to deform both elastically and plasti-
Subsidence, groundwater pollution and water table cally. Joints allow fluid to flow in the spaces between
depletion are highly coupled processes depending blocks. Pore pressures in the joints are driven by
on the thermal, chemical, hydraulic and mechanical the weight of the rock blocks and act upon the sur-
behavior of the rock mass surrounding a UCG cav- face of those blocks. As such, UDEC allows for full
ity. Due to the complexity of these issues, and the bi-directional coupling of mechanical and hydraulic
difficulties in observing physical processes occurring effects. Heat conduction and thermally generated
underground, numerical modelling is commonly used stresses are also simulated, giving one directional
as a means to further understanding of the processes thermo-mechanical coupling.
involved in UCG. A number of models have been pro- UDEC is not capable of simulating complex
duced over the past forty years which simulate a range hydraulic phenomena such as dissolution. As such,
of UCG aspects, including product composition (Dag- groundwater contamination modelling must be
gupati et al 2012, Perkins and Sahajwalla 2008, Seifi achieved using alternative methods. An additional
et al 2011), cavity growth mechanics (Biezen et al model is currently in development which uses the
1996, Britten and Thorsness 1988, Sarraf et al 2011), MODFLOW and MT3DMS codes (United States Geo-
subsidence (Morris et al 2009, Vorobiev et al 2008) and logical Survey 2015) to predict contaminant motion
groundwater contamination (Yang and Zhang 2009). based on permeability data taken from the results of
Despite their strong coupling, no model yet considers the UDEC model.
both subsidence and groundwater effects, though one The UDEC model uses the plane strain assump-
is in development (Nitao et al, 2011). tion to model the 3D cavity as a 2D cross section. This
337
Table 1. Design parameters used in the base case UCG
model.
Parameter Value Reason/Source
Cavity Width 12.4 m Lindblom et al, 1990

Cavity Height 14.2 m Lindblom et al, 1990
Cavity Depth 98.4 m Lindblom et al, 1990
Water Table Depth 27.6 m Mason et al, 1987
Operating Pressure 555 kPa 80% of hydrostatic
Model Width 310 m 25 times cavity width
Model Height 126.8 m See Fig. 2.
Fine Zone Width 37.2 m 3 times cavity width
Fine Zone Height 56.8 m 4 times cavity height
Fine Joint Edge Length 0.8 m Runtime considerations
Coarse Joint Width 31 m One tenth of model
Coarse Joint Height 5m Runtime considerations
assumption both simplifies the model and speeds solu-

tion. On the other hand, the assumption can potentially
lead to an overestimation of environmental damage as
the cavity is effectively infinite in length. Given that
the aim of the model is to predict the negative conse-
quences of UCG, this overestimation is deemed to be
acceptable as it provides an effective safety factor for
gasifier design.
The two dimensional geometry of the cavity is based
Figure 2. Geometry and joint patterns used in the UCG
on the Rocky Mountain 1 field trial site (Lindblom et al
model.
1990). Model design parameters are given in Table 1.
A dual scale representation of the rock mass is used
to limit the number of blocks and zones in the model, 3. Excavate cavity volume
considerably reducing solution times. In the region 4. Set cavity pressure equal to operating pressure
close to the cavity, the joint pattern is based on the dis- 5. Determine temperature profile around cavity
crete fracture network (DFN) as presented by Min et al 6. Find hydromechanical equilibrium after operation
(2004). This pattern allows for random variations in 7. Set cavity pressure equal to hydrostatic pressure
model behaviour whilst maintaining appropriate over- 8. Reset temperature profile to background levels
all results. Far from the cavity, a horizontal brickwork 9. Find hydromechanical equilibrium after cooling
pattern is used for simplicity. The geometry of the base 10. Analyse model results
case model is shown in Figure 2.
The rock mass is modelled using a strain soft- The thermal part of the UCG model cannot be sim-
ening Mohr-Coulomb constitutive relationship. The ulated at the same time as the mechanical or hydraulic
strain softening representation is used as UDEC’s stan- parts due to the large difference in the characteristic
dard Mohr-Coulomb relationship does not fully realise timescales of these effects, as well as differences in the
plastic strains. Material properties are derived for spatial discretization used in each aspect of the model
sandstone using simulated triaxial compression testing (Itasca Consulting Group 2006).
as described in a previous paper (Roullier et al 2015).
Rock mass permeability is calibrated against field data
from the Rocky Mountain 1 site (Mason et al 1987).
3 RESULTS
The temperature dependence of material properties is
based on the work of Ranjith et al (2012). Tempera-
Initial results of the coupled thermal, mechanical and
tures are set at the cavity wall and vary with time over
hydraulic models show reasonable agreement with
a 100 day period to simulate the consecutive stages of
both field data and empirical predictions. Figure 3
combustion, gasification and cooldown. The thermal
compares the simulated subsidence profile for the
model is set up to determine the maximum tempera-
Rocky Mountain 1 site with a simple empirical influ-
ture experienced at each location over the entire period
ence function (Brauner, 1973). Field measurement
of operation. Rock material properties are then altered
data is not available for this site as subsidence was
based on the maximum experienced temperature.
considered of little importance during the trial. Data
Solution of the coupled thermal, hydraulic and
was available however, for the Hoe Creek 3 trial site
mechanical models follows a ten step process:
(Ganow, 1984). Figure 4 compares simulation results
1. Set up model geometry and material properties with site measurements for this site. In both figures,
2. Find initial hydromechanical equilibrium average subsidence was found as the arithmetic mean
338
Figure 5. Pore pressure distribution around Rocky Moun-
tain 1 cavity. Pressure contours are in units of kilopascals.
Figure 3. Comparison of simulated subsidence profile with
empirical predictions for the Rocky Mountain 1 field trial.
Figure 4. Comparison of simulated subsidence profile with Figure 6. Simulated maximum temperature profile around
measured site data for the Hoe Creek 3 field trial. Rocky Mountain 1 field trial site. Temperature contours are
in units of Kelvin.
of ten simulations while error bars are given as the stan-
dard deviation of these simulations. Model runtimes thermal penetration length of UCG is on the order of
were found to average 54 ± 13 hours. metres, in good agreement with the literature (Sarhosis
Model results show reasonable agreement with both et al, 2013, Yang, 2006).
site measurements and empirical predictions. Simu- In addition to the results presented above, a num-
lated profile shapes are observed to be sharper than ber of further systems are currently being simulated
those given by empirical means, but to agree well with to determine the validity of the model. Sensitivity
field measurements. In addition, the variability in sim- analyses, parametric studies and further comparisons
ulation results is seen to be very large. The reasons for with field trials in both UCG and related fields (e.g.
this variability are discussed in section 4. mining) are underway. Further plans involve the appli-
As well as subsidence, the coupled model produces cation of both the UDEC model and the contamination
results for both the hydraulic and thermal effects of model (currently in development) to a range of poten-
UCG. Figure 5 shows the pore pressure distribution tial site designs. Models will consider variances in
around a cavity based on the Rocky Mountain 1 geom- cavity geometry, cavity operating conditions and local
etry. As seen, the influence of UCG on groundwater geology including the presence of faulting.
pressures does not extend far from the cavity itself. A
small cone of depression is seen up to approximately
5m from the cavity wall, and some distortion of the 4 DISCUSSION
pressure distribution is seen directly above the cavity.
Water table height was seen to be unaffected by UCG As seen in section 3, the results of the simulations
operation. This agrees with the results of modern UCG show reasonable agreement with field trial data, yet a
field trials, for example the Bloodwood creek trial site, number of issues are observed.
in which negligible changes in groundwater pressures Firstly, as seen in Figure 3, simulations tend to give
were observed (Green, 2015). sharper subsidence profiles than empirical measures.
Finally, the coupled model gives information on the It is believed that this discrepancy is due to the empiri-
temperature profile around the cavity. Figure 6 shows cal methods not taking into the account the exact shape
the maximum temperature experienced by each point of the cavity. In effect, the curved roof of the modelled
around the cavity wall. It can clearly be seen that the cavity acts to concentrate rock deformation across a
339
smaller horizontal distance than would be given by Finally, model development is complicated by the
the rectangular cavity of the empirical method. The lack of available verification data. The majority of past
close agreement between the simulated and measured UCG trials took place in the 1980s or earlier, and many
profiles in Figure 4 support this explanation. of the more recent trials are operated as commercial
A second issue is the high degree of variability in ventures (Bhutto et al 2013, Underground Coal Gasi-
model results. Differences between two simulations fication Association 2011). As such, site measurement
for the same geometry are principally due to the ran- data is rare either because it was not collected at the
dom nature of the DFN pattern used in the fine mesh time (due to technical limitations or lack of foreknowl-
region. Model variability can therefore be reduced edge) or because it is considered to be commercially
by decreasing the standard deviation of joint angle. sensitive. Due to this lack of data it is difficult to
Further reductions can be made by tightening conver- compare model results with real world experience,
gence criteria; though this can lead to considerable considerably limiting the reliability of model predic-
decreases in solution speed and stability. Some degree tions. In order to overcome this limitation, a number
of variability in model results is necessary however, of verification models are being performed against
to account for the inherent uncertainties in rock mass processes with similar geometries to UCG, including
behaviour at real sites. Measured material properties mining and tunneling. These studies will go some way
of rock masses vary considerably, even for rocks of the towards validating model results, however the consid-
same type acquired from the same site. These varia- erable differences between these systems and UCG
tions are caused by several mechanisms including the may reduce their usefulness. Further validation could
spatial variability of the rock mass, the inherent ran- be performed against future UCG operations, although
domness of joint patterns and experimental error in the these may also suffer from the issue of commercial
collection of material data. Simplifying assumptions sensitivity.
in both numerical models and analytical theory also In addition to subsidence and water table depletion,
contribute to variability. (Cai et al 2011). Given these the aim of this model included simulation of ground-
natural and modelled variabilities, it is recommended water contamination caused by UCG. As mentioned
that any results based on this or other numerical mod- previously, this capability is currently being developed
els are repeated a number of times to ensure their with the use of the MODFLOW and MT3DMS codes.
validity. Unlike the subsidence model, the lack of data is not an
An additional complication with the model is the issue for the fluid model for two reasons: Firstly, fluid
introduction of errors due to uncertainties in inter- properties such as diffusivity are much easier to obtain
nal variables. As mentioned above, material properties than rock mass material properties. This is principally
for the rock mass are based on previous work into due to the relative ease with which these properties
simulated compression testing. Although this method can be investigated in a laboratory setting. Unlike rock
gives advantages over arbitrary selection of proper- mass properties, fluid characteristics are also not site
ties, it introduces uncertainty based on the errors in the specific, with the exception of adsorption coefficients.
simulated tests. Ideally material property data would The second reason refers to verification data. Ground-
be based on site investigations, but this is often too water compositions can be readily measured at trial
difficult and expensive to achieve. Joint material prop- sites using borehole sampling methods. In addition,
erties are chosen arbitrarily, based on suggestions in many authorities place a legal requirement on opera-
the literature. Due to the difficulties in examining rock tors to measure pollutant concentrations and the results
masses without disturbing them, reliable joint mate- are often in the public domain. As such, verification
rial property data is rare. In addition, the Coulomb data for the pollution model should be much easier to
slip model used to represent joints in UDEC does not find. A further advantage of the pollutant model is its
account for effects such as roughness or non-linear relative simplicity. Compared to the UDEC model, the
closure behaviour, further increasing uncertainty. Data pollutant code is based on phenomena which are more
on water table properties are available for many sites, readily understood, hence validation of model results
however these must be assumed in cases where no data should also be easy to achieve.
is given. In-situ stresses at sites are rarely reported Once the contaminant model is developed, the fully
due to the complication of measuring them. Modelled integrated UCG model will be complete. In addition to
in-situ stresses are based on the weight of the over- the validation and verification tests mentioned above,
burden and assume a lateral earth coefficient of 0.5. a number of tests will be performed to investigate
This value is chosen arbitrarily from the literature and how certain design choices affect the environmental
can vary considerably. UDEC specific values such as risks of UCG. As mentioned in the introduction, these
finite difference zone density and the magnitude of tests will focus on the influence of cavity geometry,
convergence criteria are often chosen by investigating operating conditions and water table properties. The
how these values affect results and selecting values effects of geological discontinuities (i.e. faults) will
which give a compromise between accuracy, stability be modelled and investigations into multiple cavity,
and runtime. As mentioned previously, a number of commercial scale operations will be performed. It is
tests are currently underway to further examine the hoped that model results will show the expected trends
effects of various internal parameters on results and in subsidence, groundwater pollution and water table
ensure the validity of the model. depletion. Results from these tests could then be used
340
to derive a set of guidelines for the appropriate siting REFERENCES
and design of future commercial scale UCG opera-
tions, as well as proving the validity of the model for Barton, N., By, T.L., Chryssanthakis, P., Tunbridge, L., Kris-
use in future predictive modelling of proposed UCG tiansen, J., Loset, F., Bhasin, R.K., Westerdahl, H. & Vik,
G. 1994: Predicted and measured performance of the 62m
operations. span Norwegian Olympic ice hockey cavern at Gjovik.
International journal of rock mechanics, mining sciences
and geomechanics. 31(6): 617–641
Bhutto, A.W., Bazmi, A.A. & Zahedi, G. 2013: Underground
5 CONCLUSIONS coal gasification: from fundamentals to applications.
Progress in energy and combustion science 39: 189–214
Underground coal gasification is seen to be a promis- Biezen, E.N.J. 1996. Modelling underground coal gasifica-
ing technology for both energy generation and indus- tion. PhD Thesis. Delft University of Technology, The
trial applications. UCG is shown to have a number of Netherlands
economic, social and environmental benefits over tra- Blum, P., Mackay, R. & Riley, M.S. 2009: Stochastic sim-
ditional coal extraction and use. As with all energy ulation of regional scale advective transport in fractured
technologies however, there are environmental con- rock masses using block upscaled hydromechanical rock
property data. Journal of hydrology, 369(3–4): 318–325
cerns associated with UCG. Due to the expense and Brauner, G. 1973: Subsidence due to underground mining
difficulty of performing field scale UCG experiments, (in two parts) 1. Theory and practices in predicting sur-
numerical models are commonly used to understand face deformation. Bureau of Mines, US Department of the
and predict these issues. Discrete element modelling Interior.
is often used to account for the discontinuous nature Britten, J.A. & Thorsness, C.B. 1988: A mechanistic model
of the rock mass around a UCG cavity. for axisymmetric cavity growth during underground coal
A coupled simulation of thermal, mechanical and gasification. Journal of the American Chemical Society
hydraulic effects has been produced with the UDEC 33: 126–133
discrete element code in order to predict the local envi- Burton, E., Friedmann, J. & Upadhye, R. 2006: Best practices
in underground coal gasification. Lawrence Livermore
ronmental impact of UCG. Initial results show good National Laboratory, Livermore, USA
agreement with the literature for the prediction of sur- Cai, M. 2011: Rock mass characterization and rock property
face subsidence and water table depletion as well as the variability considerations for tunnel and cavern design.
effect of high temperatures on overburden rock. The Rock mechanics and rock engineering, 44: 379–399
variability in results is seen to be high; however this Couch, G.R. 2009: Underground coal gasification, Interna-
is on a scale comparable with the natural variability tional Energy Agency Clean Coal Centre, 2009.
of rock. In addition, the reliability of model results is Daggupati, S., Mandapati, R.N., Mahajani, S.M., Ganesh,
somewhat limited given the lack in available site data A., Pal, A.K., Sharma, R.K. & Aghalayam, P. 2012:
and verification studies. Compartment modelling and flow characterisation in non-
isothermal underground coal gasification cavities. Jour-
Further simulations are currently ongoing and it is nal of Industrial and Engineering Chemistry Research 51:
hoped that these will demonstrate the validity of the 4493–4508
model for use in future predictions of UCG behaviour. Ganow, H.C. 1984: Results of long term ground surface mea-
Investigations of the effects of varying site properties surements at the Hoe Creek 3 site. Proc. 10th Underground
on environmental issues should give trends which can Coal Gasification Symposium, Williamsburg, VA, USA.
be used to produce a set of guidelines for future gasifier 12–15 August 1984
designs. It is envisioned that the final models could be Green, M. 2015: UCG: Practical experience with UCG trials.
used as a first pass screening tool to predict the envi- Presented at the Coal Research Forum 26th annual meet-
ronmental impacts of a number of proposed sites. This ing and meeting of the coal conversion division. Leeds,
UK, 15th April 2015.
would allow further modelling efforts using high per- Itasca Consulting Group Inc. 2004: UDEC – Universal
formance computing to focus only on those sites with Distinct Element Code v4.01. Itasca consulting group,
a greater potential for safe and economically viable Minneapolis, USA.
operation. Further uses of the model could also pro- Itasca Consulting Group Inc. 2006: UDEC – Universal Dis-
vide a similar function in related fields such as nuclear tinct Element Code v4.0 user’s guide. Itasca consulting
waste storage, oil extraction, geothermal energy and group, Minneapolis, USA
carbon capture and storage. Keilich, W., Seedsman, R.W. & Aziz, N. 2006: Numerical
modelling of mining induced subsidence. In: N. Aziz (ed),
Coal 2006 – Coal operators conference, University of
Woolongong & The Australian institute of mining and
LIST OF ACRONYMS metallurgy, 2006: 313–326
Lindblom, S.R., Covell, J.R. & Oliver, R.L. 1990: Results
of phase 1 post-burn drilling and coring, Rocky Moun-
CCS – Carbon Capture and Storage tain 1 underground coal gasification site, Hanna Basin,
CRIP – Controlled Retracting Injection Point Wyoming. Western Research Institue, Laramie, USA.
DFN – Discrete Fracture Network Mason, J.M., Oliver, R.L. & Moody, C.G. 1987: Geology and
RZCS – Reactor Zone Carbon Storage groundwater hydrology of the proposed Rocky Mountain
UCG – Underground Coal Gasification 1 underground coal gasification site, Hanna, Wyoming.
UDEC – Universal Distinct Element Code Western Research Institue, Laramie, USA.
341
Min, K., Jing, L. & Stephansson, O. 2004: Determining the Seifi, M., Chen, X. & Abedi, J. 2011: Numerical simula-
equivalent permeability tensor for fractured rock masses tion of underground coal gasification using the CRIP
using a stochastic REV approach: Method and applica- method. The Canadian Journal of Chemical Engineering
tion to the field data from Sellafield, UK. Hydrogeology 89: 1528–1535
Journal, 12(5): 497–510 Self, S.J., Reddy, B.V. & Rosen, M.A. 2012: Review of under-
Morris, J.P., Buscheck, T.A. & Hao,Y. 2009: Coupled geome- ground coal gasification technologies and carbon cap-
chanical simulations of UCG cavity evolution. Proc. 26th ture. International journal of energy and environmental
annual international Pittsburgh coal conference. 20–23 engineering 3:16
September 2009. Pittsburgh, USA Shackley., Mander, S. & Reiche, A. 2006: Public perceptions
Nitao, J.J., Camp, D.W., White, T.A., Burton, G.C., Wag- of underground coal gasification in the United Kingdom.
oner, J.L. & Chen, M. 2011: Progress on a new integrated Energy Policy 34(18): 3423–3433
3-D UCG simulator and its initial application. Proc. 28th Solak, T. 2009: Ground behaviour evaluation for tunnels in
annual international Pittsburgh coal conference. 12–15 blocky rock masses. Tunneling and underground space
September 2011. Pittsburgh, USA technology. 24: 323–330
Perkins, G. & Sahajwalla, V. 2008: Steady state model Underground Coal Gasification Association. 2011: History
for estimating gas production from underground coal of UCG. Retrieved 31/07/13 from: www.ucgassociation.
gasification. Energy & Fuels 22: 3902–3914 org/index.php/ucg-technology/history-of-ucg
Ranjith, P.G., Viete, D.R., Chen, B.J. & Perera, M.S.A. 2012: United States Geological Survey. 2015: MODFLOW and
Transformation plasticity and the effect of temperature related programs. Retrieved 04/02/16 from: http://www.
on the mechanical behaviour of Hawkesbury sandstone at water.usgs.gov/ogw/modflow
atmospheric pressure. Engineering geology. 151: 20–127 Vorobiev, O., Morris, J., Antoun, T. & Friedmann, S.J. 2008:
Roddy, D.J. & Younger, P.L. 2010: Underground coal gasi- Geomechanical simulations related to UCG activities.
fication with CCS: a pathway to decarbonising industry. Proc. 25th annual international Pittsburgh coal con-
Energy and Environmental Science 2010(3): 400–407 ference. 29 September – 2 October 2008. Pittsburgh,
Roullier, B.D., Langston, P.A. & Li, X. 2015: Effects of USA.
joint geometry on the modelled material strengths of rock World Energy Council (WEC) 2010: Survey of energy
masses in the distinct element method. Proc. TC105 ISS- resources, World Energy Council, London, UK.
MGE International Symposium on Geomechanics from Yang, L. 2006: Theoretical analysis of the coupling effect
Micro to Macro. 1–3 September 2014, Cambridge, UK. for the seepage field, stress field, and temperature field
Sarhosis, V.,Yang, D., Sheng,Y. & Kempka, T. 2013: Coupled in underground coal gasification. Numerical heat trans-
hydro-thermal analysis of underground coal gasification fer part A: Applications. An international journal of
reactor cool down for subsequent CO2 storage. Energy computation and methodology. 48(6): 585–606
Procedia. 40: 428–436. Yang, L. & Zhang, X. 2009: Modelling of contaminant trans-
Sarraf, A., Mmbaga, J.P., Gupta & Hayes, R.E. 2011: Mod- port in underground coal gasification. Energy & Fuels
elling cavity growth during underground coal gasification. 23(1): 193–201
Proc. 2011 COMSOL conference. 13–15 October 2011.
Boston, USA.
342
Stability analysis of large diameter boreholes using finite element modelling
C. Morris, A. Cardoso & S. Raymackers

GeoSea n.v, Zwijndrecht, Belgium
ABSTRACT: GeoSea n.v. is currently working on the development of windfarms requiring piles to be installed
in hard clays and weak rocks. Drilling is required for pile installation and to optimize the installation it is preferable
to drill without internal support measures. The authors have reviewed methods for assessing the stability of large
diameter boreholes drilled in rock. Borehole stability is modelled using finite element modelling software
PLAXIS 2D with the Hoek and Brown constitutive model. Results from these calculations are compared against
methods from mining industry and well stability methods.
1 INTRODUCTION 2.1 Application of Hoek & Brown model

The Hoek and Brown model takes into account the
Monopile and piled jacket foundations are commonly
rock mass properties and the intact rock strength which
used for offshore wind turbine generators (WTGs).
govern the behavior and apparent strength of a rock
Pile are installed in clay and sand by driving. GeoSea
mass.
n.v. is currently working on projects where piles will
The rock mass properties are defined by the Geo-
be installed in very hard clays and weak sedimentary
logical Strength Index (GSI). Ground investigations
rocks. Driving unaided to design embedment in such
in rock are typically made using rotary core drilling to
ground conditions is unlikely to be feasible. Drilling
extract samples. The GSI is a qualitative assessment
will be required either to install piles as rock sockets
of the rock normally made by the ground investigation
or to aid driven installation.
supervisor. The GSI is based on a visual inspection of
Due to the nature of offshore work, it is desirable
the rock considering the lithology, structure and condi-
that drilling can be done without support measures
tion of internal surfaces of the discontinuities. A high
for the borehole, such as bentonite slurry. However
GSI corresponds unweathered, massive or intact rock
the boreholes to be drilled are large, typically 2–8 m
with widely spaced discontinuities that have rough
in diameter and a stability assessment of boreholes is
internal surfaces.
required to ensure the installation method is viable.
The compression strength of intact rock is defined
The authors required a suitable (analytical) method
as the uniaxial compression strength (UCS), deter-
to assess borehole stability for numerous locations dur-
mined by unconfined compression or point load tests.
ing project preparation. Locations which are thought
Intact rock is typically stronger and stiffer than soil,
to be critical could then be subject of further
behaving in a linear elastic manner until brittle failure.
studies.
The Hoek and Brown failure criterion is defined as:
This article discusses the literature review that was
performed with the goal to find analytical models
which could give an initial indication of borehole
stability based on the rock mass properties and bore-
hole diameter as well as the FE models that were
made to analyze the global factor of safety. To con-
clude both analytical and FE models are compared and
discussed.
2 LITERATURE REVIEW
where mi is a coefficient based on rock type, D is
Stresses in the following paragraphs refer to effective coefficient depending on the degree of disturbance and
stresses unless stated otherwise. The application of the σci = UCS.
Hoek and Brown (HB) rock model to this problem and The HB failure criterion is non-linear, however it
other relevant experience from industry is examined. can be approximated with reasonable accuracy by the
343
Mohr-Coulomb model over a limited deviatoric stress The methods for well stability are concerned with
range. An equivalent cohesion and friction angle can identifying zones where hole enlargement beyond the
be calculated based on the UCS and GSI using the drilled diameter is to be expected and for selecting
method defined by Hoek et al (2000). The onset of appropriate drilling mud weights.
plastic failure is defined by the linear Mohr-Coulomb Aadnoy (2010) recommends assessing wellbore sta-
failure criterion: bility by considering the stress conditions at the wall
of the well:
Radial stress: σr = Pw = borehole pressure
Tangential stress: σθ = 2σa − Pw
Vertical stress: σvo = overburden pressure
where σa = estimated average horizontal stress
Failure of the well is assumed to occur where the
Gradient of σ1 with increasing confinement σ3 : Mohr-coulomb failure criteria are exceeded, with the
radial stress equal to the minor principal stress and the
tangential stress equal to the major principal stress.
Therefore estimation of the horizontal stress is critical
for calculating tangential stress and the borehole sta-
bility. The radial pressure on the inside of the borehole
Hoek et al (2000) defines a critical internal support provided by drilling mud is critical to the stability of
pressure for tunnels, above which the behavior of the the well.
rock mass is assumed elastic and no failure occurs: It is noted that wells for oil and gas are normally
much deeper than the boreholes considered for pile
installation and the in-situ lateral stresses and resulting
tangential stresses are higher.
This assumes stress in the rock to be hydrostatic. The

radius of the plastic deformation zone, measured from 2.3 Sakurai critical strain method
the center of the tunnel:
Sakurai’s critical strain concept can be used iden-
tify situations requiring special support consideration
when the strain is above the critical strain value, as
shown in Fig 1. The critical strain is calculated as:
Elastic inward radial deformations:
The in-situ compression rock mass strength is

calculated based on UCS and GSI:
Total elastic and plastic inward radial deformation:
However, the critical strain only identifies when spe-

cial support consideration is required and does not give
Where r0 = original tunnel radius, υ = Poisson’s ratio,
detail on the behavior expected where the critical strain
E =Young’s modulus, p0 = hydrostatic soil pressure
is exceeded.
and pi = internal support pressure.
The tunnel face provides support to the tunnel
2.3.1 Local stresses around the borehole
walls and the displacements from Eq (10) are not
Hoek et al (2000) discusses spalling failure that can
expected at the tunnel face. These values are reached
occur in brittle rock. This can occur at tangential stress
approximately 1.5 diameters behind the tunnel face.
concentrations where the minor principle stress inter-
Similar behavior is expected for drilled boreholes, with
sects the borehole wall and is manifested in the form
displacements increasing as the drill advances.
of cracking, spalling and rock bursts. It is noted that
case studies presented by Hoek et al (2000) are in hard
rocks, which may not be entirely representative of the
behavior of weaker, less massive sedimentary rocks
2.2 Wellbore stability methods
considered for this study.
Methods are used for estimating the stability of wells The utilization of shear strength can be considered
in the oil and gas industry. These wells are smaller for an infinitely small soil wedge at the wall of the
diameter than boreholes being considered for installa- borehole, shown in Fig 2. The radial stress is zero
tion of piles and drilling muds are normally used for because there is no radial confinement. The tangential
internal support. stress is maximum at the borehole wall and expected to
344
Table 1. Model revisions
Description Model variations
Base case model (DC02). DC02A: Ø = 2 m

K0 = 0.5, UCS = 10 MPa, DC02B: Ø = 4 m
GSI = 50%, D = 0.5, γ = 10 kN/m2 , DC02C: Ø = 8 m
mi = 17. Excavation in 2.5 m
increments with FoS calculated
for each depth.
Same as DC02 with DC03A: Ø = 2 m
γ = 25 kN/m2 DC03B: Ø = 4 m
DC03C: Ø = 8 m
DC02B with K0 varied. DC04A: K0 = 0.75
DC04B: K0 = 1.0
DC02 extended to 100 m depth. DC07A: Ø = 2 m
DC07B: Ø = 4 m
DC07C: Ø = 8 m
Note: K0 = at rest earth pressure coefficient.
Figure 1. Percentage strain for different rock mass strength

for the Second Freeway, the Pinglin and the new Tielun Head-
determined as the reduction required for a failure
race tunnels in Tawain (after J.C. Chern reproduced from mechanism to form.
Hoek, 1998).
3.2 Model setup
The borehole has been modelled axisymmetrically.
The material properties in the base case model are
uniform with depth in order to model diameter of the
borehole as the main variable affecting stability. The
drilling of the borehole is modelled as a staged excava-
tion, excavated in 2.5 m depth increments. The FoS is
analyzed after each excavation increment is assessed.
The drilling method is assumed not to effect the bore-
hole stability. The phreatic surface is modelled above
the top level of the model. Structural geology features
are not considered and the drilling process is assumed
not to alter the rock properties.
Details of the model revisions used for extracting
results are given in Table 1.
Figure 2. The stresses acting on an infinitely small wedge

3.3 Results
of rock at the borehole wall. σvo = overburden pressure,
σθ = tangential stress, σr = radial stress. 3.3.1 Global FoS
be a function of in-situ lateral stresses present before Fig. 3 shows the FoS against failure over 100 m depth
drilling of the borehole. for 3 borehole diameters. The results confirm expec-
The utilization of the shear strength is expected to tation that increasing depth and diameter reduce the
be maximum at the borehole wall, where the tangential borehole stability. The results suggest that for the rock
stresses are highest and there is no radial confinement. properties considered there is safety margin against a
The shear strength utilization is calculated as: global failure mechanism forming. The largest reduc-
tion of FoS with depth occurs in 0–40 m interval and
beyond 40 m the decrease in FoS appears relatively
stable with depth. This appears to be because a larger
body of rock fails as the excavation depth increases
while the slip surface that shear occurs on does not
3 FINITE ELEMENT MODELLING (FEM)
increase proportionally, shown in Fig. 4.
Increased bulk density reduces the FoS, which is
3.1 Global Factor of Safety (FoS)
logical because the increased overburden stress results
The global failure mechanism will be modelled using in increased tangential stress to be resisted by the
the HB model in PLAXIS 2D. The global FoS is calcu- rock. Fig. 5 shows the effect of increasing γ 5 kN/m3 .
lated using the safety analysis function, which reduces Results in Figure 6 indicate that the in-situ lateral
the material strength progressively. The (FoS) can be stresses have a negligible influence on the global FoS.
345
Figure 3. FoS for 2 m, 4 m & 8 m diameter boreholes from
0–100 m depth.
Figure 6. Stability of a 4 m diameter borehole with K0 = 0.5

& K0 = 1.
Figure 4. Incremental displacements for DC02A at 2.5 m,

5.0 m & 7.5 m excavation depths.
Figure 7. Development of radial (left) and tangential (right)

stresses around a 2 m diameter 40 m deep borehole (DC02A).
borehole leads to an increase in tangential stress and a

decrease in the radial stress around the borehole. The
tangential stress is maximum at the borehole wall and
the effective stresses decreases to zero at the borehole
wall, as shown in Fig 7. Fig 8 shows that the tangen-
tial stresses are independent of the borehole diameter
and are equal to twice the in-situ lateral stresses before
excavation. Reduction in tangential stress occurs local
to the base of the borehole because of different stress
paths becoming available around the base of the bore-
hole. The reduction occurs further from the borehole
base for larger borehole diameters.
Figure 5. The influence of increased unit weight on FoS. The tangential stresses around the borehole are
highly dependent on the in-situ lateral stresses. Fig. 9
3.3.2 Local stresses shows tangential stresses around the borehole increase
The development of stresses around the borehole is in direct proportion to the increases in K0 . This will
examined at the end of excavation stages without result in higher utilization of the rock around the
reduction of strength properties. Excavation of the borehole.
346
Figure 10. Shear strength utilization around a 2 m diameter
Figure 8. Tangential stresses over depth interval 20–40 m 40 m deep borehole (DC02A).
for DC02A, DC02B & DC02C. In-situ lateral stress before
excavation shown as reference. 4 DISCUSSION OF FEM AND COMPARISON
WITH OTHER METHODS
4.1 PLAXIS 2D analysis

Results indicate that the relatively weak rock proper-
ties considered have a FoS > 2 against a global failure
mechanism for 2–8 m diameter boreholes up to 100 m
depth. The FoS decreases most in the 0–20 m depth
interval and then decreases at a relatively stable rate
with depth below 30 m. The FoS decreases with larger
borehole diameter.
The plastic construction stages of the model show
that very low strains result from the staged excavation.
The shear stresses at the borehole wall exceed the shear
strength of the rock mass during the excavation stages,
which is expected to cause localized failure in the form
of cracking and spalling. The extent of this failure and
whether it is self-limiting is not clear from the FE anal-
ysis. Hoek et al (2000) indicate that rock can continue
Figure 9. Tangential stresses over depth interval 20–40 m
to support load after cracking has occurred. Results
for 4 m diameter 40 m deep borehole. K0 = 0.5 for DC02B,
K0 = 1.0 for DC04B. show that tangential stresses at the borehole wall are
twice the lateral in-situ stresses and is independent
of borehole diameter. Therefore, accurate estimates of
the lateral in-situ stress is important in determining if
Fig 10. shows the utilization of shear strength localized failure of the rock occurs.
around the borehole. This indicates that the shear
strength is fully utilized at the borehole wall below
4.2 Calculations based on tunnel stability from
approximately 30 m depth. However, this occurs at
Hoek et al (2000)
very small strains and does not result in a global failure
mechanism. The borehole is assessed considering a diameter of 2 m,
It is assumed that full utilization of the rock shear 40 m deep using Mohr-Coulomb properties using the
strength local to the borehole wall will be manifested formulas described in § 2.1. Results are shown in Fig
as cracking and spalling. This agrees with experience 3. This can be compared with DC02B, which indicates
from the mining industry discussed by Hoek et al that no plastic deformation is expected, however cal-
(2000). The effect of this on the drilling process of culations using formulas of Hoek (2000) suggest that
the borehole is uncertain. If a large quantity of rock plastic deformation occurs.
is spalled it may impair the operation of the drilling Model DC02B predicts a maximum inward radial
equipment or require cleaning of the borehole prior to displacement of the borehole wall of 0.14 mm, which is
pile installation. significantly less than shown in Fig. 11. Deformations
347
The critical strain approach to assess borehole
stability gives a more optimistic result than assess-
ing the stability based on shear strength utilization
from PLAXIS models. Additionally, this method does
not account for the lateral in-situ stresses, which
PLAXIS2D results and other methods indicate to be a
critical factor. Therefore it is questionable whether this
approach is suitable for assessing borehole stability.
4.4 Wellbore stability method

The wellbore stability method from Aadnoy (2010)
assesses the stability on the assumption that tangential
stresses are double the in-situ lateral stress. This agrees
with the results of PLAXIS2D analysis. Therefore it
is reasonable to assume that the wellbore stability
method will indicate localized failure in the same loca-
Figure 11. Inward radial deformation of borehole wall & tions that PLAXIS2D shows 100% utilization of the
radius of plastic zone. rock properties. Calculations with this method are not
performed in this paper.
5 CONCLUSION
A literature study was combined with FE modelling

to find a suitable analytical model to assess bore-
hole stability on numerous locations on a windfarm
site, in order to spot the locations that require detailed
modelling.
Hoek and Brown model, Wellbore stability method
and Sakurai critical strain method were discussed and
a Plaxis 2D Hoek and Brown model was made to assess
the stability of large diameter, unsupported boreholes
in weak sedimentary rock.
Results indicate that localized failure of the rock at
the borehole wall due to exceedance of shear strength
would occur at much lower rock strengths than a global
stability failure of the borehole. Therefore, assessment
Figure 12. Analysis of borehole stability with Sakurai
critical strain. of the global stability of the borehole appears less
relevant than assessment of localized failure at the
shown in Fig 3. may overestimate the displacements borehole wall.
close to the base of the borehole because calculations Literature from the mining industry indicates that
have not accounted for the 1.5 diameters required for the localized failure at the borehole wall will occur
displacements to develop. as cracking, spalling and rock bursts. Industry expe-
rience from assessing wellbore stability shows that
failure at the borehole wall can result in borehole
4.3 Calculation with Sakurai critical strain enlargement. Localized failure may have an adverse
effect on the drilling and installation process. There-
The critical strain is calculated using Eq. (12) and fore, assessment for localized failure is considered the
compared against the strain calculated based on the most appropriate method for a first check of borehole
effective overburden pressure: stability for pile installation.
The resistance against localized failure can be
assessed using the Hoek & Brown failure criterion, set-
ting the minor principal stress equal to a radial stress
of zero and the major principal stress equal to the tan-
The in-situ strains are compared against critical gential stress. The tangential stress can be calculated
strains over the depth range 10–100 m in Fig 12. 2 as twice the lateral in-situ stress, which is in agree-
cases are considered: Case 1: UCS = 10 MPa; Case ment with the recommendations for wellbore stability
2: UCS = 0.5 MPa. The in-situ strains for Case 2 are of Aadnoy (2010) and results of PLAXIS 2D mod-
well below the critical strain, whereas the critical in- elling. Accurately estimating the lateral in-situ stresses
situ strain is exceeded at approximately 40 m depth for is critical factor in estimating the resistance against
Case 1. localized failure.
348
The Sakurai critical strain does not appear a suit- Hoek, E (1998). Tunnel support in weak rock. Keynote
able method for assessment of borehole stability. The address, Symposium of Sedimentary Rock Engineering,
results appear optimistic compared to PLAXIS 2D Taipei, Taiwan.
results for localized failure of the rock at the bore- Hoek, E. Kaiser, P.K. Bawden, W.F. 2000. Support of Under-
ground Excavations in Hard Rock. CRC Press. Boca
hole wall. Furthermore, the Sakurai method does not Raton.
take into account the lateral in-situ stresses. Sakurai, S. 1997. Lessons Learned from Field Measure-
ments in Tunneling. Tunnelling and Underground Space
Technology, Volume 12, Issue 4.
REFERENCES
Aadnoy, S.B. 2010. Modern Well Design 2nd Ed. CRC Press.
Boca Raton.
349
Thermo-mechanical hypoplastic interface model for fine-grained soils
Henning Stutz
Institute of Geo-Science, Marine and Land Geomechanics & Geotechnics, Kiel University, Kiel, Germany
David Mašín
Faculty of Science, Charles University in Prague, Prague, Czech Republic
Frank Wuttke & Boyke Prädel

Institute of Geo-Science, Marine and Land Geomechanics & Geotechnics, Kiel University, Kiel, Germany
ABSTRACT: Thermo-active geo-structures e.g. gas-/oil-pipelines, high-voltage cables, energy-piles, nuclear

waste disposals are exposed to temperature changes. These alterations have a significant effect to the behaviour
of soil-structure interfaces.
The model is an adaption of the thermo-mechanical hypoplastic model from Mašín & Khalili (2011) and
Mašín & Khalili (2012). The reformulation is done by redefined tensorial definitions for the special case of
soil-structure interfaces. The new thermo-mechanical interface model is used for the modelling of soil-structure
interface under varying temperature. By applying different temperatures and conducting a parameter study it
is be proven that the model can be used for modelling various thermo-mechanical loading paths. After the
simulations of some selected stress and temperature paths the paper discusses the benefits and advantages by
using an advanced model for soil-structure interfaces considering temperature effects.
1 INTRODUCTION in the clay-concrete tests results show an increased

shear strength with respect to an increasing tempera-
The temperature effects in soils are manifold and ture. Further a decreasing contractive behaviour with
can have different impacts on the behaviour of inter- an increase in temperature is observed. Di Donna, Fer-
faces. Thermo-active geo-structures gain more and rari, & Laloui (2015) concluded that the effect of an
more attention in the field of renewable energy pro- increasing shear stress under an increasing tempera-
duction. Typical applications as nuclear-waste stor- ture is a result of a thermal consolidation which the
age, deep and shallow geothermal applications imply sample had undergone.
new challenges for the geotechnical & geomechanical Yavari, Tang, Pereira, & Hassen (2016) conducted
community. tests with intermediate heated soils (5–40◦ C), which
If structures are embedded into soils and the tem- are typical for thermo-active geotechnical structures.
perature fluctuates a temperature dependent interface The tested soil is a Kaolin clay. In contrast to the tests
zone is established around the embedded structure. by Di Donna, Ferrari, & Laloui (2015) the samples
Especially the modelling of energy pile must consider were heated prior to the test which enables the tests to
the these temperature dependences. The shaft friction be performed without the effect of thermal consolida-
and the performance of these geo-thermal foundations tion. The results by the tests of Yavari, Tang, Pereira,
will be influenced by variations of the temperature. & Hassen (2016) indicate negligible influences for
For example Di Donna & Laloui (2015), Bodas Fre- the interface shear strength parameters. Nevertheless,
itas, Cruz Silva, & Bourne-Webb (2013) and Laloui, Yavari, Tang, Pereira, & Hassen (2016) reported a soft-
Nuth, & Vulliet (2006) emphasize the importance of ening behaviour for clay-concrete interfaces which are
the influence of the pile-soil interface for energy piles. not observed in clay-clay direct shear tests.
The temperature can trigger especially in fine- Xiao, Suleiman, & McCartney (2014) conducted
grained soils changes to the mechanical behaviour. interface shear tests with a silty soil. The results
In the case of soil-interface tests the experimental were compared with the standard direct shear tests
available data is scarce. for the same soil. They concluded that the shear
Di Donna, Ferrari, & Laloui (2015) conducted behaviour in soil-soil as well as in soil-solid tends to
sand-concrete and clay-concrete shear tests. In the increasing shear strength under increasing tempera-
sand-concrete shear test no effects of the temperature ture. Whereas, Xiao, Suleiman, & McCartney (2014)
changes had been observed and measured. Whereas, are not describing the sample preperation in detail.
351
From the perspective of continuum soil testing e.g. where σ̂ = σ/trσ, the two scalars are defined as:
triaxial and odeometric soil testing several open ques-
tions have to be answered and should be emphasized
by on-going interface research.
Mašín & Khalili (2012) hypoplastic model consid-
ers the volume change caused by heating or cooling as
where r is a model parameter. The scalar value of a is
fully reversible process. This process is described by a
defined as:
constant value of the thermal expansion coefficient αs .
The findings by (Xiao, Suleiman, & McCartney
2014, Di Donna, Ferrari, & Laloui 2015, Yavari, Tang,
Pereira, & Hassen 2016) doesnot show the same trend.
Indeed the temperature at the interface have an impact
for the soil-structure interface behaviour. Due to this Where ϕc is the critical state friction angle. α is given
reason, a thermo-mechanical hypoplastic constitutive as:
interface model is proposed. This model is based on
the thermo-mechanical hypoplastic models of Mašín
& Khalili (2011) and Mašín & Khalili (2012). The
model reformulation by preserving the tensorial nota-
tion of the model is done by an methodology presented Where λ∗ and κ∗ are model parameters. The second
in Stutz & Mašín (2015) and Stutz, Mašín, & Wut- order constitutive tensor is then defined as:
tke (2016). First the thermo-mechanical hypoplastic
model (Mašín & Khalili 2012) is introduced briefly
and the reduced stress and strechting tensors for inter-
face condition are given. Latter the model is used
to simulate different boundary conditions which are
typical for soil structure interfaces. where Y = 1 coincide with the critical stress condi-
tion of the Matsuoka–Nakai formulation. The limiting
stress condition Y is defined as:
2 THERMO-HYPOPLASTIC INTERFACE
MODEL
2.1 Hypoplastic thermo-mechanical model in

general formulation
The thermo-mechanical hypoplastic model developed
where the stress invariants are defined as:
by Mašín & Khalili (2011) is introduced briefly. The
stress-strain rate hypoplastic equation is given as:
the second order tensor m is calculated as:

where σ̇ is the stress tensor, L and N the fourth and sec-
ond order constitutive tensor, ε the strain tensor and fs
and fd the barotropy and pyknotropy factors. Mašín
& Khalili (2011) developed the thermo-mechanical
model using the hypoplastic clay model Mašín (2005)
using the factor F as:
as basis for the improvements of the model.
The stress-stretching rate equation for the thermo-
mechanical model is:
The temperature related strain is given as:

with
where is Ṫ the temperature rate. The constitutive and the Lode angle defined as:
fourth-order tensor L is given as:
352
the barotropy factor is calculated as: Finally, the evolution of the state variable e (void ratio)
is governed by:
and the pyknotropy factor: For an detailed description of the hypoplastic thermo-
mechanical model, see Mašín & Khalili (2011) and
Mašín & Khalili (2012).
2.2 Adaption of the model for interface behaviour

where pe is the Hvorslev equivalent pressure defined
as: The adaption of the full thermo-mechanical hypoplas-
tic model described in Section 2.1 is done by using
reduced stress and strain rate tensors. These are
defined as:
where the reference pressure pr = 1 kPa. The tem-

perature dependence values of λ∗ (T ) and N (T ) are
calculated as:
and for the strain
and
Parameters nt and lt control the position and slope These reduced tensors account for simple shear con-
of the Normal Compression Line (NCL) of heated ditions at the interface can be written in modified
soils. The tensorial terms HT is introduced by Mašín & Voigt-Notation as:
Khalili (2012) to incorporate the collapse effect of the
soil structure at constant effective stress for a heated
soil. HT is given by (Mašín & Khalili 2012) as:
with By using the modifications of the standard tensorial

operators used in the 3-D model from Section 2.1
and the reduced stress and stretching rate vectors the
thermo-mechanical interface model is reformulated.
A detailed description of the modified tensorial
Where fdSBS is defined as the value of fd at the State notation, which is used by the stress and strain rate vec-
Boundary Surface (SBS) passing through the current tor is outside of the scope of the paper. The interested
stress point. reader is referred to Stutz, Mašín, & Wuttke (2016)
and Stutz & Mašín (2015).
In respect to the difference in simple shear at
an interface and a soil-soil shearing the influencing
The fourth order tensor A is expressed as: parameter is the surface roughness. Especially for
fine-grained soils, the roughness is important. If this
roughness exceed an critical value the shear failure
localise in the soil specimen. Chen, Zhang, Xiao, & Li
(2015) demonstrated, if the critical surface roughness
is exceed the interfacial friction angle is equal to the
The collapse behaviour is controlled by an additional
soil-soil (internal) friction angle (Uesugi & Kishida
factor fu as:
1986). Due to this the roughness is for the modelling
of interfaces crucial.
In hypoplastic interface models this is done e.g.
Arnold & Herle (2006) by introducing a reference
353
Table 1. Parameters used for the hypoplastic thermo-
mechanical interface model.
Parameter Soil 1 CV CNL
ϕ c [◦ ] 27.5 27.5 27.5

λ∗ 0.09 0.09 0.09
κ∗ 0.01 0.02 0.04
N 0.88 0.82 0.82
r 0.2 0.2 0.2
αs 3.5 · 10−5 3.5 · 10−5 3.5 · 10−5
lt 0 var. var.
nt −0.01 var. var.
m 2.5 2.5 2.5
e0 0.5 0.45 0.45
σ0 300 300 100
T var. 40 40
Figure 1. τx − γx results for CNL simulation with different

parameter for the roughness κr . Stutz & Mašín (2015) applied constant temperatures.
choose a different approach, by modifying r which is
accounting for the shear stiffness. In clay hypoplastic-
ity, the value of r is equal to (Mašín 2013):
The value of r for the reduced shear stiffness (denoted

as rr ) reads
where ν controls the proportion of shear and bulk

stiffness and used here as a constant ν = 0.2.
3 SIMULATION OF INTERFACE FRICTION

BEHAVIOUR
Typically, the behaviour of interfaces is tested under Figure 2. εn − γx results for CNL simulation with different
different conditions. In this paper two of these bound- applied constant temperatures.
ary conditions are examined. First, the Constant–
Normal–Load condition (CNL) defined as σ˙n = 0,
ε˙n = 0. Secondly, the Constant-Volume condition
(CV) which is defined as σ̇n = 0 and ε̇n = 0.
The availability of limited number of experimen-
tal tests related to Constant-Normal-Load tests are
used in conjunction to model the effects by the new
thermo-mechanical interface model. Using a generic
set of parameters given in Table 1. The aim of this
paper is to demonstrate the application possibilities
instead of comparing measurement against experi-
mental data. The parameters given in the Table 1 are
artificial parameters for the evaluation of the mod-
els response. The reference temperature is 25◦ C. The
results of the CNL simulation are given in Figure 1
and 2 using the parameters of Soil 1. The applied nor-
mal stress is σ0 = 300 kPa. The shear stress decreases
slightly under increasing temperature, see Figure 1.
Whereas, the normal strain εn results (see Figure 2)
indicate an increasing normal strain εn by an increas- Figure 3. τx − γx results for CV simulation with different
ing temperature. For modelling a different behaviour as applied constant temperatures.
354
Figure 4. σn − γx results for CV simulation with different Figure 6. τx − γx results for CV simulation with parameter
applied constant temperatures. variation of nt at 40◦ C.
Figure 5. σn − γx results for CV simulation with parameter

variation of nt at 40◦ C. Figure 7. σn − γx results for CV simulation with parameter
variation of lt at 40◦ C.
indicated by the experimental results from Di Donna,

Ferrari, & Laloui (2015) another parameter set can be
used, see Section 3.1.
The behaviour using a Constant Volume boundary
condition is illustrated in Figure 3 and 4. Decreasing
shear stresses τx and normal stresses σn are the results
by increasing temperature.
In the next section the parameter variation is con-
ducted to estimate the influence and behaviour of
different model parameter which modify the thermo-
mechanical behaviour of interfaces.
3.1 Parameter variation

Figure 5–8 show the results for the parameter vari-
ation of nt and lt under CV conditions. Those two
parameters are the most important ones for modelling
the thermo-mechanical interface response using the
model proposed by Mašín & Khalili (2012) under con-
stant temperature. The stress paths shown, are only Figure 8. σn − γx results for CV simulation with parameter
monotonic shear and temperature tests. variation of lt at 40◦ C.
355
Figure 9. τx − γx results for CNL simulation with parameter Figure 12. εn − γx results for CNL simulation with param-
variation of nt at 40◦ C. eter variation of lt at 40◦ C.
The parameter study is conducted under constant

temperature of 40◦ C and a normal stress of 300 kPa.
In all diverse figures it is observed that a positive sign
of lt and nt will lead to a decrease of shear and nor-
mal stresses. Whereas, an negative sign will lead to an
increasing shear stress and normal stress.
Figure 9–12 show the results for the CNL boundary
conditions. The simulated tests are conducted under
100kPa. As indicated for the CV conditions the CNL
results indicate the same behaviour than the CV results.
In all diverse figures it is observed that a positive
sign will lead to a decreasing shear stress τ and nor-
mal stress σn . Whereas, a negative sign will invert the
model behaviour.
Figure 10. εn − γx results for CNL simulation with param-

eter variation of nt at 40◦ C.
4 CONCLUSIONS
In this paper we present an interface constitutive model

which can take into account isothermal conditions.
Further the model is capable to model non-isothermal
behaviour. For this purpose a hypoplastic formulation
of Mašín & Khalili (2011) and Mašín & Khalili (2012)
is used and the standard tensorial notation was pre-
served. By the use of reduced stress and strain rate
vectors in combination with modified tensorial nota-
tions described in Stutz & Mašín (2015) Stutz, Mašín,
& Wuttke (2016) the interface behaviour is modelled.
The model can take into account all relevant influ-
ences which occur at an interface under thermal and
mechanical loading. This is proven by the calcula-
tions of Constant-Normal-Load and Constant-Volume
boundary conditions. The utilization of this model
will contribute to the modelling of soil-solid interface
in various conditions which are subjected to ther-
Figure 11. τx − γx results for CNL simulation with param- mal and mechanical loading even for repeated loading
eter variation of lt at 40◦ C. cases.
356
REFERENCES Mašín, D. & N. Khalili (2012). A thermo-mechanical model
for variably saturated soils based on hypoplasticity. Inter-
Arnold, M. & I. Herle (2006). Hypoplastic description of the national Journal for Numerical and Analytical Methods
frictional behaviour of contacts. Numerical Methods in in Geomechanics 36, 1461–1485.
Geotechnical Engineering, 101–106. Mašín, D. (2005). A hypoplastic constitutive model for
Bodas Freitas, T. M., F. Cruz Silva, & P. J. Bourne-Webb clays. International Journal for Numerical and Analytical
(2013). The response of energy foundations under thermo- Methods in Geomechanics 29, 311–336.
mechanical loading. 18th Intl Conf Soil Mechanics and Mašín, D. (2013). Clay hypoplasticity with explicitly defined
Geotechnical Engineering, 3347–3350. asymptotic states. Acta Geotechnica 8, 481–496.
Chen, X., J. Zhang, Y. Xiao, & J. Li (2015, jan). Effect Stutz, H., D. Mašín, & F. Wuttke (2016). Enhancement of
of Roughness on Shear Behavior of Red Clayconcrete a hypoplastic model for granular soil-structure interface
Interface in Large-scale Direct Shear Tests. Canadian behaviour. Acta Geotechnica, 1–13.
Geotechnical Journal 52, 1122–1135. Stutz, H. & D. Mašín (2016). Hypoplastic contact model for
Di Donna, A., A. Ferrari, & L. Laloui (2015). Experimen- fine-grained soils (under review).
tal investigations of the soil-concrete interface: physi- Uesugi, M. & H. Kishida (1986). Soils and Foundations 26,
cal mechanisms, cyclic mobilisation and behaviour at 139–149.
different temperatures. Can. Geotech. J., 1–44. Xiao, S., M. T. Suleiman, & J. McCartney (2014). Shear
Di Donna, A. & L. Laloui (2015). Numerical analysis Behavior of Silty Soil and Soil-Structure Interface under
of the geotechnical behaviour of energy piles. Interna- Temperature Effects. In Geo-Congress, Number GSP 234,
tional Journal for Numerical and Analytical Methods in pp. 4105–4114.
Geomechanics 39(8), 861–888. Yavari, N., A. M. Tang, J. M. Pereira, & G. Hassen (2016).
Laloui, L., M. Nuth, & L. Vulliet (2006). Experimental Effect of temperature on the shear strength of soils and
and numerical investigations of the behaviour of a heat soil/structure interface. Canadian Geotechnical Journal,
exchanger pile. International Journal for Numerical and 1–33.
Analytical Methods in Geomechanics 30(8), 763–781.
Mašín, D. & N. Khalili (2011). Modelling of thermal effects
in hypoplasticity. Proc. 13th International Conference of
the IACMAG 1(May), 237–245.
357
A novel cavity expansion-based analytical tool

and its potential application for energy pile foundation
Hang Zhou & Gangqiang Kong

College of Civil and Transportation Engineering, Key Laboratory of Ministry of Education for Geomechanics
and Embankment Engineering, Hohai University, Nanjing, China
Hanlong Liu
College of Civil Engineering, Key Laboratory of New Technology for Construction of Cities in Mountain Area
Chongqing University, Chongqing, China
Yuedong Wu & Guowei Li

ABSTRACT: Energy pile gradually adopted in many countries, which is compatible with the principles of
sustainable development. Extensive field and laboratory experiments are being undertaken in order to evaluate
its thermal-mechanical performance. However, there is no sound theoretical method to reveal the fundamental
mechanism of energy pile, particularly in evaluating the pile capacity change due to the temperature change.
This paper develops a potential useful analytical tool for analyzing the energy pile performance and in particular
to quantify the stress changes at the pile-soil interface induced by the temperature change. The analytical tool
is formulated by incorporating Laloui et al ’s Advanced Constitutive Model for Environment Geomechanics-
Thermal effect (ACMEG-T) into the Cavity Expansion Theory (CET), namely Thermal-Cavity Expansion Theory
(TCET). The TCET has potential application in energy pile foundation and may provide a theoretical basis for
developing design methods of energy pile capacity in the future.
Keywords: Energy pile, Cavity Expansion Model, Thermoplasticity, Geo-thermal
1 INTRODUCTION out. Model test is also developed by some authors.

These studies have shed light on the coupled loading
Energy pile is gradually adopted in many countries, conditions and temperature effects on the pile. How-
which is compatible with the principles of sustainable ever, the analysis focused on theoretical calculation
development. It is instrumented with a closed-loop method on bearing capacity of energy pile influenced
heat exchanger in general to transfer heat between by surrounding soils is little.
the surrounding soil and upper building (Laloui et al. In this paper, an analytical tool, namely Thermal-
2006, Brandl, 2006; Bourne-Webb et al. 2009; Stew- Cavity Expansion Theory (TCET), is presented by
art & McCartney 2013; and Ng et al. 2014). It takes incorporating Laloui et al’s (2009) Advanced Con-
advantage of the geothermal energy induced by season stitutive Model for Environment Geomechanics-
changes and reduces the use of fossil fuels. Com- Thermal effect (ACMEG-T) into the Cavity Expansion
pared with conventional GSHP, it is not only more Theory (CET). The constitutive relations of ACMEG-
cost-effective to install due to the dual purposes of T are coupled with the stress equilibrium equation,
the material, but also it may be more efficient heat continuity equation, as well as drained condition dur-
exchangers for high thermal conductivity of concrete. ing cavity expansion to form a group of Partial Dif-
However, the thermal expansion and contraction of the ferential Equations (PDEs). The PDEs are transform
foundation or surrounding soil will led to the potential into Ordinary Differential Equations (ODEs) using
for foundation movements, which may influence the the similarity transformation technique. Unlike the
safety of the building. Therefore, the mechanisms of complex FEM calculations, the ODEs can be easily
thermo-mechanical soil-structure interaction are very solved through the commercial mathematical software
important. Several full-scale case histories are carried with few seconds. The TCET provides a possibility to
359
(2) the stress-strain relations are described by a consti-
tutive model (ACMEG-T)
(3) a consistency condition ensuring the current stress
state stays on the loading surface
(4) a continuity condition for the conservation of mass
of the solid phase
(5) drained conditions (undrained or drained during
cavity expansion process)
3.1 Equilibrium equation

By ignoring the body force, the usual stress equilib-
rium equation can be expressed as:
where p = (σr + kσθ )/(k + 1), q = σr − σθ , uw is the

pore pressure, σr , σθ is the radial and circumferential
stress components respectively, k = 1 and 2 for cylin-
Figure 1. Cavity expansion model. drical and spherical cavity expansion respectively, r is
the radial position of soil particle.
quantify the stress changes at the pile-soil interface of
energy pile foundation due to the temperature change. 3.2 Constitutive equation
2 CAVITY EXPANSION THEORY (CET)
Figure 1 shows a cavity (cylindrical or spherical cav-

ity) with initial radius a0 , which represents the energy
pile cross section, and initial internal pressure σ0 uni-
formly expands to another cavity with radius a under
the condition of temperature changes in infinite ther- where the symbol (◦ ) is used to describe the material
moplasticity soil. The initial temperature of soil is T0 time derivative associated with a given soil parti-
(before expansion)and then it increased to the value of
T . The expansion induced by the increasing of tem- cle, ,e ,
K = Kref (p /pref )n , G = Gref (p /pref )n , Kref and Gref
e
perature is investigated here and the case of cavity
contraction induced by the decreasing of temperature are the reference bulk and shear modulus, w is the
beyond the scope of this paper and will not reported radial velocity of soil particle, M is the slope of the
here The constitutive model for the soil is described by critical line in q − p , α is the material parameter
Laloui et al’s Advanced Constitutive Model for Envi-
ronment Geomechanics-Thermal effect (ACMEG-T). introducing the nonassociative behavior. is the total
Other definitions are identical to the conventional
plastic volume strain, is the plastic volume strain
cavity expansion theory (Zhou et al, 2014, 2015,
2016). induced by the isotropic mechanism, is the plastic
volume strain induced by the deviatoric mechanism.
3 GOVERNING EQUATION FOR CAVITY pref is the value of the effective mean pressure at which
EXPANSION the reference hypoelastic modulus are measured.
The key problem for cavity expansion in thermo- 3.3 Consistency equation
plasticity soil is to find the temperature-dependent
cavity wall pressure-expansion relations. In fact, cav-
ity expansion is initial and boundary value problem,
which is a mixed problem. If the governing equations
combined with the initial and boundary conditions
are given, the problem could be solved. To estab-
lish the governing equations for cavity expansion, the
following five conditions should be considered:
where the derivatives can refer to Laloui et al (2009)
(1) the stress around the cavity wall should satisfy the and will not be reported here. Other parameters are
equilibrium equation defined the same as Laloui et al.
360
p p
Note that the two variables λiso and λdev used here Introducing the transformation relation in Equation
are slightly different from Laloui’s original definition. (12) and note that a reference stress, pr (=1 kPa) is
Since Laloui used the expressions in the form of incre- used for normalizing the stress variables, the above
mental, while the author used the rate form in this equation can be sorted out as Ordinary Differential
p p p p p
paper. Actually, λiso = λiso /dt, λdev = λdev /dt (λiso and Equations (ODE).
p
λdev are Laloui’s original variables). However, it has
no influence on the derivation. 4.1 Boundary conditions
Combining Equations (4) and (5), one can obtains:
Based on the cavity expansion theory, the follow-
ing relations are established at the elastic-plastic (EP)
boundary:
Furthermore, at the cavity wall, one obtains:
3.4 Continuity equation
The above ODEs can be solved through the

function of “ode45” in Matlab and a numerical
solution can be calculated with a few seconds,
where υ is the specific volume of soil. which is more convenient than the finite element
methods (FEM).
3.5 Drained condition

4.2 Transformation of the similarity solution into
• For drained condition, one has: finite cavity expansion solution
Note that a finite cavity expansion (see Figure 1) is use-
ful particularly in the interpretation of the energy pile
• For undrained condition, one obtains problem. Therefore, the obtained similarity solution
should be changed to the finite cavity expansion solu-
tion. In fact, this problem has been pointed by Collins
and Yu (1996) that the response for the expansion of a
cavity from the finite initial radius a0 is consistent with
that for the soil in the region r ≥ a0 in the created cav-
3.6 Relations of three strain components ity problem (a cavity expands from a zero initial radius
cavity to a finite radius cavity ). Under this principle,
Another equation is about the three strain components the following relation is established:
realtions:
where η and w̃(η) can be obtained from the similarity

4 SOLUTION OF THE GOVERNING solution.
EQUATION Equations (15) actually represent cavity expansion-
plastic zone radius relations. With this relation, the
Based on the similarity technique proposed by Collins solution for the finite cavity expansion problem can
et al (1994), the above equations can be presented be completely determined.
in a non-dimensional form. The following variable
transformation is used as:
4.3 Critical value of the cavity expansion radius at
the start of the yield
It is necessary to give the critical value of the cavity
expansion radius for determining the surrounding soil
where R is the radius of the plastic zone at time t, W whether is yield or not. It is known that there is no vol-
is the cavity wall expansion velocity. ume change in the elastic zone surrounding the cavity.
361
In this case, the cavity wall circumferential strain can
be expressed as (Cao et al., 2001):
where εθ,a is the cavity wall circumferential strain, a0

is the radius of the initial cavity wall, a is the radius of
the cavity wall after expansion.
With the condition of zero volume change, one has:
Note that:
Combining Equations (16), (17) and (18) leads to:
At the start of the yield state, the stress variable

pb d
q is equal to qb = Mpb 1 − b ln p (pb = p0 ) and

n e c,b
p
G = Gb = Gref p b . Thus, the critical value for
ref
cavity expansion radius ac can be obtained as:
If a < ac , the cavity expansion is in the elastic state or

else it is in the plastic state. Thus, the complete cavity
expansion relations can be summarized as:
• if a ≤ ac Figure 2. Relations between normalized cavity wall radius
with the undrained expansion-induced (a) radial cavity wall
pressure, and (b) cavity wall excess pore pressure.
are selected for analyzing the temperature influence

• if a > ac
on the cavity wall pressure-expansion relations. In
addition, both of undrained and drained conditions
are considered for the cavity expansion. Additionally,
only cylindrical cavity expansion (k = 1) is considered
here though it is simple to obtain the spherical cavity
expansion solution (k = 2).
5 NUMERICAL EXAMPLES
5.1 Undrained cylindrical cavity expansion
In this section, the stress mechanism of the cav-
ity expansion in thermoplasticity soil is investigated First of all, the stress mechanism of the undrained
through the proposed similarity solutions. The soil cylindrical cavity expansion is presented. The stress
parameters used in the analysis is used the same as distribution and the cavity wall pressure are very
Laloui et al (2009) and will not reported here. The ini- important to the energy pile and thus are provided in
tial temperature of soil is 25◦ . Five typical increased this section. Figure 2a and 2b plot the relations between
value of temperature (60◦ C, 70◦ C, 80◦ C, 90◦ C, 100◦ C) normalized cavity wall radius with the expansion-
362
very interested to be noted that the increased temper-
ature leads to the decreasing of the value of the cavity
wall pressure. In fact, the increasing of the tempera-
ture results in the decreasing of the preconsolidation
pressure pc . This actually indicates the overconsolida-
tion ratio is gradually increased due to the temperature
increasing. It is known from the conventional cavity
expansion in MCC model that the cavity wall excess
pressure may decrease with the increasing of the value
of OCR under the small soil rigidity index condition.
Therefore, this is reason for the decreased cavity wall
pressure due to the temperature increased.
Figure 3a and 3b show the radial effective stress and
excess pore pressure distribution around the expand-
ing cavity wall for undrained condition respectively.
The normalized cavity wall radius is selected for 1.1
in the computation. It is seen that plastic zone and
elastic zone exist around the cavity wall. The plastic
zone is near the cavity wall, while the elastic zone
is far from the cavity wall. The radial effective stress
rapidly decreases with the increased normalized radial
distance, while it changes slowly in the elastic zone. It
is noted that higher value of temperature develops the
higher value of the radius of the plastic zone. This is
because the soil rigidity index (G/su ) increases with the
increasing of the soil temperature. However, the radius
of the plastic zone is not sensitive to the temperature
changes. Additionally, the excess pore pressure rapidly
vanishes in the plastic zone and the excess pore pres-
sure in the elastic zone is zero, which is consistent with
the conventional cavity expansion theory.
5.2 Drained cylindrical cavity expansion

Besides the undrained cylindrical cavity expansion,
the drained cylindrical cavity expansion is also inves-
tigated in this paper. Figure 4 plots the relations
between normalized cavity wall radius with the
drained expansion-induced radial cavity wall pressure.
Here, it is note that the cavity wall pressure is actually
equal to the radial cavity wall effective stress since the
pore pressure is zero during the drained cavity expan-
Figure 3. Figure 3 Stress distribution around the expanding sion process. Unlike the undrained cylindrical cavity
cavity wall for undrained condition (a) radial effective stress; expansion, the cavity wall pressure is not sensitive to
(b) excess pore pressure. the temperature changes though it slightly increases
with the decreasing of the soil temperature. In addi-
tion, it is also found that the drained cylindrical cavity
expansion in thermoplasticity soil needs larger value
induced radial cavity wall pressure and cavity wall of normalized cavity wall radius than the undrained
excess pressure respectively. It is obviously found that cylindrical cavity expansion to reach the limit state.
the temperature has significant influence on both of the Figure 5a shows the radial effective stress around
cavity wall pressure and cavity wall excess pore pres- the expanding cavity wall for drained condition respec-
sure. Furthermore, the cavity wall pressure increases tively. The normalized cavity wall radius is also
with the increasing of the temperature value and it selected as 1.1. It can be seen that the radial effec-
reaches the limit value when the normalized cavity tive stress is not sensitive to the temperature changes
wall radius a/a0 is close to the value of 5. In prac- which is not identical to the undrained cylindrical cav-
tice, the normalized cavity wall radius a/a0 for energy ity expansion. However, the radius of the plastic zone
pile is very small since the temperature change results is also not sensitive to the soil temperature changes. In
in slightly pile shaft expansion and thus the value of addition, the specific volume distributions around the
the contact pressure at the energy pile-soil interface expanding cavity wall for drained condition are also
may usually not reach the limit value. In addition, it is presented in Figure 5b. The initial specific volume of
363
Figure 4. Relations between normalized cavity wall radius
with the drained expansion-induced radial cavity wall
pressure.
soil is chosen as 2. It can be found that the specific

volume of soil in the elastic zone is not changed and
equal to 2 since the mean effective stress is constant
value during cavity expansion process. Furthermore,
the specific volume of soil increases with the increas-
ing of the normalized radial distance in the plastic zone
until it reaches the value of 2 at the elastic-plastic
boundary. It is interested to be noted that the spe-
cific volume increases with the increasing of the soil
temperature near the cavity wall, while it decreases
with the increasing of the soil temperature near the
elastic-plastic boundary.
Figure 5. (a) radial effective stress and (b) specific volume
distributions around the expanding cavity wall for drained
6 CONCLUSION condition.
This paper incorporates the thermoplasticity soil

model (ACMEG-T) into the cavity expansion theoret- ACKNOWLEDGEMENTS
ical framework to form a new analytical tool, namely
TCET, which may have potential useful in analyzing The work is supported by the National Natural Sci-
the energy pile performance, particularly in capturing ence Foundation of China (No. 51378178, 51306080),
the contact pressure changes at the pile-soil inter- and the Doctoral Program of Higher Education
face induced by the temperature changes. Since no and Research Grants Council Earmarked Research
analytical solutions for the cavity expansion in the Grants Joint Research (No. 20130094140001, M-
thermoplasticity soil have been published previously, HKUST603/13, GRF617213).
the present analysis serves the important purpose of
providing a set of benchmark results that could under-
pin further more sophisticated analysis. The present REFERENCES
solution provides an exact and realistic theoretical
Bourne-Webb P J, Amatya B L, Soga K, et al. (2009). Energy
framework for revealing the influence of the temper- pile test at Lambeth College, London: geotechnical and
ature on the cavity expansion mechanism, and it also thermodynamic aspects of pile response to heat cycles.
provides a valuable benchmark for validating numer- Géotechnique, 59(3): 237–248.
ous cavity expansion numerical methods involving the Brandl, H. (2006). Energy foundation and other thermo-
thermoplasticity model. active ground structures. Géotechnique, 56(2): 81–122.
364
Cao, L. F., Teh, C. I., & Chang, M. F. (2001). Undrained cavity Stewart, M.A. and McCartney, J.S. (2013). Centrifuge Mod-
expansion in modified Cam clay I: theoretical analysis. eling of Soil-Structure Interaction in Energy Founda-
Géotechnique 51(4) 323–334. tions. Journal of Geotechnical and Geoenvironmental
Collins, I. F., & Stimpson, J. R. (1994). Similarity solu- Engineering, ASCE, 140(4), 04013044-1-11.
tions for drained a undrained cavity expansions in soils. Zhou, H., Liu, H. L., Kong, G. Q., & Huang, X. (2014a). Ana-
Geotechnique, 44(1), 21–34. lytical solution of undrained cylindrical cavity expansion
Collins, I. F., &Yu, H. S. (1996). Undrained cavity expansions in saturated soil under anisotropic initial stress. Computers
in critical state soils. International Journal for Numerical and Geotechnics, 55,232–239.
and Analytical Methods in Geomechanics, 20(7), 489– Zhou, H., Liu, H. L., Kong, G. Q., & Cao, Z. H. (2014b). Ana-
516. lytical solution for pressure-controlled elliptical cavity
Laloui L, Nuth M,Vulliet L. (2006). Experimental and numer- expansion in elastic-perfectly plastic soil. Geotechnique
ical investigations of the behavior of a heat exchanger pile. letters 4(April-June), 72–78.
International Journal For Numerical Analytical Methods Zhou, H., Liu, H. L., & Kong, G. Q. (2014c). Influence of
Geomechanics, 30: 763–781. shear stress on cylindrical cavity expansion in undrained
Laloui, L., & François, B. (2009). ACMEG-T: soil ther- elastic-perfectly plastic soil. Geotechnique letters 4(July-
moplasticity model. Journal of Engineering Mechanics, September), 203–210.
ASCE, 135(9), 932–944. Zhou, H., Kong, G., Li, P., & Liu, H. (2015). Flat cav-
Liu, H., Zhou, H., & Kong, G., (2016). Upper-bound solu- ity expansion: theoretical model and application to the
tion for the flat cavity expansion model. Journal of interpretation of the flat dilatometer test. Journal of
Engineering Mechanics (accepted). Engineering Mechanics, 142(1), 04015058.
Ng, C. W. W., Shi, C., Gunawan, A., and Laloui, L. (2014). Zhou, H., Kong, G., & Liu, H. (2016). A semi-analytical solu-
Centrifuge modelling of energy piles subjected to heating tion for cylindrical cavity expansion in elastic-perfectly
and cooling cycles in clay. Géotechnique Letters, 4(4): plastic soil under biaxial in-situ stress field. Geotechnique
310–316. (accepted).
365
Numerical analysis of heat conduction in granular geo-material

using lattice element method
Z.H. Rizvi, A.S. Sattari & F. Wuttke

Geomechanics & Geotechnics, Christian Albrechts University, Kiel, Germany
ABSTRACT: Heat transfer in geo-materials especially in granular systems is important to a vast array of
practical problems, yet is poorly understood even in the simplest case, such as conduction in dry state. This is
due in part to the stress and contact heterogeneities inherent to these systems. Heat conduction in a packet bed
is analyzed computationally based on lattice element method.
A novel model is developed based on the Lattice Element Method, which not only sheds light on fundamental
issues in heat conduction in particles, but also provides a valuable test bed for existing theories. By meshing the
domain the material heterogeneity is directly included, and thus dynamic temperature distributions are obtained
at the particle level. Comparison with existing experiments shows that this model yields a quantitatively accurate
temperature field without the need for adjustable parameters or detailed microstructural information. This simple
system is suitable for providing insight into such phenomena as reactor and underground energy and steam cable
“hot spot” formation and spontaneous combustion of bulk reactive materials.
1 INTRODUCTION based approach is applied to study heat transfer in open

cell foam structure. (Randrianalisoa et al. 2015).
The importance of heat transfer in granular materials In this paper we present an alternative approach
has been brought up in many geotechnical engineer- for heat conduction in particulate matter based on lat-
ing applications such as, geothermal energy storage, tice element method. This method involves Voronoi
buried power cables, containment of radioactive waste, based techniques to divide the particle packing. The
CO2 sequencing and hydrocarbon energy recovery. domain is seeded with nodes and then lattice network is
Traditionally, effective medium analogy is applied, obtained by connecting these nodes with 1D elements.
assuming the porous media is homogeneous. Ther- These 1D elements connected in a three dimensional
mal analysis provides accurate results in bounded case space thus reduces the 3D problem to 1D heat conduc-
with finite element method (FEM) and with bound- tion. The method is applied to study heat conduction
ary element method (BEM) in unbounded case. These in vacuum. The heat transfer path is considered in
methods, based on established continuum theory are the present modelling with conduction path depend-
high computational efficient for large scale problems. ing upon the particle size. The results indicates that
However, these methods cannot take in account for the Voronoi based approach is useful in understanding
granular variation at spatial and particle level. Even, the mechanism of heat transfer in granular particles.
the heat conduction in solid phase presents problem in
transient case.
Many studies for heat transfer based on particle 2 LATTICE ELEMENT METHOD
method have been proposed by researchers. Thermal
particle dynamics method based on stress and con- Behavior of geomaterials depends upon the proper-
tact heterogeneities has considered uniform particle ties and distribution of grains and binders. Therefore,
bed. (Vargas & McCarthy 2001). Discrete element morphology based simulations which accounts for
modelling with circular particles, pipe network model material geometry itself is a growing trend. Vari-
for transient analysis (Feng et al. 2008,2009) three ous numerical models relying upon discrete structural
dimensional random network model considering con- components have been used to clarify crack nucle-
tact resistance (Yun & Evans 2010) multi scale model ation and propagation within cement based binders.
(Zhang et al. 2011), thermal contact model considering (Schlangen & Garboczi 1997). Voronoi based tech-
surface roughness (Zhou et al. 2012) have been stud- niques for discretizing particulate materials in three
ied. However all of these models are considering the dimension within the lattice model for fracture anal-
particulate matter as circular which is far from reality ysis is implied using randomly generated polyhedral.
in case of geo-materials. Recently, Laguere-Voronoi (Asahina & Bolander 2011)
367
Lattice element method (LEM) is adopted from 3.2 Heat Transfer Formulation
condensed matter physics as a method of discretiz-
A lattice based contact conductance model is devel-
ing continuum elastic media and are frequently used
oped to simulate the heat transfer in granular assem-
to simulate deformation and fracture. A LEM con-
blies in vacuum with consideration of the thermal
sists of a regular two or three dimensional network
resistance of smooth contact surfaces. The representa-
of a dimensional springs, connected to randomly dis-
tion of heat transfer between spherical particles using
tributed nodal sites, with each spring is govern by its
DEM can be found in various scientific works (Feng
own set of constitutive relations. Two type of spring
et al. 2008, 2009). The discrete thermal element mod-
models are used 1) Hookean spring model and 2)
elling of heat conduction in spherical particle systems
Born spring model. The first model exhibit central
using Pipe-network model is introduced by (Feng et
force interaction with constant Poisson’s ratio while
al. 2008, 2009). Generally, the temperature Ti of each
the second introduces non-central two body interac-
particle in the steady system should satisfy the follow-
tion limiting the rotation freedom. A number of studies
ing heat conduction equation. Under this condition,
used beam-spring network thus enriching the material
the heat flows into and out of the RVE are the same.
description with rotation degree of freedom. Lattice
type models have been applied by Schorn using truss
element, which require numerical measures to avoid
instability. Bolander applied a beam lattice with a step
wise softening law which was extended to 3D by Lilliu
and termed is as “Delft lattice”. A parallel computation In which R denotes the granular RVE and qi represents
scheme is implemented to reduce the computational the total heat exchange between particles i and others.
time and an interfacial transition zone (ITZ) is imple- Calculating qi requires heat exchange qji between two
mented to reduce the matrix strength. The majority contact particles, which is given by
of particulate models have been two dimensional with
circular and fiber inclusions with a few spherical inclu-
sions is investigated in a 3D simulation. Various failure
criteria of element failure have been suggested includ-
Where hij denotes the heat conductance between the
ing critical elastic stress, strain and energy. LEM is
two particles, Ti and Tj the temperatures of parti-
mostly associated with linear elastic fracture mechan-
cles i and j, respectively. The analytical solution of
ics with a few attempts to include plasticity. Plasticity
heat conductance hij with perfectly smooth surfaces is
is assigned to the springs which results in anisotropic
presented by (Bahrami et al. 2005).
plastic deformations which in turns fails to account for
conservation of volume.
3 REPRESENTATION OF HEAT TRANSFER

WITH LATTICE ELEMENT Where Fn is the magnitude of normal contact force
between the two contact particles, and k is the thermal
An in-house Lattice based code in under development conductivity. re and Ee are the effective radius and the
to solve the heat transfer problem in steady and tran- effective Young’s modulus, which are determined by,
sient state. The solver can handle complex geometry
and yield results in reasonable time.
3.1 Mesh Generation

In the 2D and 3D model Voronoi polygons are used
for the representation of the structured system. We
apply a special form of the Voronoi tessellation called
vectorizable random lattice (Lilliua-Mier 2003 &
Moukarzel-Herrmann 1992), in order to obtain an ini-
tial configuration of the mesh. It is performed by first
putting a regular grid onto the plane and throwing
points randomly and independently in a square of the
side length a centered on the plaquettes of the regular
grid. Using these points for the Voronoi construction,
the randomness of the tessellation can be controlled by
tuning the value of the parameter a between 0 and the
lattice spacing of the grid. After Voronoi construction,
a customized Delaunay discretization code is devel-
oped based on the neighboring cells, accounting for Figure 1. Representation of Voronoi cell and effective
lattice element springs or beams. distance between two nuclei.
368
Where ν is Poisson ratio. Application of Lattice Ele- i such that each heat contribution may be calculated
ment into discrete element problem can provide bet- from Eq. 2 and the total heat input Qi may be approx-
ter representation of heat transfer between irregular imated as the sum of the interactions of particle i with
shaped particles. In this case ri and rj should be each of its neighbors. The assumption is not correct
redefined as shown in figure (1). if the heat storage term is incorporated. However it
To scale up the heat transfer equation requires two is conceptually simple and is found to be accurate
conditions. The first assumption is that each i − j par- in steady state condition. It is also assumed that heat
ticle contact “feels” the same temperature for particle transfer is possible through the network formed from
connecting the nuclei of each Voronoi cell. The first
two assumptions are independent of material property.
The material parameter is introduced from the conduc-
tance between the two adjacent nuclei depends upon
the re and Ee are the effective radius and the effec-
tive Young’s modulus Fn is the magnitude of normal
contact force between the two contact particles. The
second condition is that the temperature change is a
gradual process and only the immediate cell can expe-
rience the change in one time step. Mathematically,
this quasi-steady temperature criterion can be shown
to be met by choosing a time-step sufficiently small.
4 NUMERICAL TESTS AND DISCUSSION
Three numerical tests are performed based upon the

developed model with regular and irregular packing
Figure 2. Voronoi cell and Denulay triangulation of a 2D of the granular assembly. These simulations are per-
sample (a)(b)(c) and 3D(d)(e)(f) with different randomness formed with 2D and 3D arrangement of particles.
factor (a) r = 0.01 (b) r = 0.4 (c) r = 0.7. Three different packing arrangements are simulated,
Figure 3. 2D conduction heat transfer in granular material with particle shape factor of 20x20 mesh size a.) 0.01 b.) 0.4 c.)0.7
d.) 0.9.
369
Figure 4. Mesh density test with different number of elements in a square plate [1, 1] a) 400 b) 2500 c) 10000 d) 22500.
Figure 6. Mesh density test with different number of

elements in a prism block [1, 1, 3] a) 10 × 10 × 30 b)
20 × 20 × 40 c) 30 × 30 × 30.
Figure 5. Mesh density test with different number of ele-
ments in a cube block [1, 1, 1] a) 15 × 15 × 15 b) 30 × 30 × 30
c) 40 × 40 × 40. granular materials are dominated by conduction and
the flow path (Yun & Santamarina 2008) is defined
the first one with a randomness of 0.01 which resem- by conduction path in granular material. Many parti-
bles a regular lattice case. The second simulation is cle based approaches to solve the problem had been
performed with a randomness of 0.4 and the last one proposed. Some of these approaches such as based of
with a randomness of 0.7. The randomness factor con- Discrete Element Method are limited with the shape
trols the construction of Voronoi cell generation. The and arrangement of particle. The other method based
figure below snows the generated Voronoi and denulay on molecular computation is computational inten-
triangulation diagrams for the considered region. sive and requires the physical parameter calculation
at molecular level. As the connections are defined
between Voronoi cell and a heat conduction paths are
4.1 Illustrative examples
developed which defines the heat conduction path,
The effectiveness of the LEM heat transfer code the method gives better picture of heat conduction in
is demonstrated with examples. Heat conduction in granular materials with limited computation power.
370
The problem of heat conduction in granular mate- REFERENCES
rial is presented. The Voronoi cells are generated in
a range of [1, 1] square geometry. Randomness fac- Asahina D., Bolander J.E.,”Voronoi-based discretizations for
tors as mentioned above of 0.01, 0.4, 0.7 and 0.9 are fracture analysis of particulate materials” Volume 213,
Issues 1–3, 10 November 2011, Pages 92–99.
applied. The number of Voronoi cells are increased Bahrami M., Culham J. R., Yovanovich M. M., Modeling
from 20, 50 and 100 in each directions. Thermal Contact Resistance: A Scale Analysis Approach,
The code is extended to simulate heat transfer in 3D J. Heat Transfer 126(6), 896–905 (Jan 26, 2005).
granular media. 1D elements are chosen and the same Feng YT, Han K, Li CF, Owen DRJ (2008) Discrete thermal
scheme to discretize and the method for generation of element modelling of heat conduction in particle systems:
Vornoi cells are adopted.Temperature boundary condi- basic formulations. J Comput Phys 227(10):5072–5089.
tions are applied at two opposite faces and the rest were FengYT, Han K, Owen DRJ (2009) Discrete thermal element
assumed to be isolated. Two different shapes, cube and modelling of heat conduction in particle systems: pipe-
prism were chosen for steady state heat transfer. The network model and transient analysis. Powder Technol
193(3):248–256.
Voronoi particle randomness factor for these simula- Lilliua G, van Mier J.G.M “3D lattice type fracture model
tions are fixed at 0.7. The conductance between two for concrete” Volume 70, Issues 7–8, May 2003, Pages
particles is calculated from equation 3. 927–941.
Moukarzel, C. & Herrmann H.J. A Vectorizable random
lattice, journal of statistical physics Vol,68.Nos.5/6 1992.
5 CONCLUDING REMARKS Randrianalisoa J., Baillis D., Martin C.L., Dendievel R.,
Microstructure effect on thermal conductivity of open cell
The present work has developed a lattice based foams generated from the Laguere Voronoi tessellation
approach for thermal conduction in granular material. method, International Journal of Thermal Sciences 98
(2015) 277–286.
Each particle is modelled as a lattice which is connect Schlangen E., Garboczi E.J. “Fracture simulations of concrete
by a simple 1D element connecting the Voronoi cell using lattice models: Computational aspects” Volume 57,
nucleolus with surrounding cells using Delaunay tri- Issues 2–3, May–June 1997, Pages 319–332.
angulation scheme. The model essentially neglects the Vargas, W.L. & McCarthy, J.J. 2001. Heat conduction in
heat storage term in the cells and thus simplifies the Granular Materials AIChE Journal Vol 47, No. 5.
solution procedure. Yun, T.; Evans, T. Three-dimensional random network model
In comparison with the other particle methods for thermal conductivity in particulate materials. Comput.
i.e. Discrete Thermal Element Method, this lattice Geotech. 2010, 37, 991–998.
based approach predefines nucleolus and lattice con- Yun, T.; Santamarina, J. Fundamental study of thermal con-
duction in dry soils. Granul. Matter, 2008, 10, 197–207.
nections. The algorithm is compatible with classical Zhang H.W., Zhou Q., Xing H.L., Muhlhaus H., A DEM
discrete element method which could be used to simu- study on the effective thermal conductivity of granular
late thermo-mechanical problem. The method could assemblies, Powder Technol. 205 (2011) 172–183.
be extended to simulate the transient heat transfer Zhou Q.,. Zhang H.W., Zheng Y.G., A homogenization tech-
in granular materials with minor changes. Numeri- nique for heat transfer in periodic granular materials,
cal simulations suggest that lattice based heat transfer Advanced Powder Technology 23 (2012) 104–114.
method gives accurate and fast results for heat trans-
fer in conduction for steady state problems. A detailed
study for heat transfer in conduction based on lattice
method is required for transient problems.
ACKNOWLEDGEMENT
This research project is financially supported by the

research grant DuoFill provided by the Federal Min-
istry for Economic Affairs and Energy, Germany
(ZIM grant number KF3067303KI3) and Federal state
funding at University of Kiel.
371
Thermo-hydraulic coupled modelling: Analysis of thermal enhanced

backfill structures and materials for near-shore power cables
K. Sembdner
Department of Geosciences, Kiel University, Kiel, Germany
H. Anbergen
APS Antriebs-, Prüf- und Steuertechnik GmbH, Rosdorf, Germany
F. Wuttke
Department of Geosciences, Kiel University, Kiel, Germany
ABSTRACT: Offshore wind power has been an expanding market during the last decade. The wind power
stations are electrically connected via buried high voltage cable systems. Due to the electrical resistance of cables,
thermal energy is emitted during current transport. The emitted heat damages and ages the cable, and weaken
significantly the cable’s performance. A thermal enhanced backfill embedding the cable helps distributing the
produced heat fast to the surrounding soil and protect the cable from excessive heating. Thermo-hydraulic coupled
simulations have been carried out to analyse the influence of thermal enhanced backfill on the temperature
distribution around these high voltage cables. A comparative study with different thermal backfill materials
embedding a high voltage cable was performed, in order to identify suitable thermal backfill properties and its
arrangement around the cable. For this purpose, three different numerical solutions have been used (OpenGeoSys,
COMSOL, FEFLOW). The results show that heat distribution around high voltage cables is strongly influenced
by thermal and hydraulic properties of the surrounding backfill and that a proper backfill is able to modify the
heat dispersion according to requirements.
1 INTRODUCTION (BSH, 2007). The heat protection is particularly con-

centrated to the upper centimeters of the seafloor, since
Offshore wind power has been an expanding mar- it exhibits the highest bioactivity compared to the lower
ket of the last decade. Several offshore wind power depths of the seabed. In summary, the backfill material
plants have been constructed or the construction is needs to provide a sufficient heat transport ability, to
in progress. These plants need to be connected via keep the high voltage cable cool and simultaneously
high voltage cable systems. While the energy trans- comply with the ecological boundary conditions.
port onshore is often solved with over-head systems, Within the framework of a joint research project
offshore and nearshore buried cables are used. Due to (KF3067306KI4) thermo-hydraulic coupled simula-
the electrical resistance of cables, thermal energy is tions have been carried out to analyze the influence of
emitted during electrical energy transport. In order to thermal enhanced backfill on the temperature distri-
be able to transfer a sufficient amount of current and bution around high voltage cables inside the seafloor.
simultaneously protect the cables from heat caused The first fundamental study is a comparative study
damages, the cable’s dimensions are enormous. The on the influence of the backfill’s material parameters
aim of this study is to increase the efficiency of power on the heat transport. For this purpose simulations were
cables and/or reduce their required size, through a performed with different backfill material and model
thermal enhanced backfill embedding the cable. Addi- geometries. The study consists of four scenarios, in
tionally ecological guidelines need to be taken into which the high voltage cable is embedded in different
account to protect the heat-sensitive marine environ- enhanced backfill materials and without any backfill
ment. Therefore, the german environmental authority material. Additionally a hydraulic flow is induced. The
(Bundesamt für Seeschiffahrt und Hydrographie) has temperature distribution around the cables is recorded
prohibited to exceed a temperature rise of 2 K at a for 168 hours at a maximum cable load, as a worst-case
seafloor depth of 0.2 m (inside german’s exclusive scenario. The simulations are performed by three finite
economic zone) and 0.3 m (inside Lower Saxony’s wet- element method (FEM) software solutions, namely:
lands), this prohibition is called the 2-Kelvin-Criteria OpenGeoSys, COMSOL, FEFLOW.
373
Table 1. Mesh parameters of the three different software
solutions (for Model A).
OpenGeoSys Comsol FEFLOW
Mesh Nodes 4211 14103 13059

Mesh Elements 8373 28385 24444
Element Type Triangular Triangular Triangular
2 METHODOLOGY
2.1 Software applications and numerical

calculations
The three software solutions utilised are: Open-
GeoSys, COMSOL, FEFLOW. They are all finite
element method (FEM) based software solutions. The
mesh parameters used by the three different software
solutions are listed in table 1.
For solving the groundwater flow coupled with
heat transport, the following thermo-hydraulic cou- Figure 1. Schematic model configuration, including bound-
pled equations were used (Rühaak et al., 2015; Mot- ary conditions and material parameter (after Anbergen et al.,
taghy & Rath, 2006; McKenzie et al., 2007; Rühaak, 2015b).
et al. 2008; Anbergen et al., 2015a):
where λw , λs , and λeff are the thermal conductivi-
ties of water, soil solids, and the resulting bulk value,
respectively; and ε is the porosity.
The bulk volumetric heat capacity (ρc)eff (J m−3 K −1 )
is calcultated in the similar manner:
with
where ρ is the density, c the heat capacity, ε the poros-

ity. The indexes w, s and eff indicate the water and soil
Nomenclature phase, respectively the bulk value.
Ss specific storage (m−1 )
ρ density (kg m-3) 2.2 Model configuration
c specific heat capacity (J kg−1 K−1 )
T temperature (◦ C) Two different model configurations were used in this
λ thermal conductivity (W m−1 K −1 ) study. Their major difference is the trench design.
∇ Laplace operator Model A contains a uniform trench backfill, whereas
model B has a combination of two backfilll material.
H heat sources (W m−1 )
Also, model A was solved by all three different soft-
Q fluid sources (s−1 )
ware applications, while model B was solved only by
K hydraulic conductivity (m s−1 ) Comsol.
h head (m)
t time (s)
2.2.1 Model A
Sw water saturation (–)
The model area is a vertical cross-section through a
L latent heat of fusion for
power cable with the surrounding water and sediments.
bulk medium (J m−3 ) The area of interest is a square of 10 m × 10 m, see
q Darcy velocity (m s−1 ) figure 1. The model is horizontally divided into halves.
Indexes The upper half represents free seawater, the lower half
eff bulk value the saturated seabed sediments. Inside the seabed the
f fluid phase high voltage cable is embedded in a 1.64 m deep and
0.36 m wide trench filled with backfill material.
The FEM code interpolates the bulk thermal con- Due to the legal constraints, a worst-case scenario is
ductivity linearly using porosity, the solid and fluid simulated in order to identify potential hot-spots in the
thermal conductivity (Anbergen et al., 2015a; Diersch, surrounding sediments. Therefore, the outer surface
2014; Comsol, 2013; Böttcher, 2014): of the cable is set to a constant temperature of 60◦ C,
following assumptions of Brakelmann (2010). Brakel-
mann (2010) simulated a worst-case-scenario with a
374
Figure 3. Three model scenarios of Model A: a) trench
filled with backfill_1, b) trench filled with backfill_2, c) no
backfill.
Table 2. Material pameters of water, seafloor and backfill

materials.
Water Seafloor Backfill_1 Backfill_2

Figure 2. Schematic model configuration of the layered
trench, filled with backfill_1 and backfill_2. Heat Cond. (eff.), 0.58 1.43 1.00 2.00
λ[W m−1 K −1 ]
maximum electrical loading for a period of seven days Heat Capacity, 4190 1273 1000 1000
c[J kg −1 K −1 ]
(168 hours). This extensive energy transport resulted Porosity, ε [–] 1.00 0.25 0.30 0.30
in a heating of the outer cable surface of 60◦ C. Hydraulic Cond., – 1 × 10−6 1 × 10−8 1 × 10−8
The temperature boundary conditions of the model k[m s−1 ]
area are set to constant 12◦ C, assuming no tempera- Density, 1000 1800 1800 1800
ture change in the surrounding soil and water. With ρ[kg m−3 ]
12◦ C the boundary condition is set slightly higher
than the mean annual bottom water temperature in the
nearshore areas of the Northsea (Schulz, 2009), hence
a worst case scenario is simulated. parameters of the backfill materials backfill_1 and
In order to simulate the water movement at the sea backfill_2 are almost identical, except for the heat con-
bed, a hydraulic flow from the left to right side of the ductivity of the saturated materials. Backfill_1 has a
model area is induced. Boundary conditions of differ- low effective heat conductivity of 1 W/mK, which is
ent hydraulic heads are implemented (left hydraulic lower than that of the seafloor (1,43 W/mK).
head of 10 m, right hydraulic head of 9 m). Backfill_2 has an effective heat conductivity of
The model includes observationpoints at several 2 W/mK, which is higher than that of the seafloor.
depths under the seabed (0.2, 1 and 2 m). A special
focus is on the depth of 0.2 m, for monitoring the
compliance of the 2-Kelvin-Criteria. 3 RESULTS
2.2.2 Model B 3.1 Model A

This model configuration contains a layered trench, in
which two different kinds of backfill material are com- The heat dispersion around the cable embedded in
bined, see figure 2. Also the depth of the total trench backfill_2 after 168 hours simulated with Comsol is
is increased by 0.36 m to 2 m. All the other already shown in figure 4. The heat distributes in a circular
described model configurations of Model A remain shape around the heat source and expands its radius
unmodified. with increasing timesteps. The results in figure 4 cal-
culated with Comsol stand exemplary for the other
software solutions, which have shown very similar
results for heat dispersion.
2.3 Model Scenarios
For Comparison of the different software solutions
Three model scenarios were simulated for Model A, and estimation of the accuracy of the results, the
see figure 3. The scenarios differ in the backfill mate- temperature at an observationpoint is analysed. The
rial of the trench. In two scenarios the high voltage observationspoint is set at 0.2 m depth perpendicular
cable is embedded in backfill materials with differ- to the cable’s center, for monitoring the compliance
ent thermal properties. In the third scenario the cable of the 2-Kelvin-Criteria. The temperature at this point
is embedded in the seafloor, without any modified is plotted versus time for all three software solutions
backfill. A fourth scenario is calculated for Model (OpenGeoSys, Comsol and FEFLOW) and for the
B; a layered trench combining two different backfill three scenarios: with backfill_1 (a), with backfill_2
materials, as shown in figure 2. (b) and without backfill (c).
The material parameters of the backfill materials, The scenario without backfill is shown in figure 5,
water and seafloor are listed in table 2. The material here the cable is embedded in marine sediment. The
375
Figure 4. Heat dispersion after 168 h with backfill_2,
Comsol.
Figure 6. Heat development at 0.2 m seafloor depth versus
time, with backfill_1.

time, without backfill.
time, with backfill_2.
temperature increases with time for all three soft-
ware applications. The 2-Kelvin-Criteria is slightly In all of the scenarios the Comsol application cal-
exceeded by OpenGeoSys and Comsol results. In both culated the highest temperatures, followed by Open-
cases the temperature increases up to around 14.1◦ C. GeoSys and FEFLOW. However the deviations are
The FEFLOW results show no exceedance of the smaller than 1 ◦ C.
2-Kelvin-Criteria for the calculated time period, the Figure 8 displays Isolines for the temperatures 15◦ C
temperature’s maximum is 13,6◦ C. and 30◦ C around the cables of all three scenarios. It can
The temperature in the scenario b) (see figure 6) is be clearly noticed that the backfill_2 (b) is directing
also increasing with time, but the maximum tempera- the heat flux towards the seabed surface. Additionally
tures are significantly lower compared to figure 5. you can see, that the temperature plume is disturbed by
Figure 7 shows the temperature development with the backfill materials, as their hydraulic conductivity
the heat source embedded in backfill_2. In this sce- is lower than that of the seafloor.
nario the temperature increase is the steepest and
the maximum temperature is the highest for all three 3.2 Model B
software solutions, compared to the other scenarios.
Moreover all of the software solutions exceed the In figure 9 and 10, the temperature at different obser-
2-Kelvin-Criteria. vationspoints for a layered trench and without any
376
Figure 8. Temperature Isolines (15◦ C, 30◦ C) around the
cable for the three scenarios of Model A: a) trench filled with
backfill_1, b) trench filled with backfill_2, c) no backfill.
Figure 10. Heat development at different observationpoints
(0.2, 1.0, 2.0 m beneath the seafloor) versus time, without
backfill, solved with Comsol.
Figure 11. Temperature Isolines (15◦ C, 30◦ C) around the

cable for Model B.
Figure 9. Heat development at different observationpoints

(0.2, 1.0 and 2.0 m beneath the seafloor) versus time, with 4 DISCUSSION
layered backfill, solved with Comsol.
The figures 5–7 demonstrate that the temperature
dispersion is obviously depending on the thermal
trench are displayed. Comparing these two results, it parameters of the backfill.
is obvious that the layered trench leads to a deviant Thermally enhanced backfill in the trench, such
temperature distribution. The upper layer of back- as backfill_2, induces a faster heat transport. Even
fill_1 acts as isolating barrier for the upper part of though this is favorable for the cable, as it leads to
the trench, to a very slight extent. As for the lay- an enhanced cooling of the cable, the legal criteria
ered trench the temperature at the observationpoint (2-Kelvin-Criteria) is exceeded. Since a faster heat
in 0.2 m depth decreases by 0.01◦ C. But it seems, transport is mainly induced in an upwards direction.
that the temperature is accumulating underneath the The backfill_1 might seem to be the best solu-
layer of backfill_1, as the temperature is raised by tion, as it is the only scenario complying with the
0.5◦ C at the observationpoint at 1.0 m depth within the 2-Kelvin-Criteria. But this first evaluation is mis-
layered trench. Additionally the heat dispersion into leading, because the backfill is insulating the cable,
the seafloor depth is strongly enhanced by backfill_2 which conflicts with the need of the thermal energy
underneath the cable. For the temperature at the obser- dispersion.
vationpoint at 2.0 m depth is increased by 4.69◦ C. This It is obvious that the first simulated solutions with
effect of a redirected heat flow into the seafloor depth Model A do not lead to the desired effect. Hence,
is also displayed in figure 11. further simulations were executed, where a layered
377
trench was constructed (Model B). Here, the trench ACKNOWLEDGEMENTS
is filled with two different backfill materials. The
upper part of the trench is filled with an insulating This research project is financially supported by the
backfill (backfilll_1), to ensure a compliance with research grant “GeoMörtel” provided by BMWi (Bun-
the 2-Kelvin-Criteria. The lower part of the trench is desministerium für Wirtschft und Energie) & ZIM
filled with a thermal enhanced backfill (backfill_2), (Zentrales Innovationsprogramm Mittelstand) with the
to ensure a fast thermal energy dispersion. Yet the grant number: KF3067306KI4.
desired effects of the layered trench have only been
achieved in parts. The fast thermal energy dispersion
particularly to deeper areas of the seafloor has been REFERENCES
accomplished. Yet the isolating effect of the upper
backfill layer is much less pronounced than expected. Anbergen, H. Rühaak, W. Frank, J. & Sass, I. 2015a. Numer-
ical simulation of a freeze–thaw testing procedure for
This may depend on the thermal and hydraulic mate- borehole heat exchanger grouts. Canadian Geotechnical
rial parameters of backfill_1. A study with a variation Journal 2015 – doi:10.1139/cgj-2014-0177
of these parameters and the geometry of the layer is Anbergen, H. Sembdner, K. Kremmer, M. Christoffers, T.
adviseable. Metge S. 2015b. Untersuchungen der geothermischen
Comparing the different software applications, cer- Eigenschaften von Bettungsmaterialien für Near-Shore
tain variations in the results are detected, even though Kabeltrassen. Der Geothermie Kongress 2015, Essen
the calculations are executed with the same simulation Brakelmann, H. 2010. Kabelverbindungen innerhalb der
parameter and model parameter input. Offshore-Windfarm Arcadis Ost 1 – Thermische und
These deviations may be based on different numeri- Magnetische Emission. Sachbericht
Böttcher, N. 2014. Thermodynamics of porous media: non-
cal solution algorithms, which are used by the different linear flow processes. Dissertation zur Erlangung des
software applications. In spite of these discrepan- akademischen Grades Doktoringenieur (Dr.-Ing.), TU
cies, the simulation results enable to evaluate at least Dresden
tendencies of the temperature distribution around a BSH, 2007. Standard – Konstruktive Ausführung von
high-voltage cable. Offshore-Windenergieanlagen. Bundesamt für Seeschif-
fahrt und Hydrographie, Stand: 12.07.2007
Comsol 2013. Heat Transfer Module User’s Guide, Comsol
Version 4.4
5 CONCLUSIONS AND OUTLOOK Diersch, H.-J.G. 2014. FEFLOW – Finite Element Modeling
of Flow, mass and heat transport in porous and fractured
The results of the three software solutions differ sub- media. Springer, Berlin, Germany
stantially. More benchmark studies, comparing the McKenzie, J.M. Voss, C.I. Siegel, D.I. 2007. Groundwater
different software applications are needed. flow with energy transport and water-ice phase change:
Numerical simulations, benchmarks,and application to
The temperature dispersion is strongly depending freezing in peat bogs. Advances in Water Resources 2007;
on the thermal parameters of the backfill and its 30:966–983
arrangement. The results show, that a proper backfill Mottaghy, D. Rath, V. 2006. Latent heat effects in sub-
is able to modify the heat distribution around the cable surface heat transport modelling and their impact on
according to its requirements. Besides the simulation palaeotemperature reconstructions. Geophys. J. Int. 2006;
of the temperature distribution, further elaboration of 164:236–245
the absolute heat-flux is needed. More in-depth studies Rühaak, W. Anbergen, H. Grenier, C. McKenzie, J. Kurylyk,
are planned, as for example the simulation of heat flux B.L. Molson, J. Roux, N. Sass, I. 2015. Benchmark-
around the cable. Only in this way the complete effect ing Numerical Freeze/Thaw Models. Energy Procedia
08/2015 -doi:10.1016/j.egypro.2015.07.866
of thermally modified backfill material can be evalu- Rühaak, W. Rath, V. Wolf, A. Clauser, C. 2008. 3D finite
ated. Furthermore accompanying experimental studies volume groundwater and heat transport modeling with
are planned, in order to confirm or rectify the above non-orthogonal grids using a coordinate transformation
presented simulation results. For this, thermal conduc- method. Advances in Water Resources 2008, 31(3):513–
tivity and heat capacity measurements of the modified 524
backfill and sediments will be executed. Addition- Schulz, A. 2009. North Sea Atlas – Temperature, Salin-
ally an experimental setup as technikum experiment is ity, Density and Heat Content – Monthly Means for the
planned, in order to measure the thermal plume around Period 1902 to 195. Bundesamt für Seeschifffahrt und
a cable directly. Hydrographie, Hamburg und Rostock
378
Development of laboratory system for cryogenic fracturing study
M. Cha
N.B. Alqahtani
King Abdulaziz City for Science and Technology (KACST), Riyadh, Saudi Arabia
B. Yao, X. Yin & Y.S. Wu

Colorado School of Mines, Golden, Colorado, USA
T.J. Kneafsey
Lawrence Berkeley National Laboratory, Berkeley, California, USA
ABSTRACT: The concept of cryogenic fracturing is that sharp thermal gradient developed on rock surface by
subjecting cryogenic fluid creates strong local tensile stress and initiates fractures. Prior field tests suggest that
field application with special equipment rated for cryogenic temperatures may bring in potential benefits. They
did not, however, identify the fracture mechanisms at work in downhole conditions. We have developed experi-
mental setups and procedures that are specifically designed to conduct cryogenic fracturing tests with confining
stress, integrated cryogen transport, measurements, and fracture characterization. A true triaxial loading system
is built to simulate reservoir stress levels and anisotropies. Temperature, borehole pressure, and liquid nitrogen
can be monitored continuously. Acoustic and pressure-decay measurements are used to characterize fractures
before and after the experiments. The laboratory design was able to effectively apply cryogenic fracturing to
laboratory rock specimens. The characterization methods were able to capture rock property changes due to
cryogenic fracturing.
1 INTRODUCTION aspects, cryogenic fracturing have the potential to offer

benefits.
Cryogenic fracturing is based on the idea that a cold Cryogenic fracturing using liquid nitrogen can
fluid can induce thermal fractures when brought into achieve extremely low temperatures and very strong
contact with a much warmer rock. When liquid nitro- thermally induced stresses. Grundmann et al. (1998)
gen as a cryogen is injected into a rock with much treated a Devonian shale well with liquid nitrogen and
higher temperature, the heat from the rock will quickly observed an initial production rate 8% higher than the
transfer to the cryogenic fluid. This rapid heat trans- rate in a nearby offset well that had undergone tra-
fer, better known as a thermal shock, will cause the ditional fracturing with nitrogen gas. Unfortunately,
surface of the rock to contract and fail in tension, thus subsequent production information was unavailable
inducing fractures orthogonal to the contact plane of because the well had to be shut in for logistical rea-
the cryogen and rock. sons. To further advance the study of cryogen fluids
The hydraulic fracturing practice uses mainly water- on hydrocarbon producing formations, McDaniel et al.
based fluids, due to the general availability and low (1997) conducted simple laboratory studies where coal
cost; however, a dependence upon water presents a samples were immersed in cryogenic nitrogen. The
number of advantages. Water can cause significant for- coal samples experienced significant shrinkage and
mation damage, which can occur as clay swelling and fracturing into smaller cubicle units, with the creation
relative permeability effects stemming from capillary of microfractures orthogonal to the surface exposed
fluid retention (Mazza, 1997). The formation dam- to the cold fluid. The researchers found that repeated
age mechanisms inhibit hydrocarbon flow and impair exposure cycles to the cryogen caused the coal to break
production rates and recovery efficiency. Also, water into smaller and smaller pieces, or become rubblized.
used in large quantities may place stresses upon the After 3 cycles of exposing the coal to liquid nitrogen
local environments where fracturing activities occur. and allowing the coal to ambient temperatures again,
Finally, downhole injection of chemicals needed in the coal was reduced to grain size particles. McDaniel
water-based fracturing operations, including slickwa- et al. (1997) also conducted field experiments by
ter and gel, are environmentally controversial. In these re-stimulating four coal-bed methane (CBM) wells
381
and one tight sandstone well with liquid nitrogen.
The wells were retrofitted with stainless steel sur-
face piping, manifolds, and wellhead component to
prevent thermal contraction problems. A free hang-
ing fiberglass tubing was used to inject the liquid
nitrogen without compromising the casing integrity.
The results were mixed: All 5 wells showed promis-
ing re-stimulating initial production rates 10–20 times
the before-re-stimulation production average; how-
ever, those rates quickly dwindled. The CBM wells
showed sustained 6 month re-stimulation production
increase of 0–45%. The tight sand well initially had
higher flow rates for 2 months after re-stimulating, Figure 1. Basic schematics of cryogenic fracturing setup
but then had a 65% loss in production from pre re- (the loading device is not shown).
stimulation performance. It is believed that the initial
success in re-stimulating these wells was not that new
fractures were created but that the damage from the
gel filter cake of previous fracturing treatments were
fracturing needs to be different from that for hydraulic
greatly reduced.
fracturing. For cryogenic thermal fracturing, cryo-
In summary, fracturing with liquid nitrogen may
genic fluid needs to keep flowing in order to cool the
be viable for the field and bring with it benefits in
borehole down, because the specimen is much hot-
reducing the formation damage and water and chemi-
ter and stagnant liquid nitrogen will quickly boil and
cal use. However, when it comes to the understanding
vaporize. Compressed nitrogen gas was used to either
of the cryogenic fracture mechanisms, few lab work
push liquid nitrogen into borehole under higher pres-
has been done systematically. Research is required to
sure (“higher-pressure LN flow”) or directly pressurize
better understand the mechanisms of cryogen fractur-
boreholes for breakdown tests. Pressure can be applied
ing process in controlled environment and how we can
as shown on the right-hand side of Figure 1.
integrate it into our current fracturing technology, if
The cryogen transport lines should be able to with-
cryogenic fracturing potential is proven high.
stand cryogenic temperature (down to −196◦ C in our
A proper laboratory design of equipment is impor-
study). Stainless steel 316 and brass generally provide
tant to effectively capture cryogenic fracturing pro-
such an ability as their brittleness-ductility transition
cess. In this study, we developed a laboratory system
temperature is lower than liquid nitrogen boiling point.
for cryogenic fracturing under true triaxial loading
However, stainless steel has higher pressure rating
conditions. It allows to characterize cryogenic frac-
at the temperature. In case of tubing, it needs to be
turing processes in the laboratory in controlled envi-
seamless annealed.
ronments, such as loading conditions and different
Figure 1 shows that LN is flown directly from the
cryogenic fracturing schemes. The devices and pro-
dewar to the specimen borehole under low pressure
cedures are improved and optimized based on our
(∼60–130 kPa). In order to have LN flow under higher
understanding in cryogen and system behavior from
pressure and flow rate, liquid nitrogen needs to be tem-
our preliminary studies including Cha et al. (2014).
porarily stored in a special container, as the Dewar
cannot accommodate any high pressure. A vessel is
specially built to temporarily store liquid nitrogen from
2 DEVICES FOR CRYOGENIC STIMULATION dewar before it is pushed into the specimen borehole
EXPERIMENTS under higher pressure and flow using compressed gas
nitrogen pressure.
The laboratory system is mainly consisted of a tri- The vessel is made of annealed seamless stain-
axial loading system, a liquid nitrogen delivery, and less steel tubing with 5.1 cm OD, 4.1 cm ID, and
measurement system. Compressed nitrogen gas source 70 cm length (Figure 2). The 5.1 cm OD tubing is
is used to either directly pressurize boreholes or to reduced to 0.64 cm OD tubing by multi-stage tube fit-
push liquid nitrogen into borehole at higher pressure. ting reduction. The vessel tubing is rated for 7 MPa
Setups for submersion tests and unconfined tests are in ambient temperature, but 3.5 MPa is maximum rec-
not covered in this paper, but is detailed in Cha et al. ommended pressure at cryogenic temperature from the
(2014). vendor. The vessel is heavily insulated to minimize
heat transfer, and its internal storage volume is 1 liter
(Figure 2).
2.1 Cryogen delivery
As mentioned above, unlike pressure-induced frac-
Once an outlet is opened, liquid nitrogen is released turing, e.g., hydraulic fracturing, cryogen need to keep
out of a dewar by internal gas nitrogen pressure gener- flowing through in order to cool the borehole down.
ated inside the dewar. This pressure is kept at relatively In order to achieve this, we devised two methods. The
low levels ∼60–130 kPa by a pressure relief valve first one is placing a packer which has an inlet and
(Figure 1). The fluid injection system for cryogenic outlet at the entrance of the borehole (Figure 1 & 3).
382
Figure 2. Liquid nitrogen vessel for higher-pressure LN
injection (Photo was taken before applying insulation).
Figure 4. Cryogen flow-through at/around borehole (casing

detached to show inner inlet pipe).
2.2 Measurements
The measurements made include pressure, tempera-
ture, photography of specimens, and liquid nitrogen
consumption.
(a) Borehole & circular dent (b) Silicone pad, packer & gasket 2.2.1 Temperature measurement
Figure 3. Locations where gaskets and packers are placed Temperature measurement is a critical part to see per-
on the top surface of the block. formance and system behavior. T-type thermocouples
were selected for range and accuracy. Thin thermocou-
ples are placed into a borehole first before applying
A gasket is used to seal between packers and the sur- gaskets and packers. Then, packers are loaded by top
face around boreholes. Some forces are applied on top load platen to create tight seal (Figure 3b). When
of the packer to pressure-seal between the borehole and using a casing, thermocouple wires are inserted into
the packer/gasket. The gaskets that we used are made borehole between casing and borehole walls. One
of PTFE, which resists temperatures down to −212◦ C. thermocouple is suspended in the borehole to see tem-
To provide good contacts between gasket and specimen perature at the borehole and know nitrogen phase state,
surface, epoxy is applied to a gasket seat to fill any and another is attached to the borehole wall to see
uneven surfaces to provide tight seal between gasket actual conduction of temperature to the rock surface
and specimen surface (Figure 3). The packer method (Figure 5)
is okay for low borehole pressure applications, but it
leaks at higher borehole pressure. 2.2.2 Pressure measurements
Thus, the 2nd method is putting a casing into a Information of pressure in the borehole is important,
borehole, which turns out to be a robust solution for and pressure and temperature are intimately related
long stimulation and/or higher-pressure stimulation. to each other. Operating a pressure transducer at cold
2.5-cm stainless steel tubing as a borehole casing is temperature or exposing the sensor to cryogenic fluid
mounted to the borehole wall by epoxy for sealing and will damage sensing elements of most pressure trans-
pressure rating at cryogenic temperature. Epoxy gen- ducers, and cryogenic-rated pressure sensor that works
erally worked well for cryogenic fracturing, but could in such an environment is very expensive. Pressure
deteriorate after 3–5 times of uses. We also developed is measured using a pressure transducer attached to
our innovative design that allow effective flow through the end of a standoff pipe (Figure 1), where a vapor
with coaxial inlet and outlet (Figure 4). Insulation is cushion is created to prevent conductance of the cold
applied between dewar and specimen inlet to minimize temperature to the sensor. The temperature at the top
heat loss. of the stand-off pipe remains above 0◦ C throughout
383
Figure 7. Triaxial loading system: specimen and actuators
inside the containment mounted in the press bed frame.
Figure 5. Thermocouple attached to borehole wall. Loading axes and specimen side numbers are shown.
2.3.1 Triaxial loading device

A true triaxial loading system was developed to sim-
ulate effects of stress levels and stress anisotropy on
the characteristics of cryogenic fracturing. The con-
tainment is designed for the selected specimen size
20 cm × 20 cm × 20 cm.The triaxial loading (TX) sys-
tem can load the specimen up to 30 MPa in x and y axes,
and 40MPa in z axis, and can independently control
loadings in the three axes. The two hydraulic pistons
for x and y axes and the hydraulic press for z axis are
powered by three pneumatic hydraulic pumps. Three
heavy-duty ratchet tie-down straps (2300-kgf capac-
Figure 6. Liquid nitrogen consumption vs. time. ity each) surround the containment for extra safety.
A practical advantage of our system is the vertical
loading frame can be removed by rolling it away after
unlocking it from the bed. This ability provides a user
the stimulation (Figure 13). The standoff pipe length with space to work on specimens and inside the con-
should not be longer than necessary because a long nar- tainment.An open system is adopted to better deal with
row pipe can create drag force and thus may decrease any unexpected situation such as a cryogen spill and
responsiveness to fast-changing pressure. The length for ease of instrumentation (Figure 7).
of standoff pipe can be reduced by decreasing the Silicone or PTFE pads are placed between rock
diameter of the pipe. Common inexpensive pressure specimens and the loading pad to provide uniform
transducers can be used to use with the standoff pipe. contacts and loading to specimen surface. PTFE and
silicone pads resist temperature down to −212◦ C and
2.2.3 Liquid nitrogen consumption −62◦ C, respectively. An automatic servo control sys-
The amount of LN leaving the dewar and flown through tem for load control is not available in our system nor
a borehole was monitored by placing the dewar on required as our problem is quasi-static. Constant forces
a scale. Typical data for liquid nitrogen consumption can be maintained by either manual control or in quasi-
over time is shown in Figure 6. The nonlinear LN automatic manner using pressure relief valves attached
consumption vs. time shows that strong vaporization to the hydraulic lines. In the manual control, a small
occurred at the beginning, which slowed the flow of additional amount of pressure is applied by pumping
LN. Then as the system cooled down, the flow rate of when certain amount of natural decay occurs.
LN increases under the same amount of outlet opening.
2.3.2 Triaxial loading tests
Forces exerted by the hydraulic pistons are calculated
2.3 True triaxial loading equipment by multiplying hydraulic pressure in the hydraulic lines
measured using pressure transducers by effective area
A large specimen size (20 cm × 20 cm × 20 cm) are of hydraulic pistons. Then the specimen stresses are
considered in order to create thermal gradient required obtained by dividing the piston forces by the specimen
for thermal tensile fracturing. surface areas.
384
Figure 8. Confining load tests – manual control of constant
stress (isotropic stress 14 MPa).
Figure 10. Prominent fracture along the centers at the sides

created due to repeated, acute anisotropic conditions.
Figure 9. Extrusion of the silicone pads under loadings (28

MPa loading). Figure 11. An example of synchronized unloading to avoid
large stress anisotropic conditions, using pressure relief
valves in the hydraulic system.
Constant forces can be maintained by either the
manual control or in the quasi-automatic manner
using pressure relief valves in the hydraulic lines.
Our hydraulic system can make use of pressure relief any local, undulating surfaces, and have exceptional
valves in the hydraulic lines to quasi-automatically temperature rating down to -212◦ C. A drawback com-
control constant pressure, which needs less attention pared to silicone pad is when the specimen is undu-
than the manual control. In the manual control, we lating more than a certain level, it cannot distribute
manually pump small amount of pressure when certain load.
amount of natural decay occurs. The manual control In one of loading tests that we performed, unloading
can be more accurate than quasi-automatic control, but processes were abrupt and not perfectly synchronized
require more attention to changes in pressure. Note that in all three axes, which created acute anisotropic stress
the test Figure 8 is manually controlled within ±2% conditions in specimen. For example, the specimen
of errors. was subjected to anisotropic condition where x and y
Silicone pads are placed between the specimen sur- axes are loaded to ∼14 MPa and z axis is loaded to
face and loading plates. The silicone pads appeared ∼0 MPa due to unsynchronized rapid unloading for a
to function well in distributing the load on slightly couple of times (each time anisotropic condition lasted
uneven surfaces of specimens as we did not observe about 0.5 sec). This created prominent fractures (Fig-
any specimen damages after loading tests. Especially, ure 10). After this observation, we devised to make use
the top surfaces are rougher, and the silicone pads help of pressure relief valves to slowly and simultaneously
distribute load over the surfaces. However, the pads unload the three axes as shown in Figure 11.
inevitably extruded out the contact surfaces against
rough surfaces at high confining stress (Figure 9), and
damaged after repeated uses. The silicone pads resist 3 EXPERIMENTAL PROCEDURE
temperature down to −62◦ C. In our short-term tests
(less than an hour), this rating was sufficient as the Cryogenic stimulation is performed by flowing liquid
block surface remained above this temperature. nitrogen from its source into boreholes of specimens.
PTFE sheets are stiffer than the silicone pads, but Baseline measurements of rock conditions relevant
is malleable enough at high pressure to accommodate to fracture indication are done before any treatment,
385
and then the same measurements are performed dur-
ing treatments or after completing the treatments
for comparison. Fracture assessments include break-
through fracturing, acoustics, pressure decay tests,
visual inspection, and CT.
3.1 Procedure
As mentioned previously, because the specimen is
much hotter than liquid nitrogen, any stagnant con-
dition will increase the borehole temperature rapidly.
Thus, liquid nitrogen is continuously flowed through
the borehole. Using our cryogenic fracturing appara-
tus under triaxial loading conditions, we performed
two different cryogenic stimulation schemes. The first
one is low-pressure liquid nitrogen flow, where liq- Figure 12. The bridge saw that was used to cut the
uid nitrogen is directly flowed from the dewar from large shale and sandstone blocks into cubic specimens
a pressure difference between inside the dewar and (20 cm × 20 cm × 20 cm).
outside the dewar upon opening the dewar’s release
valve. Pressure ranges from 35–135 kPa in the bore-
hole, depending on the internal pressure level inside When other properties such as intact rock properties,
the dewar. The other scheme is high-pressure LN flow, density, and joint stiffness are the same, the wave
where liquid nitrogen is flowed through the bore- velocity can be used as a monitoring tool for fracture
hole under higher pressure (2–2.8 MPa) for faster generation. In addition, X-ray computed tomography
cooling in the borehole due to reduced film boiling (CT) scanning is used for fracture assessment before
effects. In high-pressure liquid nitrogen flow, there is and after stimulations.
also a pressure effect, which facilitates fracture open-
ing by helping to reach tensile strength of the rock.
Normally we apply the higher-pressure stimulation
multiple times because our vessel for storing liquid 3.3 Specimen preparation
nitrogen for higher-pressure injection is small (1 liter) Shale and sandstone were collected from outcrops of
(Figure 2), which makes each stimulation cycle very producing formations. A fairly large specimen size
brief (1–2 min). (20 cm × 20 cm × 20 cm) is selected in order to cre-
ate sufficient thermal gradient in the specimen for
an extended time. Collected rock blocks are precisely
3.2 Fracture characterization
cut into 20 cm × 20 cm × 20 cm cubic shapes using a
Fracture assessments are carried out by borehole pres- laser-guided bridge saw (Figure 12). Then a 2.5-cm
sure decay test, specimen breakdown pressure, and diameter wellbore was drilled by using a diamond
acoustic measurements. Borehole pressure decay is imbedded coring drill bit with 2.6-cm outer diame-
performed by applying a pressure to the borehole, ter to a depth of 15 cm. Following drilling, a 2.5-cm
shutting the borehole in, and monitoring the pressure stainless steel-316 tube (casing) were attached to the
decay. This was tested before cryogenic treatments as wellbore by applying epoxy after the thermocouples
a baseline, and then tested between treatments and were placed inside the wellbore. The casing extends
after completing cryogenic treatments for compari- five centimeters into the wellbore.
son. After all planned stimulation is completed for Mortar concrete blocks are used as surrogate for real
each specimen, specimens were subject to gas nitro- rock. A fresh concrete with a water to cement ratio
gen (GN) pressure to fully fracture (“breakdown”) the of 0.55, and sand to cement ratio of 2.5 was poured
specimen. These breakdown pressures are compared into the 20 cm × 20 cm × 20 cm mold and sealed in a
with baseline breakdown pressure of untreated speci- plastic bag. After 24 hours, the seal and mold were
men, and also with those of specimens that were treated removed and the concrete was cured under water
in different situations. Breakdown tests are done at the (ASTM, 2014a).
last stage after all other tests and measurements are Index properties of intact rock were obtained for
done as it will fully fracture to the surface. tested specimens. Permeability and porosity were mea-
Elastic wave propagation are measured using ultra- sured using CMS 300 (CoreLab). Elastic constants
sonic transducers before and after the treatments. were obtained from measurements of elastic wave
The post-stimulation measurements were done before velocities (Cha and Cho, 2007, ASTM, 2008b). Spe-
applying breakdown pressure, so that we can know the cific heat capacity was obtained by using a calorime-
effect of cryogenic stimulations. Elastic waves are gov- ter. Splitting tensile strength and unconfined uniaxial
erned by the mechanical properties of the medium. In compressive strength were obtained using procedures
particular, the wave velocity in jointed rock masses is from the ASTM standards (ASTM, 2008a, ASTM,
a function of the density of fractures (Cha et al., 2009). 2014b).
386
Figure 14. Shrinkage and expansion of the specimen due
to specimen cooling and warming, indicated from hydraulic
pressure responses during low-pressure LN flow (followed
by pressurization).
Figure 13. Temperature and borehole pressure during

low-pressure LN flow (and borehole pressurization) under
14 MPa isotropic confining stress.
4 CRYOGENIC STIMULATION
EXPERIMENTS
Following the procedures explained earlier, we per-

formed experiments of low-pressure LN flow and
higher-pressure LN flow. Some data is presented for
temperature and pressure behavior in and around bore-
hole, followed by assessment of generated fractures for
a low-pressure LN stimulation.
Figure 15. P-wave signature before (blue) and after (red)
liquid nitrogen treatment in a concrete specimen (Side 1&3).
4.1 Low-pressure liquid nitrogen flow through
borehole 4.2 Fracture assessment and characterization
In this scheme, low pressure LN is discharged Several methods are applied to check whether frac-
(70 ∼ 135 kPa) naturally from a liquid nitrogen dewar ture has been created by cryogenic stimulations. They
and injected into borehole to initialize fractures at include the pressure decay, acoustic waves, and break-
the borehole by thermal shock. Liquid nitrogen was down fracturing, etc.
injected until borehole and borehole wall was fully
cooled down to the nitrogen boiling point (Figure 13). 4.2.1 Acoustic waves
Although not included in the initial plan, when the Elastic wave data is collected to obtain parameters on
liquid nitrogen injection was stopped, we applied gas changes in specimen stiffness and quality by waveform
nitrogen pressure to the borehole, pressurizing it at comparisons.Acoustic signals were measured between
about 3 MPa (Figure 13a). During the pressurization Faces 1&3 and 2&4 (pairs of opposing faces). For
period (about 20 seconds), rapid temperature increase each set of faces, acoustic measurements were con-
is observed following the adiabatic compression of ducted at twelve locations (Figure 15). We observed
gas (Figure 13b). This gas nitrogen pressure is in an that acoustic velocities and amplitudes decrease after
attempt to open up cracks generated by the thermal cryogenic stimulations, indicating fracture creation in
shock. the specimens.
Contraction and expansion of specimens were
observed from pressure responses of the hydraulic 4.2.2 Post-stimulation breakdown fracturing by
system (Figure 14). When the specimen was cooled borehole pressure
down, the specimen shrank and the pistons in con- When stimulations are finished for each specimen,
tact with the specimen lost pressure at a faster rate specimens were subject to gas nitrogen (GN) pres-
than at the steady-state condition. When the tempera- sure to fully fracture (“breakdown”) all the way to
ture of the specimen increases, the specimen expands the block surface (Figure 16b). These breakdown pres-
and the piston starts to pick up pressure, making sures are compared with baseline breakdown pressure
the hydraulic pressure increase. Each jump represents of untreated specimen (Figure 16a). Breakdown tests
manual pumping to make up for pressure decay in are done at the last stage after all other tests and
hydraulic lines to keep ∼14 MPa. measurements are done as it will fully fracture to
387
data were gathered from cryogenic fracturing tests
on triaxially stressed rock blocks by using the newly
developed system proves that the system works ade-
quately. With this device developed, high-quality data
can be obtained to provide better understanding in
cryogenic fracturing, and develop and improve the
process toward field applications. As waterless or
reduced-water technology is sought for more than ever,
this technology should be sought for systematically in
the lab.
ACKNOWLEDGEMENT
Support for this research was provided by Research

Partnership to Secure Energy for America (RPSEA)
(Grant no. 10122-20).
REFERENCES
Figure 16. Post-stimulation fracturing by nitrogen gas. Alqatahni, N. B., Cha, M., Yao, B., Yin, X., Kneafsey,
T. J., Wang, L., Wu, Y.-S. & Miskimins, J. L. (2016)
the surface. A result shows that cryogenic treatment Experimental Investigation of Cryogenic Fracturing of
decreases breakdown pressure (PBD) level by low Rock Specimens under True Triaxial Confining Stresses.
pressure liquid nitrogen flow-through (Figure 16). SPE Europec featured at 78th EAGE Conference and
Exhibition, Vienna, Austria, 30 May–2 June 2016. SPE-
4.2.3 Borehole pressure decay test 180071-MS.
Any permeability enhancement was assessed by com- ASTM (2008a) ASTM D3967 Standard Test Method for
paring of pressure decay over time in a borehole before Splitting Tensile Strength of Intact Rock Core Specimens.
ASTM (2008b) D2845-08 Standard Test Method for Labo-
and after cryogenic stimulation. Pressure decay test ratory Determination of Pulse Velocities and Ultrasonic
results show that low and high-pressure liquid nitro- Elastic Constants of Rock. ASTM International.
gen stimulations increase permeability of stimulated ASTM (2014a) ASTM C192/C192M Standard Practice for
specimens. In one test, low-pressure LN flowed for Making and Curing Concrete Test Specimens in the
30 min resulting in a significant increase in perme- Laboratory.
ability indicated by a rapid pressure decay. Cycles ASTM (2014b) ASTM D7012 Standard Test Methods for
of higher-pressure stimulation under triaxial loading Compressive Strength and Elastic Moduli of Intact Rock
also led to significant permeability enhancements. It is Core Specimens under Varying States of Stress and
found that specimen temperature and confining stress Temperatures.
Cha, M. & Cho, G. C. (2007) Compression wave velocity of
play noticeable role in pressure decay profile. Detailed cylindrical rock specimens: engineering modulus inter-
pressure decay results and analysis are in Alqatahni pretation. Japanese Journal of Applied Physics Part 1-
et al. (2016) Regular Papers Brief Communications & Review Papers,
46, 4497–4499.
Cha, M., Cho, G. C. & Santamarina, J. C. (2009) Long-
5 CONCLUSIONS wavelength P-wave and S-wave propagation in jointed
rock masses. Geophysics, 74, E205–E214.
We built a laboratory system for controlled cryogenic Cha, M., Yin, X., Kneafsey, T., Johanson, B., Alqahtani, N.,
fracturing study under tri-axial loading conditions, Miskimins, J., Patterson, T. & WU,Y.-S. (2014) Cryogenic
fracturing for reservoir stimulation – Laboratory studies.
which mainly consist of a true triaxial loader, liquid Journal of Petroleum Science and Engineering, 124, 436–
nitrogen delivery/control, and measurement/ charac- 450.
terization system. The true tri-axial loading (TX) Grundmann, S. R., Rodvelt, G. D., Dials, G. A. & Allen, R.
system can load to reservoir confining stresses level E. (1998) Cryogenic Nitrogen as a Hydraulic Fracturing
on 20 cm × 20 cm × 20 cm cubic blocks, and indepen- Fluid in the Devonian Shale. SPE-51067-MS. SPE East-
dently control loadings in the three axes, and keep ern Regional Meeting. Pittsburgh, Pennsylvania, Society
constant pressure to specimens. It is an open-system of Petroleum Engineers.
with movable z-axis frame. The cryogen delivery and Mazza, R. L. (1997) Liquid CO2 improves Fracturing. Hart’s
control system allows pressurized and unpressurized Oil and Gas World, 22.
McDaniel, B. W., Grundmann, S. R., Kendrick, W. D., Wil-
liquid nitrogen flow through borehole, with borehole son, D. R. & Jordan, S. W. (1997) Field Applications
gas pressurization as necessary. Our characterization of Cryogenic Nitrogen as a Hydraulic Fracturing Fluid.
tools, such as acoustics, breakdown fracturing, and SPE 38623. SPE Annual Technical Conference and Exhi-
pressure decay tests, are able to effectively capture bition. San Antonio, Texas, 1997 Copyright 1997, Society
created fractures and permeability changes. Initial of Petroleum Engineers, Inc.
388
Steady state vs transient thermal conductivity of soils

ABSTRACT: Precise determination of thermal conductivity of porous media or soils in particular is vital in
performing analysis and modeling operations in a range of fields of engineering, agriculture, hydrology etc.
Depending on the type, thermal properties and medium temperature, steady state or transient techniques are
employed to obtain the thermal conductivity of soils. Although several studies have been carried out in the
past to assess the suitability and accuracy of each technique, there is an evident lack of assessment of the
two methods for a wide range of soil types as well as verification of the results against widely applicable soil
thermal conductivity prediction models. In this document, the result of thermal conductivity study on three
fine-grained and five coarse-grained soils analyzed using steady state and transient techniques in a two-phase
dry soil condition is presented. The steady state and transient measurements of each soil are further compared
with the prediction of seven models of soil thermal conductivity and the suitability as well as the accuracy of
both methods is assessed.
1 INTRODUCTION includes methods such as the guarded comparative

longitudinal heat flow technique, which uses a series
Thermal conductivity plays a significant role in engi- of reference materials of known thermal conductivity
neering applications such as the operation of small aligned in series with the specimen (Momose et al.
and large scale seasonal thermal energy storage, 2008).
embedding of underground power cables in selected Transient methods on the other hand involve apply-
porous media, nuclear waste disposal facilities, heat ing heat to the specimen and monitoring temperature
exchanger piles of structures, assessing the effect of changes over time, and then after using the transient
berms on the heat loss from thermal energy storage data to determine the thermal conductivity, usually
tanks etc (Doughty et al. 1983, Hart & Whiddon 1984, by applying analytical solutions to the heat diffusion
Rosen & Hooper 1989). Several methods exist for equation (Mitchell & Kao 1978, Farouki 1981, Low
measuring soil thermal conductivity (Mitchell & Kao et al. 2013, Alessandro 2007). The method by which
1978). These methods of measuring soil thermal con- the transient data is analyzed can significantly affect
ductivity fall into one of two categories: steady state or the obtained thermal conductivity. Transient methods
non-steady state (transient methods) (Mitchell & Kao of measuring thermal conductivity provide better ver-
1978, Farouki 1981). satility as compared to steady state methods because of
Steady state methods involve the application of their ease of use and the required short measurement
one-directional heat flow through a specimen and the time. Most commonly used transient methods include
observation of power or temperature difference across the hot wire method, transient plane source method
the sample when a steady state is reached (Mitchell & and the thermal needle probe (single or dual probes).
Kao 1978, Farouki 1981, Low et al. 2013). The thermal The probe methods have been more commonly used
conductivity of the specimen is then obtained using for the past 50 years (Farouki 1981).
Fourier’s Law. Generally, steady state techniques are Several researchers have performed studies on
helpful when the temperature of the specimen does steady state and/or transient methods of thermal con-
not change with time. However, the main limitation of ductivity determination independently and/or in joint
steady state methods is that a well-engineered exper- combination (Low et al. 2013, Alrtimi 2014, Mohsenin
imental setup as well as ample time is needed for 1980, Hooper & Lepper 1950, Sass et al. 1984). How-
establishing a steady state condition. Steady state tech- ever, there is a noticeable lack of evaluation of the two
niques are furthermore widely classified as absolute methods for a wide range of soil types as well as com-
or comparative methods (Alrtimi 2014). The former parisons with existing widely known theoretical and
includes the guarded hot plate method, unguarded hot semi-empirical thermal conductivity prediction mod-
plate method and the heat flow meter technique, where els. The focus of the body of work presented in this
the determination of power and temperature distribu- manuscript is to contribute to the existing literature
tion across the specimen is directly obtained from the and to provide users with a comprehensive compar-
input power and temperature parameters. The latter ison between experimental steady state and transient
389
techniques as well as theoretical and semi-empirical the geometric mean as given by Equation 3 (Farouki
models of thermal conductivity prediction for a wide 1981). The geometric mean (GM) equation has been
range of soil types. widely used for predicting two-phase thermal conduc-
For this purpose, the steady state and transient tivity of porous media (Farouki 1981, McGaw 1969).
thermal conductivities of three fine-grained and five Several researchers (Johansen 1975, Côté & Konrad
coarse-grained soils in two-phase dry condition, 2005, Lu et al. 2007) have also adopted the geomet-
obtained using a steady state thermal conductivity ric mean method in the development of semi-empircal
meter and a transient thermal needle probe respec- models with satisfactory accuracy.
tively, are analyzed experimentally.The results are then
compared with seven theoretical and semi-empirical
models of thermal conductivity, and the accuracy as
well as suitability of each experimental technique is
Hashin & Shtrikman (1962) proposed the relations
assessed.
given by Equations 4 (HS-U) & 5 (HS-L) for the
upper λu and lower λl bounds of thermal conductivity
respectively.
2 THEORETICAL & SEMI-EMPIRICAL
ANALYSIS
Three theoretical models (volume fraction & log-

models (Yun & Santamarina 2007) and cubic cell
model (Gori & Corasaniti 2004)) and four semi-
empirical models (Johansen 1975, Gavriliev 2004,
Côté & Konrad 2005 and Lu et al. 2007 models)
have been used in this research to verify the experi-
mental steady state and transient thermal conductivity
measurements. Details of each thermal conductiv- According to the cubic cell model (Gori & Corasaniti
ity prediction model are presented in the following 2004), the effective dry soil thermal conductivity can
sections. be obtained as:
2.1 Theoretical models

Numerous theoretical approaches have been devel-
oped to model the thermal conductivity of two-phase
soils based on the volumetric proportions of the dif- Yun & Santamarina (2007) fitted the volume fraction
ferent phases as well as the texture/fabric within the (VF) model (equivalent to the Complex Refractive
medium (Farouki 1981, Yun & Santamarina 2007, Index Method, CRIM) and the log-model for obtaining
Abuel-Naga et al. 2008). Unlike most empirical mod- the effective dry thermal conductivity of soils as given
els, the validity of theoretical models is not limited by Equations 7 & 8 respectively.
to specific conditions and they can be used to predict
thermal conductivity for a wide variety of soil types
(Abuel-Naga et al. 2008).
The parallel and series heat flow models can be con-
sidered as the upper and lower bounds of theoretical
models, respectively (Farouki 1981, Abuel-Naga et al.
2008). The two-phase parallel λParallel and series λSeries
thermal conductivities are given by Equations 1 & 2
respectively (Farouki 1981).
Figure 1 shows plot of two-phase dry thermal conduc-
tivity of a soil with a quartz fraction q = 0.70 (taken
as the average of the eight soils used in our study),
obtained using the different theoretical models for a
wide range of soil porosity.
where, n is soil porosity, λf is the fluid thermal

2.2 Semi-empirical models
conductivity (water, λw = 0.594 W m−1 K−1 or air,
λa = 0.024 W m−1 K−1 at 20◦ C, for saturated or dry Several semi-empirical models exist for estimating
soil states respectively), λs is the thermal conductivity soil thermal conductivity as a function of water con-
of solid soil particles. At low λs /λf ratios, parallel flow tent, porosity and other hydro-mechanical parameters
dominates, while, at higher λs /λf ratios, it is series flow (Farouki 1981, Johansen 1975, Côté & Konrad 2005,
which dominates (Farouki 1981). Lu et al. 2007, Kersten 1949, Van Rooyen & Win-
An important and useful ‘average’ of the parallel terkorn 1957). However, although most of the models
and series heat flow models is obtained by taking provide better accuracy as compared to theoretical
390
Figure 1. Thermal conductivity λ vs porosity n of a soil in Figure 2. Thermal conductivity λ vs porosity n of a soil
dry state obtained using theoretical models [VF is the vol- in dry state obtained using semi-empirical models [Lu et al.
ume fraction model, GM is the geometric mean model and model is valid for porosity n (0.2 < n < 0.6)].
HS-U & HS-L are the upper and lower bounds of the Hashin &
Shtrikman model respectively].
where, χ (W m−1 K−1 ) and η are parameters for parti-
models, they are generally limited to specific soil types cle shape effects with values of 1.70 W m−1 K−1 and
under specific boundary conditions. 1.80 for crushed rocks, 0.75 W m−1 K−1 and 1.20 for
Johansen (1975) proposed a semi-empirical rela- natural mineral soils, and 0.30 W m−1 K−1 and 0.87
tionship to obtain dry thermal conductivity of natural for organic fibrous soils.
soils λd (W m−1 K−1 ), based on soil dry density Lu et al. (2007) suggested a simple linear function,
ρd (kg m−3 ) and density of soil solids ρs (kg m−3 ), Equation 13, that describes the relationship between
Equation 9. thermal conductivity at dry condition λd (W m−1 K−1 )
and porosity n (0.2 < n < 0.6) for mineral soils, based
on two empirical parameters a and b, as the magnitude
of heat transfer in dry soils is related to soil porosity. Lu
et al. calculated values of 0.56 and 0.51 for empirical
Johansen also proposed that the thermal conductivity parameters a and b, respectively, by fitting heat pulse
of soil grains λs (W m−1 K−1 ) could be determined measured data with Equation 13.
using a geometric mean function, Equation 10, based
on the fraction quartz content of the total soil q
and thermal conductivities of quartz, λq = 7.7 W m−1
K−1 , and other minerals, λo = 2.0 W m−1 K−1 or Figure 2 presents plot of two-phase dry thermal con-
3.0 W m−1 K−1 for soils with q > 0.2 or q ≤ 0.2 respec- ductivity of a soil with a specific gravity Gs = 2.654
tively (Johansen 1975, Lu et al. 2007, Clauser & (taken as the average of the eight soils used in our
Huenges 1995). study), obtained using the different semi-empirical
models for a wide range of soil porosity.
The value of λs for all the models used in this study 3 EXPERIMENTAL PROGRAM
was obtained by calculating the quartz fraction based
on the coarse grain fraction of each soil using Equation 3.1 Tested soils
10 (Lu et al. 2007, Peters-Lidard et al. 1998).
Three silty clay soils (referred here as silty clays A-C)
Gavriliev (2004) suggested an empirical relation-
and five sandy soils (referred here as sands A-E) were
ship for estimating dry thermal conductivity of mineral
investigated. The common geotechnical properties of
soils λd (W m−1 K−1 ) based on dry density ρd
the soils (obtained following ASTM D420–D5876
(g cm−3 ), Equation 11. The relationship is valid for
(2001)) are listed in Tables 1 and 2.
mineral soils with dry density lower than 2 g cm−3 .
3.2 Determination of steady state thermal

conductivity
Côté & Konrad (2005) suggested a new empirical
equation for the dry thermal conductivity of soils λd Heat transfer phenomena play a vital role in many
(W m−1 K−1 ) as: problems which deal with transport of flow through
a porous medium. Under isotropic medium condi-
tions, where local thermal equilibrium is ensured
391
Table 1. Geotechnical properties of the studied fine-grained
soils.
Silty clay soil A B C
Gravel (%) 16.1 14.9 2.8

Sand (%) 8.8 12.2 7.1
Silt (%) 58.2 52.5 65.4
Clay (%) 16.8 20.2 24.6
Gs (–) 2.67 2.67 2.62
LL (%) 34.6 43.1 48.2
PI (%) 14.81 19.45 23.19
ρd (g cm−3 ) 1.49 1.43 1.38
USCS CL CL CL
Gs specific gravity, LL liquid limit, PI plasticity index, ρd

dry density, USCS unified soil classification system.
Figure 3. Steady state thermal conductivity measurement
setup.
Table 2. Geotechnical properties of the studied coarse-
grained soils.
Sand A B C D E
Gravel (%) 9.62 0 4.78 5.69 9.54

Sand (%) 90.27 100 95.16 94.14 89.95
Silt (%) 0.11 0 0.06 0.17 0.51
Clay (%) 0 0 0 0 0
Gs (–) 2.65 2.65 2.66 2.66 2.65
D50 (mm) 0.64 0.36 0.60 0.40 0.50
D10 (mm) 0.27 0.21 0.27 0.21 0.22
Cu (–) 3.00 1.81 2.63 2.19 2.72
Cc (–) 0.77 1.05 0.88 0.99 0.83
ρd (g cm−3 ) 1.73 1.52 1.77 1.77 1.76
Gs specific gravity, D50 grain diameter at 50% passing, D10

grain diameter at 10% passing, Cu coefficient of uniformity,
Cc coefficient of curvature, ρd dry density of soil
(Ts = Tf = T , where Ts and Tf are the temperatures

of the solid and fluid phases, respectively), Equations
14–17 can be obtained by combining heat transfer
equations in solid and fluid phases (Bejan & Lorente
2004, Bejan 2004).
Figure 4. Schematic representation of steady state thermal
conductivity cell (top) and dimensional analysis (bottom).
conductivity of the investigated soils in dry condi-

tion. The apparatus consists of a control panel, loading
machine, data acquisition and data analysis software,
a top heating plate, a bottom cooling plate and a
reference disc with known thermal conductivity, Fig-
ures 3 & 4.The specimen has a diameter of 100 mm and
a height of 30 mm, and is sandwiched between the top
where, the subscripts m, s and f refer to the porous heating and reference plates, Figure 4. The plates con-
medium, solid and fluid phases, respectively, n is soil sist of extremely thin PT 100 temperature sensors with
porosity, ρ is density, c is the specific heat of the solid, an accuracy of 0.05◦ C. The system measures sample
cp is the specific heat at constant pressure of the fluid, deformation with a TRS-0050 linear transducer with
λ is the thermal conductivity, q (W m−3 ) is the heat an independent linearity of 0.15% and repeatability of
production per unit volume, T is temperature, t is time 2 µm.
and v is the seepage velocity. The thermal conductivity of the specimen can be
A steady state thermal conductivity and diffusiv- calculated from the temperatures, distances and the
ity meter was used to obtain the steady state thermal known thermal conductivity of the reference plate,
392
as the temperature gradient within a homogeneous
body is linear (Stegner et al. 2011, Sass & Stegner
2012). In contrast to the divided bar apparatus (such as
King et al. (1982)), the temperature sensor is built-in
directly into the reference plate, thus lowering con-
tact resistances. For the system shown in Figures 3 &
4, Equation 18 applies provided that T1 > T2 > T3 ,
ensuring that heat flow occurs from the top plate via
the sample and reference disc to the bottom plate,
and T1 > Tambient > T3 , ensuring that heat flow from
the environment does not penetrate to the specimen
(Stegner et al. 2011, Sass & Stegner 2012). The
steady state thermal conductivity and average sam-
ple temperature are expressed in Equations 19 & 20
respectively.
Figure 5. KD2 Pro thermal needle probe setup.
where, λ (W m−1 K−1 ) is the thermal conductivity of

the sample, D (m2 s−1 ) is the thermal diffusivity of the
sample, r (m) is the radial distance to the line source, t
(s) is the amount of time that has passed since heating
has started and Ei is an exponential integral (transcen-
dent mathematical function). For a large measuring
time, the exponential integral is usually approximated
by the following expression:
where, T1 is the temperature of the top (heating) plate
(◦ C), T2 is the temperature of the reference disc (◦ C),
T3 is the temperature of the bottom (cooling) plate
(◦ C), Tav is the average sample temperature (◦ C), λp
and λv are the thermal conductivities of the sample where, γ = 0.5772. . . is Euler’s constant and α = r 2 /
and the reference disc respectively (W m−1 K−1 ), Sp , 4Dt. For a large enough time, the terms beyond ln α
Sv and S23 are the distances (m) as shown in Figure 4 in the series become negligibly small and the graph of
(bottom). T vs ln t becomes linear with slope equal to Q/4πλ.
Hence, the thermal conductivity of the medium can be
computed as:
3.3 Determination of transient thermal
conductivity
The transient thermal conductivity of the studied soils
was measured with a Decagon KD2 Pro thermal nee-
dle probe, based on transient line source measurement
in compliance to ASTM D5334-08 (2008) and IEEE 4 RESULTS AND DISCUSSION
(1992) standards, Figure 5. A thermal needle probe
(TR-1) with a diameter of 2.4 mm and a length of 4.1 Analysis of steady state conduction test
100 mm was used to measure the transient thermal con- parameters
ductivity of the soils. The sufficient length to diameter In Figure 6, the measured values of reference disc
ratio ensures that conditions for an infinitely long and temperature T2 required for obtaining the steady-
infinitely thin heating source are met, and the mea- state thermal conductivity of the investigated soils
surement error recorded for all samples was kept well are plotted. For all the measurements, a top heating
below the 0.015% limit. The KD2 Pro includes a linear plate temperature of 30◦ C and a bottom cooling plate
heat source and a temperature measuring element with temperature of 20◦ C were applied.
a resolution of 0.001◦ C. Generally, the coarse-grained soils exhibit a higher
The physical model behind the measurement measured value of T2 and hence a higher thermal
method is based on the application of heat at a con- conductivity λ near steady state condition, Figure 6.
stant rate Q (W m−1 ) to an infinitely long and infinitely This is attributed mainly to the higher quartz frac-
small line source. The temperature response of the tion q and lower porosity n (higher dry density ρd )
source over time is described as: of the sandy soils as compared to the silty clay soils.
The results comply well with previous findings which
state that soils with the highest quartz fraction exhibits
the highest thermal conductivity and vice versa
393
and a specific gravity Gs = 2.654 (taken as the average
of the eight soils used in our study).
As expected, all the measured thermal conductivity
values fall in between the series (lower bound) and par-
allel (upper bound) as well as the Hashin & Shtrikman
lower and upper bounds of heat flow conditions, Fig-
ures 1 & 7. Overall, with the exception of the Gavriliev
(2004) model, the semi-empirical models, Figure‘7
(bottom), provide better accuracy of prediction of
the experimental results of steady state and transient
measurements for both sands A, C, D & E (with rela-
tively higher dry density) as well as sand B and silty
claysA-C (with relatively lower dry density). However,
the theoretical models, Figure 7 (top), provide much
lower accuracy of prediction than the semi-empirical
models, with the log and volumetric fraction (VF)
models fitting comparatively better with the experi-
mental results when compared to the prediction of the
cubic and geometric mean (GM) models. The cubic
cell model significantly underestimates the measured
thermal conductivity values, while, the GM model
overestimates the measured thermal conductivity of all
soils. GM model only gives satisfactory results when
the ratio of thermal conductivity of solids to fluids
λs /λf < 15 (Côté & Konrad 2005b). In a two-phase
dry condition, where air is the fluid, λs /λa > 100 leads
to a significant overestimation of thermal conductivity
Figure 6. Observed values of reference disc temperature
by GM model (Farouki 1981, Johansen 1975).
T2 for the fine-grained soils (top) and coarse-grained soils The lack of accuracy of prediction of two-phase
(bottom). thermal conductivity of soils using theoretical solu-
tions based on properties such as geometry, porosity
(Johansen 1975, Côté & Konrad 2005, Lu et al. 2007, and volumetric fractions is mainly attributed to their
Abu-Hamdeh & Reeder 2000). Moreover, second inability to quantify factors such as the microstruc-
only to moisture content, dry density plays a sig- ture, gradation and soil texture (when having higher
nificant role in soil thermal conductivity (Salomone ratios of thermal conductivity of solid particles to air
et al. 1984, Salomone & Kovacs 1984, Salomone & λs /λa > 100), which play significant role in the ther-
Marlowe 1989). With an increase in the soil’s dry mal conduction of dry soils (Farouki 1981, Johansen
density, the number of soil particles packed per unit 1975).
volume increases creating higher number of contact
points between the soil grains. As a result the heat
flow path and consequently the thermal conductivity 4.3 Comparison of steady state and transient
of the resulting soil mass are increased. thermal conductivity with prediction models
In particular, sand B exhibits the lowest measured The accuracy of the steady state and transient thermal
T2 and hence a low thermal conductivity when com- conductivity test results is checked by comparing the
pared to the other sands due to its comparatively high root mean square error (RMSE) between the prediction
porosity (lower dry density) and poor gradation (uni- models and experimental results. The RMSE and is
form soil). Unlike poorly graded soils, well graded calculated using the following equation:
soils have a good distribution of coarse-grained matrix
bridged by finer particles, which in turn increases the
number of soil particles packed per unit volume and
the number of contact points, resulting in a higher soil
thermal conductivity.
The findings illustrate that both the experimen-
tal steady state and transient thermal conductivity
4.2 Steady state and transient thermal
measurements are rated to be of high accuracy accord-
conductivity results
ing to the semi-empirical models (Johansen (1975),
Figure 7 shows plot of experimental results of ther- Côté & Konrad (2005) and Lu et al. (2007) models in
mal conductivity measurements of the investigated particular). This result is to be expected as most semi-
soils as compared to the different theoretical and semi- empirical models are calibrated to account for factors
empirical prediction models of two-phase dry thermal such as soil fabric, gradation and texture in addition
conductivity of a soil with a quartz fraction q = 0.70 to the commonly used parameters such as porosity,
394
Figure 8. Bar chart of RMSE between model prediction and
experimental results.
when compared with both the Côté & Konrad (RMSE

of 0.028 W m−1 K−1 to 0.036 W m−1 K−1 ), Lu et al.
(RMSE of 0.028 W m−1 K−1 to 0.044 W m−1 K−1 )
and log models (RMSE of 0.038 W m−1 K−1 to
0.042 W m−1 K−1 ), Figure 8. Whereas, the transient
thermal conductivity measurements are more accurate
when compared with the Johansen model (RMSE of
0.029 W m−1 K−1 to 0.034 W m−1 K−1 ).
5 CONCLUSIONS
The thermal conductivity of three fine-grained and

five coarse-grained soils was analyzed experimen-
tally using steady state and transient techniques in
a two-phase dry soil condition. The steady state and
transient thermal conductivity measurements of each
soil were then compared with the prediction of seven
models of soil thermal conductivity and the suitabil-
ity as well as accuracy of both methods was assessed.
As expected, with both thermal conductivity measure-
ment techniques, majority of the sandy soils exhibited
a higher measured thermal conductivity than the silty
clay soils, due to their higher quartz fraction and
higher dry density (lower porosity). When assessing
the accuracy of the two techniques of thermal con-
ductivity measurement, the steady state thermal con-
Figure 7. Thermal conductivity λ vs porosity n compar- ductivity technique provides the most accurate results
isons of experimental results with theoretical (top) and when compared with both the Côté & Konrad (RMSE
semi-empirical (bottom) model prediction [steady state and of 0.028 W m−1 K−1 to 0.036 W m−1 K−1 ), Lu et al.
transient results are shown with gray and no fill markers (RMSE of 0.028 W m−1 K−1 to 0.044 W m−1 K−1 )
respectively].
and log models (RMSE of 0.038 W m−1 K−1 to
grain geometry as well as volumetric fraction (Farouki 0.042 W m−1 K−1 ). Whereas, the transient measure-
1981, Johansen 1975, Côté & Konrad 2005, Lu et al. ments are more accurate when compared with the
2007 and others). Bettered only by the semi-empirical Johansen model (RMSE of 0.029 W m−1 K−1 to
models, the experimental results comply well with the 0.034 W m−1 K−1 ).
prediction of the log and to a lower extent the VF theo-
retical models used in our study, as both models were ACKNOWLEDGEMENTS
calibrated based on soil thermal conductivity data by
Yun & Santamarina (2007). The authors would like to acknowledge the financial
Furthermore, when evaluating the accuracy of the support provided by the German Federal Ministry for
steady state and transient thermal conductivity exper- Economic Affairs and Energy (BMWi) under Grant
imental results, the steady state thermal conductiv- numbers 0325547B and KF3067302HF3, as well as
ity measurements provide the most accurate results the support of Project Management Jülich (PTJ).
395
REFERENCES Johansen, O. 1975. Thermal conductivity of soils. Ph.D. diss.,
Norwegian Univ. of Science and Technol., Trondheim
Abu-Hamdeh, N.H., Reeder, R.C. 2000. Soil thermal con- (CRREL draft transl. 637, 1977).
ductivity: Effects on density, moisture, salt concentration, King, M.S., Pandit, B.I., Hunter, J.A., Gajtani, M. 1982.
and organic matter. Soil Sci. Soc. Am. J. 64: 1285–1290. Some seismic, electrical and thermal properties of sub-
Abuel-Naga, H.M., Bergado, D.T., Bouazza, A. 2008. Ther- seabottom permafrost from the Beaufort Sea. Proceedings
mal Conductivity Evolution of Saturated Clay under of the Fourth Canadian Permafrost Conference, Ottawa,
Consolidation Process. Int. J. Geomech. 8(2): 114–122. pp. 268–273.
Alessandro, F. 2007. An apparatus for the routine measure- Low, J.E., Loveridge, F.A., Powrie, W. 2013. Measuring soil
ment of thermal conductivity of materials for building thermal properties for use in energy foundation design.
application based on a transient hot-wire method. Applied Proceedings of the 18th International Conference on
Thermal Engineering 27(14–15): 2495–2504. Soil Mechanics and Geotechnical Engineering, Paris,
Alrtimi, A.A. 2014. Experimental Investigation of Thermal France.
Conductivity of Soils and Borehole Grouting Materials. Lu, S., Ren, T., Gong, Y., Horton, R. 2007. An improved
Ph.D. diss., School of Civil Engineering and Geosciences, model for predicting soil thermal conductivity from water
Newcastle University, November 2014. content at room temperature. Soil Sci. Soc. Am. J. 71(1):
ASTM. 2011. D420–D5876: Annual book of ASTM stan- 8–14.
dards. Volume 04.08 Soil and Rock (I): D420–D5876 McGaw, R. 1969. Heat conduction in saturated granular mate-
and Volume 4.09 Soil and Rock (II): D5877—latest, West rials. Special Rep. No. 103, Highway Research Board, pp.
Conshohocken, Pennsylvania, United States. 144–131.
ASTM. 2008. 5334-08: Standard test method for determi- Mitchell, J.K., Kao, T.C. 1978. Measurement of soil ther-
nation of thermal conductivity of soil and soft rock by mal resistivity. Journal of the Geotechnical Engineering
thermal needle probe procedure. Division 104: 1307–1320.
Bejan, A., Lorente, S. 2004. The constructal law and the ther- Mohsenin, N.N. 1980. Thermal properties of foods and
modynamics of flow systems with configuration. Int. J. agricultural materials. Gordon and Breach, Science
Heat Mass Transfer 47: 3203–3214. Publishers, Inc.: New York, pp. 407.
Bejan, A. 2004. Convection Heat Transfer. 3rd ed., Wiley, Momose, T., Sakaguchi, I., Kasubuchi, T. 2008. Development
New York. of an apparatus for measuring one-dimensional steady-
Clauser, C., Huenges, E. 1995. Thermal conductivity of rocks state heat flux of soil under reduced air pressure. European
and minerals. In Rock physics and phase relations:A hand- Journal of Soil Science 59(5): 982–989.
book of physical constants, American Geophysical Union, Peters-Lidard, C.D., Blackburn, E., Liang, X., Wood, E.F.
pp. 105–126. 1998. The effect of soil thermal conductivity parameteri-
Côté, J., Konrad, J.M. 2005. A generalized thermal conduc- zation on surface energy fluxes and temperature. J. Atmos.
tivity model for soils and construction materials. Can. Sci. 55: 1209–1224.
Geotech. J. 42: 443–458. Rosen, M.A., Hooper, F.C. 1989. A model for assessing the
Côté, J., Konrad, J.M. 2005b. Thermal conductivity of base- effects of berms on the heat loss from partially buried
course materials. Can. Geotech. J. 42: 61–78. heat storage tanks. Proceedings of the 9th International
Doughty, C., Nir, A., Tsang, C.F., Bodvarsson, G.S. 1983. Heat Transfer Conference, Jerusalem, Israel, August 19–
Heat storage in unsaturated soils: Initial theoretical anal- 24, 1989.
ysis of storage design and operational method. Proceed- Salomone, L.A., Kovacs, W.D., Kusuda,T. 1984.Thermal per-
ings of the International Conference on Subsurface Heat formance of fine-grained soils. Journal of Geotechnical
Storage in Theory and Practice, Stockholm, Sweden. Engineering 110(3): 359–374.
Farouki, O.T. 1981. Thermal properties of soils. CRREL Salomone, L.A., Kovacs, W.D. 1984. Thermal resistivity
Monograph 81-1, US Army Corps of Engineers, Cold of soils. Journal of Geotechnical Engineering 110(3):
Regions Research and Engineering Laboratory, Hanover, 375–389.
N.H. Salomone, L.A., Marlowe, J.I. 1989. Soil rock classifica-
Gavriliev, R.I. 2004. Thermal properties of soils and surface tion according to thermal conductivity. Palo Alto: Electric
covers. in: D.C. Reston (ed.), Thermal analysis, construc- Power Research Institute, EPRI CU-6482.
tion, and monitoring methods for frozen ground, ASCE, Sass, J.H., Stone, C., Munroe, R.J. 1984. Thermal conductiv-
VA, vol. 492, pp. 277–294. ity determinations on solid rock – a comparison between
Gori, F., Corasaniti, S. 2004. Theoretical prediction of the a steady-state divided-bar apparatus and a commercial
thermal conductivity and temperature variation inside transient line-source device. Journal of Volcanology and
Mars soil analogues. Planet. Space Sci. 52: 91–99. Geothermal Research 20(1–2): 145–153.
Hart, G.K., Whiddon, W.I. 1984. Ground source heat pump Sass, I., Stegner, J. 2012. Coupled measurements of ther-
planning workshop. Summary of Proceedings, Palo Alto: mophysical and hydraulical properties of unsaturated and
Electric Power Research Institute, EPRI Report RP unconsolidated rocks. Proceedings of the 37th Workshop
2033-12. on Geothermal Reservoir Engineering, Stanford Univer-
Hashin, Z., Shtrikman, S. 1962. A Variational approach sity, Stanford, California, Jan. 30–Feb. 1, 2012.
to the theory of the effective magnetic permeabil- Stegner, J., Nguyen, D., Seehaus, R., Sass, I. 2011. Develop-
ity of multiphase materials. J. Appl. Phys. 33(10): ment of a thermal conductivity and diffusivity meter for
3125–3131. unconsolidated rocks. Proceedings of the 18. Tagung für
Hooper, F.C., Lepper, F.R. 1950. Transient heat flow appa- Ingeniergeologie, Berlin. (original in German).
ratus for the determination of heating piping and air Van Rooyen, M., Winterkorn, H.F. 1957. Structural and tex-
conditioning. American Society of Heating andVentilation tural influences on thermal conductivity of soils. Highway
Engineering 56. Research Board Proceedings 39: 576–621.
Kersten, M.S. 1949. Thermal properties of soils. Bulletin 28, Yun, T.S., Santamarina, J.C. 2007. Fundamental study of
Engineering Experiment Station, Minneapolis: University thermal conduction in dry soils. Granul. Matter 10(3):
of Minnesota. 197–207.
396
Determination of time-dependent undrained shear strength of mining landfill
M. Uhlig & I. Herle

Chair of Soil Mechanics and Foundation Engineering, Institute of Geotechnical Engineering,
Faculty of Civil Engineering, Technische Universität Dresden, Dresden, Germany
C. Karcher
RWE POWER AG, Bergheim, Germany
ABSTRACT: Today, more than a quarter of the energy produced worldwide is based on coal. In Germany
in 2014, lignite is the main resource for energy production. Lignite is mined in open-pit mines. With open-
pit mining, the overburden of the coal must first be removed. Normally, the material from the upper layers is
transported to the other side of the open-pit mine. There the material is deposited, filling the existing excavation.
If the material is fine grained and the consistency index is low then in the first step dams have to be built and
the wet material can be filled in behind to create a polder. Many dams and polders create together a large slope-
system, with a height from the coal base to the original surface elevation. This slope-system can be as high as
400 m. For engineering and planning these dams and polders, shear parameters (for example the undrained shear
strength) of the involved materials are necessary. Furthermore, the undrained shear strength changes over time
should be taken into account because the freshly dumped material is consolidating. In this paper two models are
presented showing how the undrained shear strength can be calculated using the initial state of the soil, statistics
and finite element analysis or analytical approach. For the finite element model the determination of the soil
consistency is the basis for the statistical analysis. From the statistical analysis the void ratio can be calculated.
Using this variable and the parameters of the hypoplastic soil model by Mašín, consolidation simulations are
performed. Finally, time-dependent void ratios are used to determine the undrained shear strength. The second
model uses analytical equations of the consolidation problem to calculate the effective stress. From the effective
stress the undrained shear strength can be determined. A comparison of both models is also presented on example
calculations.
1 INTRODUCTION The company developed sections to define where

the soft and liquid fine grained soil, that need to be
In open-pit mines the majority of the excavated mate- filled behind dams made of sand or gravel, have to
rial is overburden, which overlays the desirable lignite. be dumped. After the soft material has been filled in,
In West-Germany near Cologne, the material is exca- the polders are covered by stiffer clays. Another layer
vated with bucket-wheel excavators. This material is of sand cover is used to build a surface, where the
then transported by belt-conveyors to the other side of heavy spreader can be safely driven. Figure 1 shows
the mine. There it can be deposited using a spreader, a typical section of a completed polder and a dam.
which lifts the material by belt conveyors to a spe- The yellow areas are coarse grained soils, which have
cific height and then lets the material to fall free to the equal or less than 30% of fine (so called M1-material).
surface. With dry coarse grained material no prepa- Materials with more than 30% fine grains are divided
ration has to be done because it builds natural angles into two different material classes: the material class
of repose. However, with soft and wet fine grained M2T is a fine grained material which builds an angle
soils this procedure can not be done without preparing of repose during its deposition, hence the consistency
dams and polders, otherwise the material would flow varies from stiff to hard. This material, shown in Figure
away. To separate the soils in open-pit mines in North 1 in brown, is in an overconsolidated state and exists
Rhine-Westphalia there is a central collection station predominately in the form of lumps. Mostly, these are
for all belt conveyors coming from different operating high plasticity clays. The class M2N (in Figure 1 in
levels (excavators) and belt conveyors to the spreaders red) includes wet and soft or liquid soils, which are
at different levels. In this station it is possible to couple not able to build any reasonable angles of repose. Nor-
belt conveyors coming from the excavators and going mally, these materials are mixtures of fine sand with
to the spreaders with each other so that every spreader silt or clay, low plasticity silts or clays with a high water
can get the required material at any given time. content.
397
an extensive testing would be too high. Addition-
ally, it may be dangerous to enter freshly dumped
deposits with very low undrained shear strength to
acquire samples for testing. For this reason, it is neces-
sary to develop models to predict the undrained shear
strength.
The undrained shear strength can be calculated from
the effective critical friction angle ϕc using the critical
Figure 1. Section of a deposition. Blue 1-D column used
state soil mechanics theory (Schofield et al. ). In this
for calculations.
case, the undrained shear strength cu is, as shown in
equation 1, the half of the deviator stress q, which
Each of these sections forms one level. On each follows from the slope of critical state line M at a
level a spreader deposits the material evenly across the particular mean effective stress p :
width of the mine.The height of each level is always the
same. The spreader drives continuously forwards and
backwards, finishing a horizontal section across the
width of excavated area. Layers are stacked building
M is a function of the critical friction angle ϕc in
the complex temporary slope which changes every day
triaxial compression:
as the deposition moves on. These slopes go from the
bottom of the lignite level up to the natural ground
level or above it. The slope-system can reach heights
of more than 400 m.
The slopes and sections have to be engineered and
designed to be safe. First, the slopes must be sta- The effective mean pressure p can be obtained from
ble under their own weight. Also, the slopes must be a particular constitutive model. Using the Modified
designed for the load of the spreaders on the freshly Cam Clay model,
deposited soils (mostly 3 months old).
Calculations of the slope stability for fine grained
soils can be split into two models. If the soil is loaded
almost immediately after its deposition being fully where pi is the actual effective mean pressure, ηp is
saturated and unconsolidated, the load on the surface the isotropic overconsolidation ratio (p0 /pi ), r is the
will produce excess pore water pressures. Undrained pressure ratio between the normal compression and
shear strength cu is meaningful for the design in this critical state line p0 /pcs and is a ratio of the com-
case. If the soil is fully consolidated the load is compression λ and swelling index κ, both defined in a
pletely carried by the grains; the soil is drained and the (ν)-ln(p )-space:
effective shear parameters ϕ and c can be used. Dur-
ing the consolidation process the soil shear strength
changes between these two extremes. The undrained
shear strength is a conservative value here, therefore it
is used in the sequel to describe the evaluation of the The specific volume ν is defined as:
shear strength during the consolidation.
2 UNDRAINED SHEAR STRENGTH All mentioned quantities are depicted in Figure

2. The red point represents the actual specific vol-
The undrained shear strength cu depends on state of the ume (νi ) and the corresponding pressure (pi ). The soil
soil, which can be expressed in void ratio and effective state is overconsolidated, the maximum previous pres-
stress. During consolidation, the soil state changes. In sure is defined by p0 . If the behaviour is assumed to
normally consolidated soils the void ratio decreases be undrained, the specific volume remains constant
while the undrained shear strength increases. and shearing of the soil results in the critical state at
Commonly, to determine the undrained shear pressure pcs .
strength of a soil different tests can be performed. In Using equations 1 and 3 with r = 2 (Wood 1990) and
situ direct investigation methods such as cone penetra- simplifying the equation for a normally consolidated
tion or shear vane tests can be carried out to determine soil (pi = p0 ), the undrained shear strength cu1 can be
cu . Alternatively undisturbed samples can be taken out calculated from:
of the field. The samples can then be tested in the labo-
ratory using triaxial apparatus (UU-test), falling cone
or shear vane tests. However, it is usually not possible
to test the soil in several subsequent time increments
(for instance 1 week) to validate the undrained shear In the hypoplastic soil model for fine grained
strength changes over time because the costs of such soils according to Mašín (Mašín 2005), p0 is defined
398
surface, j as a control variable ( j = 1, 2, 3, …), h as
the height of the column and cv as the consolidation
coefficient:
Directly after surface loading (t = 0) the total stress

is equal to the pore water pressure. This means that the
initial pore water pressure distribution due to the load
at the top is a rectangle. After a very short time the
pore water pressure at the boundaries becomes zero.
In the case of a deposit of a significant height, how-
ever, the self-weight effect is an important factor. For
this problem equation 9 describes the analytical solu-
Figure 2. Definitions of the quantities in the Modified Cam tion (Lee 1979, Lee & Sills 1981). γ is the effective
Clay model.
unit weight of the material and n is a control variable
(n = 1, 2, 3, …).
differently and the undrained shear strength follows
from:
Herein p∗e is the Hvorslev equivalent pressure (see

Figure 2). This pressure is defined in a double loga- For the self-weight consolidation, the initial pore
rithmic space of specific volume and mean pressure. water pressure distribution is a triangle. Thus, the pore
water pressure at the top boundary is zero from the
beginning. Analogously to the case of constant load at
3 ANALYTICAL SOLUTION the top, the bottom value becomes zero after a short
time.
The following analysis is performed for the case of a Using the conditions mentioned before, the pore
1D consolidation. This means that water can only flow water pressures can be calculated as a superposition
vertically like in oedometer tests. Also, the deforma- of m different loads (u1,m ) and the self weight (u2 )
tions are only possible in the vertical direction. This consolidation. The summation must be done time t-
condition occurs roughly in the centre of a polder and depth z-wise. While the total pressure is known
as illustrated by the blue column in Figure 1. Closer and the pore water pressure follows from the men-
to the dams, which are made of more permeable tioned equations, the effective vertical stresses σv can
material, there are also horizontal drainage paths, be calculated from:
which decrease the consolidation time. The 1D case
is therefore the most conservative one.
The solutions of the consolidation problem follow
from the underlying partial differential equations. The
following conditions and assumptions apply: Here, a K0 -condition defined by Jaky:
1. drainage on both sides (top and bottom)
2. permeability is constant
3. stiffness is constant
4. load at the top is constant can be assumed. In the last step the mean pres-
sure can be inserted into equation 1 to obtain the
Well known and often used is the solution by
time-dependent undrained shear strength.
Terzaghi, which represents the simple case in which
the column is loaded on top with a constant stress.
The soil itself is weightless. The analytical equation
can be found in different books, e.g. (Atkinson and 4 NUMERICAL SOLUTION
Bransby 1977) or (Verruijt 2012). The analytical equa-
tion for the pore water pressure at time t and at a depth The basis for this procedure is a set of laboratory tests
z is shown in equation 8, using σ as the stress on the on samples taken from the spreader to determine the
399
continuous distribution was approximated with a dis-
crete distribution using ni intervals. Figure 3 shows an
example with 36 intervals. For each interval an IClog -
value (centre of interval) and a frequency f (height of
bar) is known. From the IClog -values the water content
can be calculated using equation 12 and the average of
the liquid limits and the plastic limits of all samples.
Using the water content, the void ratio was determined
from:
including the average of the grain density ρs of all

samples and the water density ρw . For all calculations
full saturation (Sr = 1) was assumed.
Figure 3. IClog -distribution continuous (red line) and dis- Finally, for each of the calculated void ratio a sim-
crete (blue bars – ni = 36 intervals). ulation of 1D consolidation is performed. During the
consolidation process the void ratio is changing con-
tinuously with time and depth. From this void ratio the
initial state. The obtained water content in situ w, the equivalent Hvorslev pressure can be calculated:
liquid wL and plastic wP limit can be used to calculate
a logarithmic consistency index IClog :
Subsequently, the undrained shear strength can be

determined from equation 7. For every simulation nd
Under the assumption that the IClog -values of all points at different depths are considered for output.
samples follow a normal distribution, see Figure 3, Thus nd void ratios are available for nt time steps and
the mean µ and standard deviation σ can be deter- ni simulations.
mined. The statistical distribution represents the initial As a result of all simulations ni · nd · nt undrained
consistency of the soils in the deposits. shear strengths are obtained. A further simplification
Subsequently, the finite element method is used of the results is needed. The focus of the presented
to solve the consolidation problem. The 1D calcu- analysis is on the time effects. Therefore the influence
lations were done by the software Tochnog (2013). of the depth and the initial state can be averaged using
The hypoplastic model for fine grained soils accord- the following.
ing to Mašín (2005) was used. The constitutive model First the average of cu from surface level to the bot-
reproduces the non-linear soil behaviour and includes tom of the column was calculated for every time step
barotropy and pycnotropy. For the basic formulation and simulation, denoted as cu . Herewith, the results
five parameters are necessary, they can be deter- are reduced to ni · nt undrained shear strengths.
mined by mechanical laboratory tests (oedometer, In the next step, the frequency f of each initial
CU-triaxial or direct shear box test) and are listed void ratio (IClog -interval) is used to sum all ni sim-
below: ulations. The idea behind is the following: if a void
ratio is rare in the deposition, then the undrained
• ϕc , critical friction angle
shear strength, calculated from this void ratio is also
• N , logarithmic specific volume at 1 kPa
exceptional. On the contrary, a void ratio measured
• λ∗ , slope of the normal compression line
often in the field (i.e. close to the peak of the IClog -
• κ∗ , slope of the reloading line
distribution) must be reflected in a corresponding
• r, initial shear stiffness.
calculated undrained shear strength with a high prob-
N , λ∗ , κ∗ are defined in a double-logarithmic space ability. Consequently, the frequencies of void ratios
of specific volume and mean pressure. from the initial IClog -distribution have to be correlated
The model has the same boundary conditions as directly with the calculated undrained shear strengths.
the analytical model described in section 3. The soil Figure 4 demonstrates the procedure.
column has an initial effective stress of 5 kPa corre- Figure 4a shows ni (here 36) intervals (ordinate) and
sponding to the initial load at the top. To simulate the the frequency of each interval (abscissa). This corre-
dumping, the gravity force is increased from 0 (ini- sponds directly to Figure 3. Each interval represents a
tially) to 1g within a few days (depending on the real IClog -value or initial void ratio for one simulation. Fig-
dumping time). ure 4b shows the calculated cu -values for each interval
The initial void ratios were calculated as follows: after a certain time (in Figure 4 one month). Then, new
First, the continuous distribution of IClog of sam- cu -intervals should be created in a new diagram show-
ples from the open-pit mine were considered. This ing the frequency f over the undrained shear strength
400
Table 1. Properties of the material for analytical model for
reference set (ref) and the parameter study sets (para 1 to
para 4).
Es [kPa] k [m/s] ϕc [–] equation for cu
ref 5000 1 · 10−7 27◦ 7

para 1 5000 1 · 10−7 27◦ 6
para 2 5000 1 · 10−7 20◦ 7
para 3 10 000 1 · 10−7 27◦ 7
para 4 5000 1 · 10−8 27◦ 7
Figure 4. Procedure to determine the discrete distribution

of cu .
cu (Figure 4c). The frequencies of the cu -intervals have

to be calculated using the frequencies fi from Figure
4a from all cui -values in Figure 4b which are located
in the cu -interval. For example the red box represents
the cu -interval from 200 kPa and higher. Three values
of cu are in between. The sum of the frequencies in
the red box on Figure 4a is equal to the height of the
red bar on Figure 4c. In this way a complete discrete
distribution of cu can be created.
In the last step, this discrete distribution can be
transformed in a continuous log-normal distribution
and the distribution characteristics can be calculated.
Here only the median value is discussed as a represen- Figure 5. Pore water pressure distribution at different times.
tative value (Herle et al. 2011). In such a way calculated
undrained shear strength depends only on time.
load due to own weight is completely carried by pore
water pressure.
5 EXAMPLE Beginning from the bottom, the consolidation starts
and proceeds very quickly as can be seen by comparing
The chosen parameter sets originate from the open- curves with time 0.01 d, 0.5 d and 5.99 d in Figure 5.All
pit mine. As a simplification, the parameters of both these lines are solid and represent only the self-weight
fine grained layers (see Figure 1) are summed togeth- consolidation. Then the cover is installed and the load
ers (M2T + M2N = M2). The cover layer, shown in at the top is added and reflected in the pore water
yellow in Figure 1, is simulated as a load at the distribution (curve 6.01 d). The maximum pore water
top. Time starts (t = 0) with the deposit of the M2- pressure moves up with time (see 6.01 d to 30.0 d) and
material. The M2-material is completely deposited in after a certain period stabilizes at the middle, which
3 days after t0 . Another 3 days later (t = 6 d) the cover corresponds to the point with the longest drainage
layer is installed. This layer loads the M2-material path. Finally the pore water pressures decreases to zero
with 130 kPa. The soil density was considered equal (360 d and 720 d), the soil is now fully consolidated.
1.8 g/cm3 . The soil own weight for both, analytical At both ends the undrained shear strength changes
approach and numerical analysis, is assumed to be the very fast. In Figure 6, the curves for 0.01 m and
same. 22.69 m show clearly this effect. Furthermore both
curves are the boundaries for the minimum and maxi-
mum final shear strength. The shear strength increases
5.1 Analytical method
from top to the bottom, due to the effect of the own
The applied parameters are summerized in Table 1. weight. Additionally, in Figure 6 the average undrained
They represent a clay silt mixture (M2-material). shear strength cu calculated over 6 heights is shown as
Using the reference parameter set the calculated a black dashed curve.
pore water pressure distribution is shown in Figure A parameter study was performed to evaluate the
5. As discussed previously the pore water pressure uncertainty of the input parameters. Figure 7 presents
distribution at the beginning is a triangle because the the results of this parameter study (the reference set is
401
Table 2. Parameters of the material for the FEM-model.
N λ∗ κ∗ ϕc r k
[–] [–] [–] [–] [–] [m/s]
0.82 0.07 0.004 27.0◦ 0.4 1 · 10−7
Table 3. Parameters for calculating the initial void ratio.
IClog -distribution Atterberg limits grain density
µ σ wL wP ρs
[–] [–] [%] [%] [g/cm3 ]
0.56 0.25 44.0 21.0 2.52

Figure 6. Undrained shear strength at different depths
(reference parameter set).
final undrained shear strength but it increases the con-

solidation time. Reducing the permeability 10 times
increases the consolidation time by a factor of 10. As
it can also be seen in Figure 7 (purple curve), in the
first 100 days no evident increase of cu in the middle
of the soil column can be observed.
5.2 Numerical method

In all simulations, the parameter set from Table 2 was
used. These average values were obtained from lab-
oratory test results using 6 representative soils from
the open-pit mine. In the FEM code the permeability
k was the same as in the analytical model.
The initial void ratios are calculated from a given
IClog -distribution (see section 4) and the liquid and
plastic limit. For the simulations the IClog -distribution
was split into ni = 36 intervals (see Figure 3).
For each initial void ratio a consolidation simulation
was performed using FEM. The void ratios were calcu-
lated for nd = 22 points (with a vertical distance of 1 m
Figure 7. Influence of different parameters: Curves for between the points) in a 1D column over a given time
depths of 11.40 m (solid lines) and 22.69 m (dashed lines).
period. As previously discussed, the undrained shear
strength cu can be calculated from these void ratios
as an average of all 22 points. Using the statistical
shown the with blue curves). The calculation results weighting (see section 4) the median undrained shear
are presented at two different depths. The results for strength was calculated. Figure 8 shows the result of
the point in the middle of the 1D column (11.4 m) the simulation.
are shown with solid lines. Additionally, a point 1 cm The analytical result in Figure 8 uses the cu of 22
above the bottom (22.69 m) was chosen. The results points in each calculated time step.This curve is princi-
for this point are shown in Figure 7 with dashed lines. pally the same as shown in Figure 6, the only difference
The green curves show the cu values calculated is the number of calculated points. As can be seen
using equation 6 (according to the Modified Cam Clay in Figure 8 the numerical curve starts and ends with
model). The value of λ is 10 times greater than of κ. a higher cu than the analytical calculation. The rea-
This equation yields the highest final shear strength son can be explained with help of Figure 9. Here the
(green lines in Figure 7). However, the difference is changes of specific volume (1 + e) over effective stress
smaller than 10% comparing with the reference set. are shown. As mentioned before, one simulation is per-
Decreasing the critical friction angle (red lines) the formed for a number of different initial void ratios. The
final undrained shear strength decreases too. The con- initial void ratios are decreasing during consolidation
solidation time is not affected by the variation of ϕc as can be seen in Figure 9. E.g., starting at a void ratio
or using equation 6 (Modified Cam Clay model). The of 1.02 (blue points), which represents the IClog of 0.09,
oedometric modulus of 10 000 kPa does not change the the normal compression line is reached very quickly
402
The explanation is the stiffness, which changes with
effective stress in case of hypoplastic model.
6 CONCLUSIONS
In this paper two methods for calculating time-

dependent undrained shear strength of mining landfills
are presented.
Comparing the analytical and numerical method the
following remarks can be formulated:
• The analytical solution performs very well, showing
the basic effects and many variations of parameters
can be calculated very fast (several minutes).
• Numerical model can take into account vari-
Figure 8. Undrained shear strength for numerical (num)
and analytical (ana) model. able stiffness or permeability changing with stress
(and/or density). Furthermore, the initial state of the
soils (normal consolidated or overconsolidated) can
be considered.
• The analytical model yields a conservative esti-
mate, whereas the FEM using statistc distributions
can produce more realistic results (at a higher
computational cost).
REFERENCES
Atkinson, J. & P. Bransby (1977). The Mechanics of Soil – An
Indorduction to Critical State Soil Mechanics.
Herle, I., D. Mašín, V. Kostkanova, C. Karcher, &
D. Dahmen (2011). Experimental investigation and theo-
retical modelling of soft soils from and mining deposits. In
International Symposium on Deformation Characteristics
Figure 9. Trend of specific volume (1 + e) with effective of Geomaterials, September 1–3, 2011, Seoul, Korea.
vertical stress for different initial states in a simulation. Lee, K. (1979). An Analytical and Experimental Study of
large Strain Soil Consolidation. Ph. D. thesis, University
of Oxford.
Lee, K. & G. Sills (1981). The consolidation of a soil stratum,
and is followed afterwards. If the initial void ratio is including self-weight effects and large strains. Interna-
smaller the normal compression line is reached at a tional Journal for Numerical and Analytical Methods in
higher effective stress. In some cases (see e0 = 0.41 – Geomechanics 5, 405–428.
black dots), the normal compression line is even not Mašín, D. (2005). A hypoplastic constitutive model for
reached at all because the maximum effective stress clays. International Journal for Numerical and Analytical
is too small. In this way, the overconsolidated states Methods in Geomechanics 29(4), 311–336.
are included in the numerical calculation. This effect Schofield, A., P. Wroth, L. in Engineering, at Cambridge, &
results in higher cu values in FEM when comparing University. Critical State Soil Mechanics.
Tochnog (2013). version 16, [finite element method soft-
both models. In the analytical model, the soil state is ware]. http://www.feat.nl/.
following only the normal compression line. Verruijt, A. (2012). Soil Mechanics. Delft University of
Further differences between the analytical and the Technology.
numerical model can be noticed in Figure 8. The per- Wood, D. M. (1990). Soil Behaviour and Critical State Soil
meability was the same for both models but the numer- Mechanics. Cambridge University Press.
ical model reaches a fully consolidated state faster.
403
Enhancement of soil thermal conductivity in dry condition
D. Shrestha, H. Hailemariam & F. Wuttke

Marine and Land Geomechanics and Geotechnics, Kiel University, Germany
ABSTRACT: The efficiency and performance of an underground power cable is strongly influenced by thermal
conductivity of soil in which it is placed. Most soils have extremely low thermal conductivity in dry state as
compared to that in saturated condition. It is essential to improve the thermal conductivity in dry state for the
optimum performance and safe operation of the cable, as backfill soils dry due to continuous heat generated
from the cable. This paper presents extensive laboratory study to develop backfill soils with higher dry thermal
conductivity by modifying grain size distribution into fuller curve gradation. Experimental results clearly show
a significant improvement of the thermal conductivity of modified soils. The improvement is attributed to fuller
curve gradation which contributes to obtain lower porosities and improve interparticle contacts. A result of
thermal simulation run for single cable clearly shows large improvement in heat dissipation for modified soil as
compared to unmodified soil.
1 INTRODUCTION dissipate the heat into the surrounding environment

for the efficient and safe operation of the cable. Thus,
Soil thermal conductivity plays preponderant role the thermal conductivity of backfill soil should be, in
in geo-engineering applications including thermal principle, higher than surrounding soil in dry as well as
effects, such as embedding of underground power wet state.The thermal conductivity of the backfill soils
cables, small and large scale seasonal thermal ground are highly moisture dependent. Typically their thermal
energy storage, oil and gas pipelines, heat exchanger conductivity decreases by five- to ten-fold when the
piles and nuclear waste disposal facilities. The effi- soil moisture is decreased to zero (Adams & Baljet,
ciencies and performances of some applications such 1968). It is due to fact that the water bridges formed
as underground power cable and energy piles are between soil solid particles improve contact between
strongly affected by soil saturation conditions; espe- the particles and thermal conductivity of water is 25
cially low saturation conditions, which alone can times larger than that of air. The reduction in thermal
negatively affect the overall soil thermal conductivity. conductivity as a result of low saturation largely influ-
Thus, it limits further possibility to extend their appli- ences the current carrying capacity of the underground
cation in hot-arid environments region, where extreme cables (Sandiford 1981, Len & Anders 2008) as well as
environmental conditions such as high air tempera- cable life (Karahan & Kalenderli 2011). On the other
ture, high evapotranspiration rate and poor rainfalls hand, energy piles used as ground heat exchanger loss
result low saturation condition and poor recharge of their efficiency up to 40% in dry conditions as com-
underground water and thus high thermal resistance. pared to that in fully saturated conditions (Venuleo
Therefore, it is essential to improve the soil ther- et al. 2016). The goal of this paper is to compensate
mal conductivity to enhance the performance of such such reduction by improving the thermal conductivity
applications. of backfill soils for lower saturation conditions.
High voltage underground power cable, alter- Soil thermal conductivity depends on several fac-
natives to overhead power cable, needs proper tors such as soil fabric (refers particle shape and
burial. The performance of an underground power size), porosity, water content, packing density, mineral
cables is critically influenced by the thermal composition and temperature (Vries 1963, Johansen
conductivity of the medium in which it is placed, as 1977, Farouki 1981, Rao & Sing 1999, Côté &
well as by the thermal properties of the cable itself. Konrad 2005a). As explained by Yun (2007) and
Heat is propagated due to power losses in the conduc- Nasirian (2015), a bigger particle size, well graded
tors, insulation, sheath and other components of the grain size distribution, higher solid thermal conductiv-
cable system (Sandiford 1981, Mozan et al. 1997 & ity, lower porosity, saturation and effective stress are
Afa 2010). The main role of backfill materials is to the factors to enhance thermal conduction and thus
405
increase the thermal conductivity of soils. In dry soils,
heat mainly transfers via solid inter-particle contacts
and thus, thermal conduction is controlled by num-
ber of contacts per volume and its quality. The number
of contacts per particle and contact quality depends
on the particle shape, grain size distribution and
packing density. Keeping these facts in consideration,
the grain size distribution of soil was modified into
fuller curve gradations to achieve the goal. The fuller
curve gradations (Fuller & Thomsan 1907) consists of
wide range of particle arrangement (coarse to fine par-
ticles) contribute to obtain lower porosities (or dense
mixes) and improve inter-particle contacts. The fine
grains act as thermal bridges between the grains to
increase the overall thermal connectivity of the soil
solid matrix.
In order to validate these assumptions, the thermal
conductivity of unmodified and modified backfill soil Figure 1. Particle size gradation.
samples was measured in dry conditions. The mea-
sured thermal conductivity data were also compared
with existing theoretical and semi-empirical mod-
els applicable to natural soils and crushed rocks to 2.2 Soil thermal conductivity measurement
analyse the thermal improvement of modified soils. The thermal conductivity of studied mixes was mea-
Furthermore, thermal simulation was carried out for sured with a thermal needle probe, Decagon KD2
single cable using Finite Element Methods software Pro, based on transient line source measurement in
(Comsol multiphysics) to observe the heat dissipation compliance to ASTM D5334-08 [23] and IEEE stan-
characteristics with the modified soils. dards [24]. The sufficient needle length to diameter
ratio ensures that conditions for an infinitely long and
infinitely thin heating source are met. The measure-
ment error recorded for all samples was kept well
below the 0.015% limit. The KD2 Pro includes a lin-
2 EXPERIMENTAL PROGRAM ear heat source and a temperature measuring element
with a resolution of 0.001◦ C, and computes the thermal
2.1 Analysed soil conductivity of the analysed material by the following
The thermal conductivity of unmodified sand sam- equation,
ples and modified sand samples was measured and
compared at dry conditions. Two different types of
sands referred in this paper as sand A (from Weimer,
Germany), and sands B (from Kiel, Germany) were
selected for this study. The gradation of both sands was
modified to three different fuller curve gradations of where λ (W m−1 K−1 ) is the thermal conductivity of
2 mmF, 4 mmF and 8 mmF. There were altogether six sample, Q (W m−1 ) is a constant rate of application of
modified soils, three from each sand. The modified heat, T is the temperature response of the source over
soils of 8 mmF, 4 mmF and 2 mmF have maximum time, and t (s) is the amount of time that has passed
particle size of 8 mm, 4 mm and 2 mm respectively. since heating has started.
Selection of bigger size particles was done on the basis
that bigger particles enhance the thermal conduction.
The fine soils, sodium bentonite in this study, was
added to complete fuller curve gradation attaining par-
ticle size lower than 125 µm. It acts as inter-granular
3.1 Experimental results analysis
bridges to increase the overall thermal connectivity
of the soil solid matrix. The particles size distribu- Figure 2 shows thermal conductivity results as a func-
tion of all sands and modified sands are presented in tion of porosity in dry state for all studied mixes. As
Figure 1. expected the thermal conductivity values increase with
A total of 33 samples including 8 unmodified sand a decreasing porosity. In this case, the thermal con-
samples were analysed with measurement of thermal ductivity is increased exponentially with decreasing
conductivity. All the samples were oven dried and the porosity. The experimental results clearly show the
bulk density was controlled. The measurements were significant improvement in thermal conductivity for
done at room temperature and atmospheric pressure dry conditions. About two to threefold increment in
conditions and were repeated at least three times for dry λ is noticed in all modified sand samples at poros-
each sample. ity less than 0.3.The thermal conductivity values of dry
406
Figure 2. Thermal conductivity as function of porosity in Figure 3. Thermal conductivity with respect to relative
dry state for modified (fuller) and unmodified sands. density in dry state for modified (fuller) and unmodified
sands.
unmodified sand obtained in this study are lower than enhance thermal conduction. The thermal conductiv-
0.4 W m−1 K−1 , range from 0.28–0.39 W m−1 K−1 ity is decreased with decrease in maximum parti-
for the porosities between 0.45 and 0.3). For cle size and consequently 2 mmF has lower thermal
modified sands the measured dry λ ranges from conductivity.
0.4–1.1 W m−1 K−1 for porosities between 0.35 and
0.2, which is significantly higher than that of ordinary
dry soils. As increase in the maximum particle size
of fuller gradation from 2 mm to 8 mm, the decrease
in porosity was observed and lowest porosities of 0.2
are attained with 8 mm fuller gradation of both sand The significant increase in thermal conductivity
samples. However, the dry thermal conductivity values with reduction in porosity reflects the increase in the
for the same porosity are not remarkably affected by number of contacts per volume and the improvement
maximum particle size of fuller gradation. For exam- in heat conduction efficiency. Wide range of particle
ple, thermal conductivity values are 0.59 W m−1 K−1 , arrangements in fuller gradation tends to attain denser
0.61 W m−1 K−1 and 0.60 W m−1 K−1 at a porosity of packing and a higher coordination number as num-
0.27 for 8 mmF, 4 mmF and 2 mmF fuller gradations ber of contacts per unit volume helps to improve the
respectively. quality of contacts. Due to larger surface area, the fine
In terms of relative density, the improvement in the particles act as thermal bridge at contacts to improve
thermal conductivity of modified sand samples can quality of contacts. The fuller curve gradation is one
be distinctively observed as unmodified and modified of the most influential parameters which can be used
sand samples have different minimum and maximum to improve dry thermal conductivity for granular type
porosities. Minimum and maximum porosities are in soils.
range of 0.20–0.21 and 0.31–0.35 respectively for the
modified sand samples while those for sand samples
3.2 Comparison with semi-empirical models
are 0.31 and 0.45 respectively. In Figure 3, the thermal
conductivity is gradually increased for unmodified Several semi-empirical models exist for estimating soil
soil as compared to modified soil. Densification of thermal conductivity as a function of water content,
soil affects significantly to improve the thermal con- porosity and other hydro-mechanical parameters (Ker-
ductivity in case of modified soil as compared to sten 1949, Van Rooyen & Winterkorn 1957, Johansen
unmodified soil. However, the modified soil has higher 1975, Farouki 1981, Côté and Konrad 2005, Lu et al.
conductivity even in loose state than unmodified soil. 2007). However, although most of the models provide
Figure 4 shows improvement in thermal conductiv- better accuracy as compared to theoretical models,
ity (I), calculated using Equation 2, with respect to they are generally limited to specific soil types under
relative density. The thermal conductivity values are specific boundary conditions.
increased by 50–180% in all state from loose to dense. Johansen (1975) proposed a semi-empirical rela-
As expected 8 mmF has higher thermal conductivity tionship, Equation 3, to obtain dry thermal conduc-
than others which confirm the bigger particle sizes tivity of natural soils λd (W m−1 K−1 ), based on soil
407
Figure 4. Improvement in thermal conductivity with respect Figure 5. Comparison of experimental results with
to relative density in dry state. semi-empirical models for dry condition.
dry density ρd (kg m−3 ) and density of soil solids ρs

(kg m−3 ).
Côté and Konrad (2005) suggested a new empirical

equation for the dry thermal conductivity of soils λd
(W m−1 K−1 ) as:
where, χ (W m−1 K−1 ) and η are parameters for parti-

cle shape effects with values of 1.70 W m−1 K−1 and
1.80 for crushed rocks, 0.75 W m−1 K−1 and 1.20 for
natural mineral soils, and 0.30 W m−1 K−1 and 0.87
for organic fibrous soils.
Gavriliev (2004) suggested an empirical relation-
ship for estimating dry thermal conductivity of min- Figure 6. Predicted versus measured dry thermal conduc-
tivity of modified soils.
eral soils λd (W m−1 K−1 ) based on dry density ρd
(g cm−3 ), Equation 3. The relationship is valid for
mineral soils with dry density lower than 2 g cm−3 .
empirical thermal conductivity models. All the mod-

els underestimate the measured thermal conductivity
Lu et al. (2007) suggested a simple linear function, of modified sands except Gavriliev model which pre-
Equation 5, that describes the relationship between dicts quite close to measured λ at porosity greater than
thermal conductivity at dry condition λd (W m−1 K−1 ) 0.27. This is probably due to the fact that the models are
and porosity n (0.2 < n < 0.6) for mineral soils, based primarily based on dry density (or porosity) of media
on two empirical parameters a (=0.56) and b (=0.51), and lack considerations of inherent presence of con-
as the magnitude of heat transfer in dry soils is related tacts and bulk conductivity of each constituent in dry
to soil porosity. soils.
Figure 6 presents the comparison between the mea-
sured and predicted thermal conductivity data from
five models. The measured λ of modified soil are
Figure 5 shows measured thermal conductivity larger than predicted data from all models. However,
values against porosity and trends predicted with these models are able to predict for the unmodified
408
Figure 7. Temperature (◦ C) distribution around underground cable with sand (left) and modified sand (right) at 24, 72 & 120
hours (2nd row to bottom) for dry condition. Geometry of model with mesh generation (top).
soil.The root mean square, RMSE and bias in model where m is the number of measurements, and
prediction were calculated as: λm and λp are the measured and predicted
values of dry thermal conductivity λd , respec-
tively. The RMSE and bias for the modified
sands were: 0.333 W −1 K−1 and 0.289 W −1 K−1 for
Johansen model, 0.382 W −1 K−1 and 0.332 W −1 K−1
for Côté and Konrad: soils, 0.202 W −1 K−1 and
0.139 W−1 K−1 for Côté and Konrad: crushed rocks,
409
Table 1. Thermal property of the simulated geometry.
Thermal Heat capacity at

Material Conductivity Density constant pressure
λ [Wm−1 K−1 ] ρ [kgm−3 ] Cp [Jkg−1 K−1 ]
Soil 0.18 1450 900

Sand 0.36 1650 800
Modified-sand 1.0 2000 800
0.187 W −1 K−1 and 0.106 W −1 K−1 for soil Gavriliev,

and 0.386 W −1 K−1 and 0.328 W−1 K−1 for Lu model.
The RMSE of Johansen model, Lu model and Côté
and Konrad model are considerably higher than that
of other two models.
Figure 8. Temperature versus time at two observation points

for sand and modified sand.
3.3 Temperature distribution around
underground cable
The thermal simulation for single cable has been
carried out using Finite Element Methods software Figure 8 shows temperature increment with respect
(Comsol multiphysics) to observe the heat dissipation to time at two selected points for unmodified and mod-
characteristics around underground cable with unmodified sands. The heat dissipations in case of modified
ified and modified backfill soil. Figure 7 (top) shows sand are faster as compared to unmodified sand in both
geometry of model with mesh generation, which con- observation points. At OP-2, there is abrupt change
sists of single cable, directly buried in the trench at in temperature in the beginning and 63◦ C is reached
1.3 m depth. The trench has a width of 0.5 m and depth in just 30 hrs for modified sand, whereas 200 hrs are
of 1.5 m, while the size of the model, which represents needed to reach the same temperature for unmodified
surrounding soil, is 6 m in width and 5 m in height. sand. Then-after the temperature increases steadily
The trench first was first filled with sand and then after 50 hrs and 100 hrs for modified and unmodi-
with modified sand for the simulation. The properties fied sand respectively. The temperature is gradually
of material, used in this simulation, are presented in increasing at OP-1 and reaches to 50◦ C and 32◦ C for
Table 1. modified and unmodified sand respectively in 200 hrs.
The geometry was simulated in 2-D environment,
while the influence of air thermal conductivity was
taken into consideration. Each material was assumed 4 CONCLUSION
to be homogeneous with constant thermal conductivity
and a constant specific heat.The cable temperature was In this study, backfill materials with high thermal con-
fixed at 90◦ C, maximum temperature the underground ductivity in dry state were developed and analysed.The
power cable can reach. At the boundary between trench modification of grain size distribution into fuller curve
and surrounding soil, continuity of heat flux and tem- gradation shows a strong effect on increasing thermal
perature was assumed. The soil surface and other three conductivity of sand for dry condition. Comparing
sides were represented as an isothermal boundary at the thermal conductivity of modified sand with that
the temperature of 20◦ C and 15◦ C respectively. The of unmodified sand shows improvement upto 180%.
transient heat transfer equation was solved numerically Hence, the backfill materials developed in this study
for 200 hrs using the Cosmos Multiphysics software. can be efficiently used to bury high voltage under-
Figure 7 (2nd row to bottom) shows the outputs of ground cables under all moisture conditions; as well
simulation showing the temperature distribution for graded soils have a great capability to hold moisture
24 hrs, 72 hrs and 120 hrs respectively with unmodi- content (Adams & Baljet 1968, Moya et al. 1999)
fied sand (left) and modified sand (right). The results and good thermal stability (Adams & Baljet 1968,
clearly indicate the large improvement in heat dissipa- Radhakrishna et al. 1980). The work presented in
tion for the modified soils which helps to prevent the this study also confirms previous finding such as the
cable overheated and thus prevent the cable failure and decrease in thermal conductivity with increase in soil
extend their life. porosity, particle size distribution influence on dry
Two observation points OP-1 & OP-2 were plot- thermal conductivity, improvement of quality of con-
ted at depth of 0.5 m & 1.5 m respectively as shown tacts due to addition of fine particles which in turn
in Figure 7 (top) to observe the temperature rise enhances thermal conduction. The output of thermal
characteristics within the selected period of 200 hrs. simulation done in this study for underground cable
410
clearly indicate the large improvement in heat dissipa- Gavriliev, R.I. 2004. Thermal properties of soils and surface
tion for the modified soils which helps to prevent the covers, In D.C. Reston (ed.), Thermal analysis, construc-
cable overheated and thus extend the cable life. Rea- tion, and monitoring methods for frozen ground: 277–294.
sonable agreement could not be obtained with existing VA: ASCE.
IEEE. 1992. Guide for soil thermal resistivity measure-
semi-empirical models while comparing with experiments, Inst. of Electrical and Electronics Engineers, Inc.
mental results. The work in this research can be used New York.
further to develop a complete set of prediction model Johansen, O. 1975. Ph.D. diss. Norwegian Univ. of Science
over the full range of saturation. and Technol, Thermal conductivity of soils. Trondheim
(CRREL draft transl. 637, 1977).
Karahan, M. & Kalenderli, O. 2011. Heat transfer-
Engineering applications, In V. Vikhrenko (eds.), Coupled
ACKNOWLEDGMENT Electrical and Thermal Analysis of Power Cables Using
Finite Element Method: 205–230. Croatia
The authors wish to acknowledge the financial sup- Kersten, M.S. 1949. Thermal properties of soils. Engineering
port provided by the German Federal Ministry for Experiment Station Bulletin 28, University of Minnesota,
Economic Affairs and Energy (BMWi) under Grant Minneapolis.
numbers KF3067302HF3 and 0325547B. Lu, S. Ren, T. Gong, Y. & Horton, R. 2007. An improved
model for predicting soil thermal conductivity from water
content at room temperature, Soil Sci. Soc.Am. J. 71: 8–14.
Moya, R.E.S. Prata, A.T. & Cunha Neto, J.A.B. 1999. Exper-
REFERENCES imental analysis of unsteady heat and moisture transfer
around a heated cylinder buried into a porous medium,
Adams, J.I. & Baljet, A.F. 1968. Thermal behaviour of cable International Journal of Heat and Mass Transfer 42:
backfill materials, IEEE Transaction on Power Apparatus 2187–2198.
and Systems 87(4): 1149–1161. Mozan, M.A. El-Kady & Mazi, A.A. 1997. Advanced ther-
Afa, J.T. 2010. Subsoil temperature and underground cable mal analysis of underground power cables, Record of
distribution in Port Harcourt City, Res. J. Appl. Sci. Eng. the Fifth International Middle East Power Conference
Technol. 2(6): 527–531. MEPCON’97, Alexandria, Egypt: 506–510.
ASTM. 2008. Standard test method for determination of ther- Nasirian, A. Cortes, D.D. & Dai, S. 2015. The physical
mal conductivity of soil and soft rock by thermal needle nature of thermal conduction in dry granular media,
probe procedure, ASTM 5334-08. Géotechnique Letters 5: 1–5.
Côté, J. & Konrad, J.M. 2005a. A generalized thermal con- Radhakrishna, H.S. Chu, F.Y. & Boggs, S.A. 1980. Thermal
ductivity model for soils and construction materials, Can. instability and its prediction in cable backfill soils, IEEE
Geotech. J. 42: 443–458. Transactions on Power Apparatus and Systems 99(3):
de Len, F. & Anders, G.J. 2008. Effects of backfilling on cable 856–867.
ampacity analyzed with the finite element method, IEEE Sandiford, P. 1981. Cable backfill materials-state-of-the-art,
Transactions on Power Delivery 23(2): 537–543. Proceedings of the Symposium on Underground Cable
de Vries, D.A. 1963. Physics of plant environment, In W.R. Thermal backfill, Toronto, Canada: 3–9.
Van Wijk (eds.), Thermal properties of soils: 210–235. Van Rooyen, M. & Winterkorn, H.F. 1957. Theoretical and
Amsterdam: North-Holland Publ. Co. practical aspects of the thermal conductivity of soils and
Farouki, O.T. 1981. Thermal properties of soils, CRREL similar granular systems, U.S. Highway Research Board,
Monograph 81-1, US Army Corps of Engineers, Bulletin 159: 58–135.
Cold Regions Research and Engineering Laboratory, Venuleo, S. Laloui, L. Terzis, D. Hueckel, T. & Hassan,
Hanover, N.H. M. 2015. Effect of microbially induced calcite precipi-
Fuller, W.B. & Thomsan, S.E. 1907. The laws of proportion- tation on soil thermal conductivity, Géotechnique Letters
ing concrete, Trans. ASCE 59(2): 67–143. 00: 1–6.
Gangadhara Rao, M.V.B.B. & Singh, D.N. 1999. A general- Yun, T.S. & Santamarina, J.C. 2007. Fundamental study
ized relationship to estimate thermal resistivity of soils, of thermal conduction in dry soils, Granul. Matter 10(3):
Can. Geotech. J. 36: 767–773. 197–207.
411
Heat energy recovery from waste water in the Glasgow Subway system
N. Hytiris, K. Ninikas & R. Emmanuel

Glasgow Caledonian University, Glasgow, Scotland, UK
P.L. Younger
University of Glasgow, Glasgow, Scotland, UK
ABSTRACT: This paper investigates the feasibility of utilizing the waste subsurface water ingress inside the
Glasgow Subway system. At present this unused excess water is being discharged into the city’s drainage system
as waste. This valuable resource could be channeled through a water source heat pump to produce heat energy
for domestic or public use. A study was carried out in order to calculate the heat contained in the water. Water
flow and water temperature were recorded over a full year within fifteen different places around the network of
underground tunnels. A feasibility study to review the number of support factors that could profit the subway
system was undertaken as well. Options were discussed and a selection of a site inside the tunnels for a pilot
study was decided. The findings of this study are expected to develop an appropriate renewable solution and
design a cost effective heat pump system. This waste water will be collected and used to recover heat energy.
During this process energy will be produced from a waste product using a sustainable and environmental friendly
method.
1 INTRODUCTION
The need for replacing conventional fuel finding alter-

native sources becomes day by day bigger. A basic
factor that led the UK government into more environ-
mental friendly ways of heating are the obligation of
reducing the CO2 emissions in 1990’s levels by 2020
(Scottish Government, 2012) and to fully decarbonise
heating until 2050 (Department of Energy & Climate
Change, 2013). Ground source heat pump (GSHP)
systems have shown potential to reduce energy con-
sumption and as result a CO2 reduction of more than
50% compared with conventional heating systems Figure 1. A typical Glasgow Subway map.
(electricity, oil).
A study was carried out from May 2014 until
April 2015 in the Glasgow Subway system to inves-
tigate the possibility of using the water ingress inside 2 BACKGROUND
Glasgow’s Subway tunnels for space heating and
Domestic Hot Water (DHW) through a water source Heat pumps have always been the key technology in
heat pump (WSPH). the use of excess heat in lower temperatures. They
The Glasgow Subway forms a circle in the center- use compression (the same principal as a refrigera-
west of the city. The entire passenger railway is under- tor) to extract tepid low grade heat to produce heat
ground, contained in twin tunnels, allowing clockwise for space and/or water heating in general. They can
circulation on the “outer” circle and counterclock- also be reversed to produce cooling. Heat pumps,
wise on the “inner”. Fifteen stations are distributed as shown in Figure 2, are designed to move thermal
along the route length of just over ten kilometers. The energy opposite to the direction of spontaneous heat
river Clyde dissects the circular route, with eight sta- flow by absorbing heat from a cold space and releasing
tions in the North and seven in the South as shown in it to a warmer one. A heat pump uses some amount of
Figure 1. external power to accomplish the work of transferring
413
Figure 2. A typical water source heat pump (WSHP).
Figure 4. A typical sump located between the rail tracks in

the Glasgow Subway system.
network adjacent to the vicinity of the nearest Subway

station.
The two most critical measurements are the water
flow and the water temperature in each sump. Some
additional data were collected at the same time in
Figure 3. Sump locations in the Glasgow Subway system.
order to help us evaluate the water flow through out
a year (Average Glasgow temperature, humidity, and
precipitation).
energy from the heat source to the destination (space

heating/hot water). 3.1 Water flow measuring methods
3.1.1 First method of water flow measurement
(Drop test)
3 METHODOLOGY In each sump there is a probe that when the water
reaches a certain level the pump starts pumping the
A series of measurements were undertaken from May water out of the sump. This period is called “active
2014 to April 2015 and these consisted of the fol- time”. When the pump will stop functioning; it takes
lowing: measurements of temperature and flow of some time before the water flux will raise (inside the
the ingress water, consideration of geological and sump) the water level up to a point that the pump
geotechnical parameters to identify sites suitable for will start working again. This period is called “inactive
heat pump locations. During this period, the seasonal time” and this is the actual water flow. Given that the
variations of the water ingress were identified. Water dimensions of each sump are known; the water flux
sampling was carried out at all sumps for the chemical can be calculated by measuring the water level inside
analysis of the water regime. The 21 sumps located the sump (Figure 6) between the “active time” and the
inside the tunnel system that were monitoring are “inactive time”.
displayed in Figure 3. The higher water level (when pump starts working)
In addition to the above, atmospheric data were also and the lower level (when pump stops working) are
compiled in order to monitor the Glasgow weather thus measured (Figure 5). This height difference mul-
changes i.e. do temperature, humidity, atmospheric tiplied by the area (surface) of the sump, gives us the
pressure and rainfall. This was done by using Glas- water volume that is being pumped out. Dividing this
gow Caledonian University’s meteorological station volume by the time (seconds) that the pump is active
to collect the readings for each day of the site vis- gives the discharged volume during the active time of
its. A monthly meteorological table was also produced the pump. From the lower point when the pump has
to parallelize the tunnel conditions with the Glasgow stopped the time that the water takes to reach again
climate conditions. the highest point in the sump is being measured (until
The sumps (Figure 4) are generally flat bottomed the pump will start working again). This depth raised
rectangular chambers formed either within the tunnel multiplied by the sump surface gives us the water vol-
invert or station platforms and range from 0.50 m to ume that inflows the sump in a specific period of time
2.50 m in depth from access level. The pumping sta- (F). This is the average water flux in each sump. At
tions inside each sump are generally equipped with least two readings are taken during each visit in each
two submersible pumps. The excess water from the sump are taken to have more data and therefore more
tunnels is being pumped out and discharged into the accurate averages of the water flux.
414
Figure 7. The ultrasonic flow meter showing the water
flux (l/s).
Figure 5. Measuring the sump water level.
Figure 6. Two consecutive water flow measurements in Figure 8. The transducers clamped onto the discharge pipe.
sump No. 1.
Readings were taken with a rigid measuring stick 3.2 Temperature measurements
as well as with an automated depth meter, in which a Water temperature is being measured inside each sump
water sensitive sensor at the end of the measuring cord and also from all the water inlets into the sump and
completes a circuit when it touches the water level, the average is calculated. A digital probe thermome-
sounding a buzzer. This calibrated cord indicated the ter (Tiny Tag, TGP4020, range = −40◦ + 125◦ C,
distance from the both water levels to the surface. accuracy = ±0.35◦ C in the 0–60◦ C range) was used to
record the temperature every 10 sec. The thermome-
ter is kept in place for 2 minutes (as a minimum)
3.1.2 Second method of water flow measurement so the minimum 12 temperature measurements were
(Ultrasonic flow meter) received from each measuring point (Figures 9 & 10).
In order to crosscheck the water flux in each sump Atmospheric data were compiled in order to identify
a second method was used as well. A flow meter if the weather changes i.e. atmospheric temperature,
(Dynasonics TFX/DMS 1002 ultrasonic flow meter pressure and rainfall, had any effect on the subsur-
with clamp on pipe transducer) (Figure 7) was used face tunnel water. This was undertaken by using the
providing more accurate water flow measurements. University’s meteorological station.
The device was calibrated prior to each measurement
inputting in the software the pipe material (uPVC or
steel), the diameter of the pipe (∅120 or ∅160) and the
3.3 Water flow and water temperature readings
liquid (water).
The portable transducers (Figure 8) were clamped Water flow and water temperature readings were taken
onto the pipe applying also liquid silicon to the from May 2014 to April 2015. These were compared
transducers to assist “reading” the flux. with the average Glasgow temperature and humidity
415
Table 1. Readings from sump 1.
Sump No: 1
WF1: Water flow WT1: Water Temp. GMT: Glasgow Mean
Temp. GMH: Glasgow Mean Humid.
WF1 WT1 GMT GMH

Month Year (l/s) (◦ C) (◦ C) (%)
May 2014 6.7 14.17 11.40 80

June 2014 6.3 13.45 16.70 83
July 2014 5.3 14.95 15.80 77
August 2014 3.9 16.03 16.00 88
September 2014 1.9 15.40 15.00 67
October 2014 1.8 16.13 12.00 67
Figure 9. The digital thermometer. November 2014 1.8 13.72 10.00 75
December 2014 2.0 13.20 04.10 96
January 2015 2.1 14.13 05.80 77
February 2015 2.2 12.12 05.70 80
March 2015 1.5 12.81 05.40 81
April 2015 1.6 14.02 09.40 68
Table 2. Heat Loads for six Subway stations (Accord. BS

EN: 12831).
Total design Total design

Station name heat load (W) heat load (kW)
Kelvinbridge 4755 4.8

St. George’s Cross 5185 5.2
Cowcaddens 3369 3.4
Buchanan Street 3778 3.8
Kelvinhall 2599 2.6
Hillhead 6641 6.6
Figure 10. The same digital thermometer inside the sump

taking readings.
The heat energy H (in kW) can be calculated from
the following formula (Banks, D. 2009):
the same days where the readings were taken. InTable 1
are displayed the result from the sump that has been H = Q ∗ ρSvc ∗ t (1)
chosen for the trial. 3
Q: Water flow of the system (m /s)
ρ: Density of the water 1000 (kg/m3 )
Svc: Heat capacity of water 4.18 (kJ/kg ∗ K)
3.4 Subway station’s heat demand & heat energy t: Temperature difference (◦ C)
output from the water Example: If we take the average water flow from
May 2014 to April 2015 for sump 1 (Table 1), which
A constant water flow has been monitored in three is 3.07 l/s = 0.00307 m3 /s, we can estimate the output.
out of twenty one sumps. Those are: 1, 2 & 17A.
The following five stations are the closest ones to the H = 0.00307 ∗ 1000 ∗ 4.18 ∗ 4 = 51.33 kW
three sumps with the highest and constant water flux:
Buchanan Street, Cowcaddens, St. Georges Cross, St. Georges Cross station’s heat load is 5.2 kW
Kelvinbridge and Hillhead. (See Figure 3). (Table 2), so a selection of a system to provide heating
To assess where heat output can be delivered and and Domestic Hot Water (DHW) is feasible according
used, a heat load calculation (in kW) for all the Station to the heat energy output from sump 1.
has been completed (Table 2). The calculations have Table 3 is an example of the calculation of
been done according the BS EN: 12831-2003. St. Georges Cross station thermal needs according
St. Georges Cross subway station has been chosen BS EN 12831:2003.
for the pilot installation of a Water Sourced Heat Pump A basic design for this pilot installation has been
(WSHP). This station is in the closest proximity from undertaken and a water source heat pump (WSHP)
the source point (water sump no.1) to the sink point of 9 kW output is required to meet the Subway sta-
(station’s ticket office). tions heating and domestic hot water demand (DHW).
416
Table 3. Thermal needs of St. Georges Cross Subway station.
Transmission Ventilation Higher Heating-up Total design

heat load heat load temperature factor capacity heatload
St. George’s Cross T ,i V ,i f RH ,i HL,i
room name W W p.u. W W
Ticket-Office 845.62 215.37 1 239.20 1300.19

Hallway (Ground) 448.60 74.72 1 82.81 606.13
Hallway (Upper) 337.90 36.62 1 40.56 415.08
Canteen 987.77 232.25 1 257.40 1477.43
Female Toilet 276.31 78.83 1.6 58.24 448.32
Male Toilet 276.31 78.83 1.6 58.24 448.32
Store 377.76 31.36 1 80.21 489.33
Total 3550.28 747.98 – 816.66 5185
to accommodate the space heating and the domestic

hot water for the station.
This was the first installation which is expected to
prove the efficiency of the suggested system before
scaled up in the other stations in the Subway system.
A basic design for this system was undertaken and
the installation has been completed in November 2015.
A WSHP of 9 kW has replaced the station’s heat-
ing system to cover the heating demand and DHW
(Domestic Hot Water).
The old four electric radiators that were providing
only heating for this station have been replaced by four
new low temperature fan coil radiators (with a return
temperature of 45◦ C).
Figure 11. St. Georges Cross station drawing. This system is covering the space heating, the DHW
and is expected to provide as a by-product cooling
during summer months.
This case study will feed four new low temperature A COP for a return temperature of 45◦ C and inlet air
fan coil radiators which will be sufficient to heating temperature) of 5◦ C is around 3. This means that for
the station replacing the existing four electric radia- every kW of energy spent through the compressor the
tors (2 kW each) (Figure 11). This system is expected heat energy output is 3 kW. It is expected with the same
also to provide cooling as a by-product during summer return temperature of 45◦ C and inlet air temperature
months. of 13.5◦ C even during December to have a COP of
more than 4, which means that the energy consumption
will be reduced approximately 75% compared to the
existing system.
4 DISCUSSION & CONCLUSIONS The CO2 emissions are expected to be reduced more
than 60%.
Out of the 21 sumps inside the Glasgow subway
system; only in three there is a constant water flux.
Because of the position of Glasgow, north of the
NOMENCLATURE
UK, it is common that heating starts even from
September.
Water Source Heat Pump (WSHP), Domestic Hot
Currently the Subway stations are heated with elec-
Water (DHW), Carbon Dioxide (CO2), Coefficient of
tric radiators where the energy cost and the CO2
Performance (COP). British Standard (BS), European
emissions from the use of electricity are high.
Norm (EN), Kilowatt (kW).
A WSHP generally can perform with seasonal per-
formance factor (OFGEM) of 2.80 (Energy Saving
Trust).
The (Coefficient of Performance CoP) can be up to ACKNOWLEDGMENTS
4. The higher COP equates to lower operating costs.
It is feasible to install a WSHP in four Subway This research was funded through a Knowledge
stations to cover its thermal needs. transfer Partnership (KTP) scheme operated by Inno-
St. Georges Cross Subway station was chosen, and vate UK. The partnership members are Glasgow Cale-
the first WSHP has been installed in the ticket office donian University (GCU) and Strathclyde Partnership
417
for Transport (SPT). The authors would also like to Department of Energy & Climate Change, March 2013,
thank the following from the SPT: Mr. Gordon McLen- the future of Heating: Meeting the challenge. See
nan, Mr. Charles Hoskins and Mr. Stuart McMillan https://www.gov.uk/government/uploads/system/uploads/
whose belief in the approach described herein made it attachment_data/file/190149/16_04-DECC-The_Future_
of_Heating_Accessible-10.pdf
possible for us to carry out the work. Energy Saving Trust, our calculations, http://www.energy
savingtrust.org.uk/content/our-calculations
OFGEM, seasonal performance factor, https://www.
REFERENCES ofgem.gov.uk/key-term-explained/seasonal-performance-
factor-spf
Banks, D. 2009. An introduction to thermogeology ground Scottish Government 29 March 2012, Scotland beats 2011
source heating and cooling, Wiley-Blackwell, Second green energy target. See http://www.scotland.gov.uk/
Edition. 93–99 News/Releases/2012/03/geenenergytargets29032012.
Coefficient of Performance CoP, http://hyperphysics.
phy-astr.gsu.edu/hbase/thermo/heatpump.html
418
Potential of nano material (SiC) for improving the thermal

properties of sand
Z.H. Rizvi, K. Sembdner, A.S. Sattari & F. Wuttke

Geomechanics & Geotechnics, CAU Kiel, Germany
ABSTRACT: For the present study, two different SiC particles sizes with average diameter of 63 nm and
125 nm were mixed with medium sand and measurements were performed. Thermal conductivities of mixtures
were measured by the transient hot-wire technique. Classical models were used to predict the thermal conductivity
of the mixes. Test were performed to observe the improvement in mechanical properties such as shear strength
and consolidation. Finally, a numerical model is setup to observe the improvement in heat dissipation with the
modified sand SiC mix. The possible mechanism of the thermal conductivity enhancement in nanoparticle sand
mixture is discussed.
1 INTRODUCTION 2 EXPERIMENTAL
Soil thermal properties play an important role in many 2.1 Thermal conductivity measurements
engineering projects and heat transfer situation in the
Different measurement techniques of the soil thermal
soil i.e. roads, airfields, pipelines, buildings in cold
conductivity have been presented in detail[Mitchell &
regions and underground power cables. Thermal con-
Kao 1978]. These methods are broadly classified
ductivity of soil varies predominantly with type of soil,
as: steady heat flow method and transient heat flow
grain size distribution, soil texture, porosity, moisture
method. The steady heat flow method requires a long
content and temperature, etc. Porosity and the mois-
testing time to setup the thermal equilibrium. There-
ture content are the dominating factors on the thermal
fore, based on the transient theory, KD2 pro thermal
conductivity. Some studies [Ochsner & Horton 2011,
property analyzer is used for measurements (Decagon
Usowicz 1995] suggest that the thermal conductivity
Devices, Inc. USA). The sensor needle can be used for
is more closely correlated with air-filled porosity than
measuring thermal conductivity of soil in the range of
with volume fraction of water.
0.02–5 W/m-K with an accuracy of ±2%. Each mea-
In fine grain soils aggregate structure that forms
surement cycle consists of 900 s. During the first 60 s,
pore network with relatively large continuous and
the instrument will equilibrate; heating and cooling of
interconnected inter-aggregate pore spaces (10−3 to
sensor needle for 420 s each then follow. At the end
10−6 m) and very small intra-aggregate pore spaces
of the reading, the controller computes the thermal
between the textural grains (10−9 to 10−6 m) highly
conductivity using the change in temperature θ–time
influence the thermal conductivity. These inter and
data from
intra-aggregate pores store air or water in dry and sat-
urated conditions respectively [Carminati et al. 2008,
Horn et al 2005, McGarry et al. 2000]. To improve the
thermal conductivity, Nano material is suitable due to
its size, large surface to volume ratio and high ther-
mal conductivity to fill these inter and intra aggregate This instrument follows the ASTM D5334 and IEEE
pores. 442-1981 standards. For calibration of the probe
To our knowledge no comprehensive studies have thermal conductivity of distilled water, glycerin and
been performed to improve soil thermal conductivity ethylene glycol were measured. The measured values
using Nano material (SiC 125 63). Therefore, the aim for distilled water, glycerin and ethylene glycol were
of the work was to observe the enhancement in the 0.614, 0.289 and 0.257 W/m-K, respectively, which are
thermal conductivity of dry sand, which acts as a bottle in agreement with the literature values of 0.613, 0.285
neck in many engineering applications. and 0.252 W/m-K, respectively, within ±2% accuracy.
419
This equation is strictly applicable only when weight
percentage is small, since it was derived on the assump-
tion that the solid spheres are far enough apart that they
don’t mutually interact.
The Hamilton and Crosser [Hamilton & Crosser
1962] equation for computing the thermal conductiv-
ity of solid mixture for non-spherical particles was
applied. This model considers the effect of the shape
as well as that of the fraction percentage of the dis-
persed particle. This complicated model can predict
the thermal conductivity of solid suspensions contain-
Figure 1. Grain size distribution of Duofill sand. ing micrometer or millimeter-size particles. Therefore,
it is used to predict the behavior in dry case. It is
assumed that the sand is a continuous media and
Nano particles are evenly disperse. Homogeneity of
At each point at least three measurements were per- the mixture is achieved by thorough mixing and the
formed. The uncertainty of the calculated (measured) then validated with density tests. 10 samples were col-
thermal conductivity was calculated from the uncer- lected and measured from a mix type. If the results
tainty value of experimental data and was estimated to don’t match within 2% of error range it is again
be lower than 5%. The thermal conductivity calculated homogenized till it achieves the homogeneity criteria.
from experimental data was based on several assump- The equation is expressed as
tion such as the long heat source treated as an infinitely
long heat source and the medium is homogeneous,
isotropic, and at uniform initial temperature.
SiC/Sand mixture up to 5% weight percentage were
used for thermal conductivity measurements for two where φ is the volume fraction of the particle and n is
different types of SiC Nano particles (63 nm and the shape factor. It can be calculated empirically by
125 nm). The investigated sand, called Duofill sand,
has a Quartz content of more than 98.7% and an
air-dried moisture content of lower than 0.1%.
Particle-size distribution was determined by sieving where k is the sphericity, which is defined by the ratio
for sand (Figure 1) [ASTM D422]. SiC/Sand mix- of the surface area of a particle to that of a sphere with
ture with two different particle sizes were prepared. the same volume as the particle. The sphericity of a
The particle size between 125 micron to 63 micron is spherical particle is 1, while that of a cylindrical parti-
termed as 125 while size below 63 micron is tagged cle is 0.5. So the smaller the sphericity of the particle,
as 63. the larger the shape factor will be. Sphericities of 1
and 0.5 were used for SiC 63 and 125, respectively,
in the Hamilton and Crosser equation to calculate the
2.1.1 Theoretical models effective thermal conductivities.
At present, there is no theory to predict the thermal
conductivity of the sand nano-material mix. From the
experimental results of many researchers, it is known
that the thermal conductivity of nanomaterial depends 3 RESULT & DISCUSSIONS
on the surface area, shape of the nanomaterial &
the temperature [Murshed et al. 2008, Choi et al. Figure 2 shows some of the thermal conductivity data
2001]. It is also established that the thermal conduc- obtained for the Duofill sand with SiC 63 and SiC 125
tivity of soils depend upon various factor i.e. type varying percentages. It is apparent that the thermal
of soil, particle size distribution soil structure, poros- conductivity of Sand-SiC nanomix increases nonlin-
ity, saturation degree and temperature [Johansen 1975, early with increase in percentage of SiC content in the
Donazzi 2001]. mix. Significant improvement is observed in case of
Many equations have been proposed for the thermal SiC 63 with 5% weight content. SiC 63 increases the
conductivity of the thermal conductivity of a mul- thermal conductivity more than SiC 125 due to larger
tiphase soil mixtures [Gemant 1950, Russell 1935, surface to volume ratio of the Nanoparticles. Also, the
Eucken 1932]. However, validity and application of smaller size allows to achieve a higher density by fill-
these equations are limited to specific system. ing the inter and intra granular pores. Although, there
Maxwell’s [Maxwell 1881] equation of randomly is no theoretical model for the sand/SiC mix, Hamilton
distributed solid sphere (kSiC ) in a continuous soil and crosser model acts as lower and upper bound for
media (Ksand ) is applied the prediction of thermal conductivity. For SiC 63 in
Duofill sand mix, a thermal conductivity increase of
24.4% at a weight percentage of 5% is observed. Fig-
ure 2 provides an increase in thermal conductivity for
both sand mixtures with SiC 63 and SiC 125.
420
more heat to transfer, with higher grain to grain con-
tact. Sand/SiC 63 shows more improvement as finer
SiC particles are filling the intra granular voids and
improving the quality of contact among the grains.
4 CONCLUSION
To investigate the thermal properties of SiC/Sand mix-

ture, sand mixtures with two different SiC (63 and
125 nm) were prepared and the thermal conductivity
Figure 2. Percentage enhancement of thermal conductivity
of Duofill sand (DS) with weight percentage of SiC nano was measured.
particles (SiC-63 and SiC-125). The thermal conductivity measurements suggest
that the small amount of SiC particles improve the ther-
mal property of the sand. At low weight fraction range,
the thermal conductivity enhancing ratios increase lin-
early with weight fraction of solid particles. SiC with
63 micron particles size have a large surface area per
volume, therefore, an improvement in the effective
thermal conductivities of the mixtures is expected with
decreasing particle size.
The thermal measurement results have been com-
pared with the Hamilton & Crosser model. The model
behaves as a low and upper bound for most of the
measurements. A further investigation is required to
Figure 3. SEM image of showing two kinds of SiC (a) SiC develop a model for Sand/SiC mixture.
63 and (b) SiC 125.
ACKNOWLEDGEMENT
This research project is financially supported by

the research grant DuoFill provided by the Federal
Ministry for Economic Affairs and Energy, Germany
& ZIM by grant number KF3067303KI3.
Figure 4. SEM image of (a) Sand-SiC 63 and (b) Sand-SiC

125 showing thermal bridge formation in both mixes. REFERENCES
T.E. Ochsner, R. Horton, T. Ren, A new perspective on soil
thermal properties, Soil Sci. Soc. Am. J. 65 (2001) 1641–
1647.
In the calculation of Hamilton and Crosser model, B. Usowicz, Evaluation of methods for soil thermal conduc-
the thermal conductivities of sand was obtained from tivity calculations, Int. Agrophys. 9 (2) (1995) 109–113.
experimental results as 0.34 W/m.K Duofill sand. The A. Carminati, A. Kaestner, P. Lehman, H. Flühler, Unsatu-
rated water flow across soil aggregate contacts, Adv. Water
effect of boundary scattering of phonons and/or elec-
Res. 31 (2008) 1221–1232.
tron in micro and nano scale regime is well known. We R. Horn, A. Smucker, Structure formation and its conse-
assumed the thermal conductivity of SiC particles as quences for gas and water transport in unsaturated arable
490 W/m.K based on reported value in literature. and forest soils, Soil Till. Res. 82 (2005) 5–14.
As the heat transfer in dry soil is dominated by con- D. McGarry, B.J. Bridge, B.J. Radford, Contrasting soil phys-
duction through grain to grain contact mechanism, it ical properties after zero and traditional tillage of an
was expected that the heat transfer will be more with alluvial soil in semi-arid subtropics, Soil Till. Res. 53
the filling of the void with high conductive material. (2000) 105–115.
The advantage of nano particles is that it cannot only Mitchell J.K., Kao TC (1978) Measurement of soil thermal
resistivity. J Geotech Eng ASCE 101(GT10):1307–1320.
fill inter granular pores but also the intra granular pores
ASTM D 5334-00. Standard Test Methods for Determina-
thus increasing the thermal conductivity. tion of Thermal Conductivity of Soil and Soft Rock by
Scanning Electron Microscopy (SEM) images show Thermal Needle Probe Procedure. vol. (04)08.
that with addition of SiC 63/125 a thermal bridge ASTM, 100 Barr-Harbor Dr., West Conshocken, PA 19428-
is between the sand grains is formed. Thus allowing 2059, 2000.
421
IEEE STD 442-1981. IEEE Guide for Thermal Resistivity F. Donazzi, Soil thermal and hydrological characteristics in
Measurements, The Institute of Electrical and Electronics designing underground cables. Proc IEE 123:506–516
Engineers, Inc., 345 East 47 Street, New York, NY 10017 A. Gemant, Journal of Applied Physics. 21 750 (1950)
Standard Test Method for Particle-Size Analysis of Soils, H.W. Russell, Journal of American ceramics Society. 18 1
ASTM D422 – 63(2007) (1935)
S.M.S. Murshed, K.C. Leong, C. Yang Thermophysical and A. Eucken, Forsch. Gebiete Ingenieurw, B3, VDI-Forschung-
electrokinetic properties of nanofluids – a critical review sheft, 353 (1932)
Appl. Therm. Eng., 28 (2008), pp. 2109–2125 J.C. Maxwell “A Treatise on Electricity and Magnetism”
S.U.S. Choi, Z.G. Zhang, W. Yu, F.E. Lockwood, E.A. (second ed.) Clarendon Press, Oxford, UK (1881)
Grulke Anomalous thermal conductivity enhancement R. L. Hamilton and O. K. Crosser, “Thermal Conductivity
in nanotube suspensions Appl. Phys. Lett., 79 (2001), of Heterogeneous Two-Component Systems,” Industrial &
pp. 2252–2254 Engineering Chemistry Fundamentals,Vol. 1, No. 3, 1962,
O. Johansen Thermal conductivity of soils. PhD. Thesis, pp. 187–191
University of Trondheim, Trondheim, Norway (1975)
422
Bearing capacity and settlement of raft-pile foundations

under cyclic loading
Ilizar T. Mirsayapov & Marat I. Shakirov

Kazan State University of Architecture and Engineering, Kazan, Russia
ABSTRACT: Developed raft-pile foundation settlement calculating method under cyclic loading, considering
joint deformation of ground base, piles and raft grillage. Depending from the loading cycle, cyclic loading leads
to a redistribution of load between the elements of the raft-pile foundation, ground base at various levels from
the grillage raft foundation and ground massive in the space between piles.
This calculation method takes into account the specificity of raft-pile foundation elements stress-deformed
state and shows good agreement between the calculated and factual researched parameters values.
1 INTRODUCTION
In modern conditions at construction of buildings and

structures, trend of increasing loads to the ground base
and using as a base weak grounds contributed to the
fact, that one of the common ways to increase bearing
capacity and reduce settlement, using raft-pile founda-
tions. This foundations and their ground bases together
with static, exposed to cyclic loads, which in some
cases are the main, determine safety using of build-
ings and structures. Herewith the question of cyclic
loads influence to the plate-pile foundations behavior
researched not enough.
In this regard were conducted experimental and
theoretical researches of raft-pile foundations under
cyclic loading.
2 EXPERIMENTAL RESEARCHES
Laboratory studies were carried out in a steel tray with

the same length, width and height equal to 1.0 m. As
the raft in raft-pile foundation used reinforced con-
crete plate with dimensions 400 × 400 × 40 mm. Piles
modeled by hollow plastic tubes (compressive strength
R = 92.0 MPa; deformation modulus E = 700 MPa)
with a diameter equal to 7 mm, wall thickness 1 mm
and length 400 mm. As the ground base were used
sandy loam with a density ρ = 1.4 t/m3 and humid-
Figure 1. Efforts in piles diagram at different number of
ity W = 11%. Were fixed foundation and ground base cycles, H (Pmax = 1000 kg, ρp = 0.5).
settlement values at different points from plate, and
also deformations in pile models and ground base
during cyclic loading. In the system “raft-pile founda- deformation of all system elements takes place in
tion – the ground base” under cyclic loading jointly connected conditions.
deform materials with different strength and defor- When the load is applied to the plate grillage, at the
mation properties and loading conditions. Herewith load initial stages, takes place a ground compaction
423
At cyclic loading, bearing capacity reducing of
raft-pile foundation model based on the experimen-
tal researches results described by the equation of
regression:
Pst = bearing capacity of raft-pile foundation model

under static loading; N = number of cycles at the limit
Figure 2. Ground stress diagrams at different number of bearing capacity (Mirsayapov I.T. & Shakirov M.I.
cycles, kPa (Pmax = 1000 kg, ρp = 0.5). 2012).
The limit value of the cyclic load, sensing by
raft-pile foundation, depends from joint deformation
conditions of ground, piles and plate grillage, and their
strength and deformation properties.
throughout the depth of compressible thickness. Fur- Sress-strain pile base condition of raft-pile foun-
ther cyclic loading leads to an increasing of plastic dation is very complicated. In this bases, together
deformation zones, which are combined in the area, deform materials with a different strength and defor-
capturing all the entire upper part of the ground base. mation properties. Deformation development of pile
At cyclic loading takes place changing efforts in pile base under cyclic loading will take place in the con-
models. From piles effort diagram in (Fig. 1) can be ditions of ground and piles interaction in connected
seen, that with increasing number of cycles increase conditions:
efforts on the entire length of the piles. a) to the free deformation of the ground is constrained
Figure 2 shows the stress diagrams at different zones by piles;
in the space between piles. With the increasing of b) to the free deformation of the piles prevents the
the loading cycles number, occurs stress decreases in surrounding ground.
the ground at different levels from grillage, herewith
the greatest stresses are fixing directly under the plate. As a result of this interaction between the elements
To the free deformation of plate grillage ground in the pile base occurs is an additional stress state
base prevent piles, and the piles free deformation and take place a redistribution of efforts between the
limited by plate grillage ground base. As a result ground and the piles under cyclic loading. The stress
of such joint deformation take place a redistribution in the piles increases, but stress in the ground reducing
of forces between the raft-pile foundation elements between the reinforcing elements compare to the first
during cyclic loading. cycle.
Measured in the step cyclic loadings base settle- Then current stress in pile base are presented as:
ments, after different numbers of repeated load, change
similar as ground deformation in the space between
piles.
After various numbers of cycles, ground base settle-
ment changes analyze shows that the main settlement
increment occurs by increasing their residual part. The
settlement increase compare to the first loading cycle
σpmax (N1 ), σgr
max
(N1 ) = maximum stress at the first
up to 30%. (Mirsayapov I.T. & Shakirov M.I. 2014,
Mirsayapov I.T. & Shakirov I.F. 2015) loading cycle in the piles and ground respectively;
σpmax (N1 ), σgr
max
(N1 ) = additional stresses in the pile
base during the process of cyclic stressing in the piles
and ground respectively.
In the pile base, pile due to its grip along the side
3 FOUNDATION BASE CALCULATION
surface with surrounding ground becomes an internal
MODEL
communication, preventing to the free ground defor-
mation in the space between piles under cyclic loading.
Raft-pile foundation settlement changing patterns
Straitened vibrocreep ground deformations leads to
under cyclic loading can be expressed in a regression
the appearance in the pile base more balanced inter-
equation received by experimental researches:
nal stresses. In the ground occurs tensile stresses, and
in the piles – compression stresses. Under the vibro-
creep deformation difference influence between the
free ground and piles, straitened vibrocreep ground
deformation in the space between piles represented as
where S1 = raft-pile foundation settlement under static
load; Pmax = maximum cycle load; N = number of
cycles at the limit of bearing capacity.
424
εpl (N ) = additional (residual) vibrocreep pile base
gr
deformations; εpl (N ) = free ground vibrocreep defor-
p
mation; εpl (N ) = free pile material vibrocreep
deformation.
Then averaged additional tensile stresses in the
ground accepted:
Eo (N ) = ground deformation modulus under cyclic

loading;
For piles εpl (N ) elastic deformation, and therefore
there are compressive stresses:
Ep (N ) = the pile material elastic modulus.

Efforts equilibrium equations from the additional
stress state symmetrical pile base has the form:
Figure 3. Interaction design scheme of raft-pile foundation
with the ground massive.
where Ap = the total piles cross-sectional area

in the limit field of the foundation pile base;
Agr = foundation ground base area.
From (8) after a number of simplifications, obtained
analytical expressions for determining the additional
(residual) stresses:
– in the ground between the pile space
– in piles
Figure 4. Interaction design scheme of single pile with a

homogeneous ground massive with the sizes 2A × 2B.
gr
εpl = ground vibrocreep deformation; Ap1 = one
pile cross-sectional area; h = the total number of piles
in the calculated base area.
To simplify the calculation, accepted design scheme
which consists from the pile, surrounding it ground
4 EQUILIBRIUM AND MECHANICAL STATE and the part plate grillage, attributable to one pile. The
EQUATIONS main stress-strain state components behavior of such a
cell will match to behavior of the pile as part of raft-pile
For an analytical description non-free deformation foundation (Figs. 3, 4).
of the system elements, adopted calculation schemes The cell dimensions = 2A × 2B × L, pile
(Figs. 3, 4) and developed ground and pile-ground sys- dimensions = 2a × 2b × l. On the borders of a cell
tem mechanical state equations, as well as the efforts taken free vertical movement conditions.At the bottom
equilibrium equations. The joint solution of these of the cell accepted complete lack of movement.
equations allows to obtain the required settlement and To solve the task it is necessary to find the four
the bearing capacity values of raft-pile foundation unknown – p1 , p2 , p3 , and τ0 (Figs. 3, 4). Using the
under cyclic loading. forces equilibrium equations on the pile and the whole
425
diagrams (Fig. 5) at intersection point, which can be
calculated by the formula:
where c(N ) = ground cohesion under cyclic loading,

takes in accordance with [19]:
Figure 5. Mobilized shear stress and limit shear stress

diagrams.
cell, equality movements of ground and at the heel and Depending from the pile length, to heel level may
top piles level, obtain a system of equations: account different share of the load, since in the case of
increasing the length of the pile, increasing the lateral
surface area.
Stress occurring in the soil under the raft can be
found by the formula:
Stress level appearing at the top of the pile can be

expressed as follows:
Here:
E
G= – soil shear modulus;
2 · (l + v)
where k(l) = dimensionless coefficient taking into
account the effect of the depth of a rigid stamp applica-
tion on its length; ω = coefficient taking into account
the shape of the die; v = Poisson’s ratio; α = 5/l,
l = pile length.
Stress under the heel of the pile can be calculated
using the formula:
The solution of the system allows to find the settlement

value, as well as the stresses arising in the pile trunk,
in the ground under the grillage and in end of the piles Shear stress τ0 (N ) can be expressed as follows:
under cyclic loading.
Limit equilibrium zones, taking into account the
rigidity of the pile material, determined by the mobi-
lized shear stress (τ(N )) and the shear stress limit
426
Calculation results comparisons with the experi-
mental values are shown in figures 1, 2, 5, 6. As can be
seen from the figures, there is good agreement between
the calculated and experimental stress and settlement
values (deviation of no more than 15%).
REFERENCES
Mirsayapov I.T., Koroleva I.V. (2010). Features of the defor-
mation of clayey soils under cyclic triaxial compression.
Figure 6. Compare of experimental and calculated International journal of Geotechnics, No. 6. 64–67.
raft-pile model settlement values under cyclic loading Mirsayapov I.T., Shakirov M.I. (2012) Bearing capacity
(Pmax = 1000 kg, ρp = 0.5). and settlement patterns raft-pile foundations under cyclic
loading. Integration, partnership and innovation in build-
ing science and education: the International Collection of
Raft-pile foundation settlement can be calculated by papaers. Scien. Conf. 2 t. Vol. 2, 528–531.
the formula: Mirsayapov I.T., Shakirov M.I. Plate-pile foundations under
cyclic loading // Geotechnics Belarus: Science and Prac-
tice. 2013. P. 314–320.
Mirsayapov I.T., Shakirov M.I. (2014) Experimental study
of bearing capacity and the settlement of bases raft-pile
foundations under cyclic loading. Perspective directions
of development of the theory and practice of rheology
and soil mechanics. Proceedings of the XIV International
rheology Symposium, 68–74.
Ground base bearing capacity in depending from the Mirsayapov I.T., Shakirov M.I. (2014) Research of the cyclic
loading effect on model combined raft-pile foundation
ratio τ(N ) ≤ τ ∗ (N ) is estimated by conditions
construction. Forming environment of life: the Interna-
tional Collection of papaers. Scien. Conf., 423–429.
Mirsayapov I.T., Shakirov I.F. (2015) Selecting the type of
foundations and basements of multifunctional complex
“Fatikh, Amir and Khan” in the Fatikh Amirhan street in
Kazan. Journal News of the KSUAE, 86–92.
Mirsayapov I.T., Koroleva I.V. (2016) The strength and
Function σ1u (N ) accepted deformability of clay soils under the regime spatial stress
state in view of cracking. Grounds, foundations and soil
mechanics, No. 1. 16–23.
Voznesensky E.A. (1997) The behavior of ground under
dynamic loads. MGU, 286.
Zaretsky Yu.K. (1989) Lectures on modern soil mechanics.
Rostov-na-Donu: University Press, 607.
5 CONCLUSIONS
Represented calculation method takes into account

the specificity of raft-pile foundation elements stress-
deformed state under the action of cyclic loading.
427
Calculation model of foundation base settlement at the static

and cyclic regime loading
Ilizar T. Mirsayapov, Irina V. Koroleva & Danil D. Sabirzyanov

Kazan State University of Architecture and Engineering, Kazan, Russia
ABSTRACT: The article presents results of experimental and theoretical researches of the clay ground bases
soils under triaxial long-static, cyclic and regime static-cyclic loading. Described the source and transformed the
clay ground deformation diagrams for long-static, cyclic-static and modal prolonged cyclic loading. Test results
are presented and processed in the form of graphs and used for comparison with theoretical studies. Based on the
analysis of the transformed strain diagrams developed engineering method for calculating ground settlement,
based on the method of layering summation taking into account spatial changes of stress – strain state of the
ground in the process of regime static and cyclic loading. At the end of the article compares the results of triaxial
with a trough tests at long-static, cyclic and static and modal prolonged cyclic defor-mation of the clay base raft
foundation.
1 INTRODUCTION
In real conditions, the construction and operation of the

load on the foundation soil is applied in stages as the
construction of buildings and structures. At this stage
of the active loading during the construction stage
transformed into stages sequential alternation of long
static and cyclic loading, if the building or construction
equipment is installed, which create a dynamic impact
on the ground base. The maximum and minimum val-
ues of cyclic loading and prolonged static load depends
on the technology and mode of equipment operation.
However, existing methods of calculation can not to Figure 1. Dependence of vertical stress from vertical defor-
take into account the particular strain of foundations mation under triaxial short-term cyclic loading.
under such loadings.
Based on that, there is a need to develop a method
for calculating bases at regime-static cyclic loading. ρ = 1.94 g/cm3 ; Ip = 17.3%; IL = 33%. Samples twins
This is particularly relevant for the grounds composed were prepared in accordance with GOST 30416-96.
of clay grounds, as in this case, the stress – strain state According to the results of experimental studies
varies with time and depends on the previous history obtained graphic dependences between the intensity of
of the loading. Scheme strained substrate conditions tangential stresses τi and the intensity of shear strain
and calculation mode shown (Mirsayapov & Koroleva γi , between the axial stress σz and axial strain εz ,
2011). between the bulk stresses σm and volume deforma-
tions εv ; changes of linear module E0Z , bulk modulus
Kv and the shear modulus G with time (Mirsayapov &
Koroleva 2009, Mirsayapov & Koroleva 2010, Mir-
2 EXPERIMENTAL RESEARCHES sayapov, Koroleva & Sabizyanov 2013, Mirsayapov,
Koroleva & Sabizyanov 2014).
In connection with the foregoing at the Department Figure 7 shows the change in strength at various
“Bases and foundations, dynamics of structure lateral pressures of 0.08 MPa and 0.16 MPa. Graphs
and engineering geology” in KSUAE were car- show changes in the short-term strength of triaxial
ried out experimental studies of the clay ground static, long static, short-term cyclic loading points and
under triaxial static, cyclic and regime loading. showed the values strength at the regime long triax-
Were tested clay ground samples with the following ial static and cyclic loadings. The graphs show that
characteristics: W = 23%; Wp = 22.8%; WL = 40.1%; the strength of the clay samples at regime-static cyclic
429
Figure 5. Loading mode under triaxial regime prolonged
static and cyclic loadings.
Figure 2. Dependence of vertical deformation from number
of cycles under triaxial cyclic loading.
Figure 6. Dependence of vertical deformation under triaxial

prolonged static loading.
Figure 3. Dependence of vertical stress from triaxial pro-

longed static loading.
Figure 4. Dependence of vertical stress from time under

triaxial prolonged static loading.
loading is significantly different from the strength of

the long-term static and cyclic loading.
As can be seen from the graphs at viewed regimes
take place changing of all bases parameters, character-
izing the stress and strain state of the ground in time,
which allows to conclude about necessity to develop
a new parameter characterizing the mechanical condi-
tion and about necessity to create calculation models.
As this parameter is accepted analytical diagram of
ground deformation in the coordinates σ1 − ε1 Figure 7. Change the regime of loading strength: a) for the
for triaxial (where σ1 , ε1 vertical stress (deviator) and lateral pressure of 0.08 MPa; b) for the lateral pressure of
linear deformation under triaxial compression). 0.16 MPa.
430
Figure 8. Initial ground deformation diagram with
short-term static triaxial loading.
Figure 9. Original and transformed ground deformation
diagram long triaxial static loading.
3 ANALYTICAL DIAGRAM OF GROUND
DEFORMATION UNDER TRIAXIAL Ground creep deformation at given moment of time
COMPRESSION at long static loading are determined by the formula:
Based on the graphs (Figs. 1–7) were built initial dia-

grams (phase diagram) of ground deformation at short
triaxial static loading. As a limit point to the coordi- f (t1 t0 ) = 1 − e−γ(t − t0 ) = growth creep deformation
nates (σ) takes the value of ground tensile strength function; γ = ground creep parameter.
σ1 = Rgr,u (deviator) under triaxial short-term static
compression. Limit point on the y-axis (ε) takes the
value of linear deformation εu1 = 0.0869. View of
diagram shown in figure 8.
At clay ground state diagrams under triaxial long
static loading as initial used stress-strain diagram c∞ (t1 τ) = ground creep limit measure at time t.
σ1 − ε1 for the case of short-term triaxial static At viewing the clay ground state diagrams under
loading. Transforming the initial phase diagram under triaxial cyclic loading as initial were used stress-strain
triaxial short-term static loading, obtain analytical diagram σ1 − ε1 for the case of short-term triaxial
dependences for the description diagrams of the clay static loading. By transforming the initial state dia-
ground deformation under triaxial prolonged static gram under triaxial short-term static loading, obtain
loading. By the form of transformed diagram taken analytical dependences for the description of clay
similar initial phase diagram based on the following ground deformation diagrams under triaxial short-
provisions (Fig. 9): term cyclic loading. Transformed diagram form taken
similar to initial state diagram based on the following
– Limit point of the vertical pressure at the top of provisions (Fig. 10):
diagram accept stress in the ground, which is equal
to the limit prolonged resistance under triaxial load – Vertical pressure limit point takes from the top of
action Rgr,long − (t, τ) and deformation, answer to ground stress diagram, which is equal to the limit of
deformations at the top of the state diagram under endurance under triaxial cyclic loading Rgr,long (t, τ)
triaxial short-term static loading εgru,red = εgru ; and deformation answering to deformation at the
– To limit point, define the limits of state diagrams top of state diagram under triaxial cyclic loading
on the y-axis, deformations equal to limit defor- εgru,red = εgru ;
mations under triaxial short-term static loading – For limit point, which define the state diagram
εgr,red = εgr,R , and on major dependencies calculate limits at the Y-axis, deformation equal to limit
by (1) stresses in the ground; deformation under short-term triaxial cyclic load-
– The origins of diagrams taken offset to value, equal ing εgr,red = εgr,R , and stress in the ground calculate
to the creep deformation at the viewing moment of by the major dependencies (4);
time: εpl (t) – at long static loading; – Diagram start coordinates taken offset to value,
– Obtained tilt angles deformation diagrams taken in which equal to vibrocreep deformations at viewing
view of the changes of the clay ground module loading cycle: εpl (N ) – at short-term cyclic loading;
deformation under triaxial long static loading. – Obtained deformation diagram tilt angles, which
taken with changes of the clay ground deformation
The dependence of vertical pressure (deviator) lim- module under triaxial short-term cyclic loading.
iting point at the top of diagram and the loading time
under triaxial long static loading takes the form: The dependence between the limiting point of the
vertical pressure at the top of the state diagram and the
loading cycles number under cyclic loading takes the
form:
α = values obtained by experiments.
431
Figure 10. Initial and transformed ground deformation
diagram under triaxial cyclic loading.
β = values obtained by experiment; N = number of

cyclic loadings.
Ground creep deformation in viewing cycle at short
cyclic loading determine by the formula:
f (N ) = 1 − e−γ(N − N0 ) = growth creep function; γ = Figure 11. Initial and transformed ground deformation
ground creep parameter under cyclic loading; diagrams at regime alternating static and cyclic loading.
fatigue cracks in the equilibrium limit planes in the

calculated ground volume (Mirsayapov & Koroleva
c∞ (t1 τ) = limit ground creep measure. 2016).
Analytical dependences for describing ground Thus, at transforming the initial diagram for the
deformation diagrams at regime triaxial loading are case of regime loading must be taken into account
obtained by transforming dependencies (1–6) for tri- the influence of alternation long-term static and cyclic
axial long static and cyclic loading (Fig. 11). In the loads.
ground deformation regime description under triaxial In the limits of starting block, independently from
loading is necessary to consider the influence from ver- sequence from loading block, transformed diagrams
tical pressure (σ1 ) of the previous block for strength, obtain similarly to not regime long-static and cyclic
modulus and relative deformation at the top of the loading. In each subsequent block take place further
diagram in the subsequent loading after the regime deformation diagram transforming, but the initial dia-
change. Within each block, in which develops regime gram for each of them is an transformed diagram at
loading, changing deformation diagrams described by the end of the previous block.
the same formulas as for cycling or long – static load- Top coordinates of transformed diagrams at the
ing, but appropriate for each block σi , Ni or Ti and beginning of each subsequent block depends from the
taking into account the change in the strength and level of stress in the preceding block and the sequence
deformability of the previous blocks. alternation of blocks with different types of loading.
It is necessary to take into account two specific case
when:
– at first acting the long – static loading σ1 (t), and
4 CALCULATION MODEL OF FOUNDATION
then cyclic loading;
BASE SETTLEMENT
– at first acting cyclic loading σ1 (N ), and then pro-
longed static loading.
Based on the analysis of the transformed strain dia-
In the first case, the overall durability increases as grams developed the engineering method for cal-
compared with that, when cyclic loading is being with culating ground base settlement, which based on
the stress value σ1 (N ). This is due to with increasing the method of layering summation with taking into
the effective destruction of the ground surface energy account changes in the spatial stress – strain state of
on the first stage load due to recovery and the increas- ground in the process of triaxial long-static, cyclic and
ing of coagulation bonds (Mirsayapov & Koroleva regime static – cyclic loading.
2016). Accepted volumetric stress state of foundation base
In the second case, strength reduction at the tran- ground presented in figure 12. Breaking base com-
sition to the block with prolonged static load will be pressible thickness to the layers, for each layer based
extremely slow. This is explained by “delay effect” of on (Fig. 12) for the stress deviator (σ1 ) determine the
432
Figure 12. Stress state scheme of the foundation base during
the regime loading.
Figure 14. Raft foundation comparing settlements graphs:

a) at long static loading; b) under cyclic loading; c) at the
regime-static-cyclic loading.
layer; εz,i (t, τ) = axial strain increment of i-th layer

under the static load action; t = time appropriate to
the moment of observation; t0 = application time of
the load.
– at triaxial cyclic loading:
εz,i (N ) = axial strain increment of i-th layer under

cyclic loading action;
Figure 13. Calculation scheme for determining ground base
settlement during the regime loading. – under triaxial regime prolonged static and cyclic
loading:
deformation εz,i appropriate to value of vertical pres-
sure, and then the deformation values at the limits of
the compressible thickness summed up.
Thus, the base settlement calculate according to the
formulas:
m = the number of cyclic loading blocks; k = the
– with triaxial prolonged static loading: number of long-term static loading blocks.
Comparing theoretical and experimental researches
results graph of raft foundation model settlement (with
dimensions of 0.4 × 0.4 m) on clay ground base in
a laboratory tray (tray with the size of 1 × 1 × 1 m)
n = the number of layers in which divide compress- the ground characteristics that defined above, for the
ible base thickness; hi = the thickness of the i-th triaxial loading conditions are shown in figure 14.
433
5 CONCLUSIONS Mirsayapov I.T., Koroleva I.V. (2009). Research strength and
deformability of clay soils with prolonged triaxial com-
Based on the analysis of the transformed strain dia- pression. Proceedings of the Kazan State Architectural
grams developed the engineering method for cal- University. No. 2 (12), 167–172.
Mirsayapov I.T., Koroleva I.V. (2010). Features deforma-
culating ground base settlement, which based on
tion of clayey soils under cyclic triaxial compression.
the method of layering summation with taking into International Journal of Geotechnical. No. 6, 64–67.
account changes in the spatial stress – strain state of Mirsayapov I.T., Koroleva I.V., Sabizyanov D.D. (2013).
ground in the process of triaxial long-static, cyclic and Strength and deformation of clayey soils under triaxial
regime static-cyclic loading. modal alternating static and cyclic loading. Geotechnics
As seen from figure 14, comparing the results of tri- Belarus: Science and Practice, 297–304.
axial with a trough tests at long-static, cyclic and static Mirsayapov I.T., Koroleva I.V., Sabizyanov D.D. (2014).
and modal prolonged cyclic deformation of the clay Warp clay soils combined with the regime and the long-
base raft foundation shows a good agreement between term cyclic loading. Perspective directions of development
of the theory and practice of rheology and soil mechan-
the calculated and experimental settlement data.
ics. Proceedings of the XIV International Symposium on
rheology, 130–135.
Mirsayapov I.T., Koroleva I.V. (2016). The strength and
REFERENCES deformability of clay soils under the regime spatial stress
state in view of cracking. Grounds, foundations and soil
Mirsayapov I.T., Koroleva I.V. (2011). Bearing capacity of mechanics, No. 1, 16–23
foundations and rainfall with prolonged loading. Integra- Voznesensky E. A. (1997). The behavior of soils under
tion, partnership and innovation in building science and dynamic loads. MGU, 286.
education: the International Collection of papers. Scien. Zaretsky Yu.K. (1989). Lectures on modern soil mechanics.
Conf. 2 t. Vol. 2, 342–347. Rostov-na-Donu: Pabl growth. University Press, 607.
434
The role of high-pressure flow-through experiments for evaluating

the mechanical behaviour of gas hydrate-bearing soils
C. Deusner, E. Kossel, N. Bigalke & M. Haeckel

GEOMAR Helmholtz Centre for Ocean Research Kiel, Kiel, Germany
S. Gupta
Technical University Munich, Garching bei München, Germany
M. Freise, H. Anbergen & T. Wille

APS GmbH Wille Geotechnik, Rosdorf, Germany
ABSTRACT: Results from two recent field trials, onshore in the Alaska permafrost and in the Nankai Trough
offshore Japan, suggest that natural gas could be produced from marine gas hydrate reservoirs at compatible
yields and rates. However, both field trials were accompanied by different technical issues, the most striking
problems resulting from un-predicted geomechanical behaviour, sediment destabilization and catastrophic sand
production. So far, there is a lack of experimental data which could help to understand relevant mechanisms and
triggers for potential soil failure in gas hydrate production, to guide model development for simulation of soil
behaviour in large-scale production, and to identify processes which drive or, further, mitigate sand production.
We use high-pressure flow-through systems in combination with different online and in situ monitoring tools
(e.g. Raman microscopy, MRI) to simulate relevant gas hydrate production scenarios. Key components for soil
mechanical studies are triaxial systems with ERT (Electric resistivity tomography) and high-resolution local-
strain analysis. Sand production control and management is studied in a novel hollow-cylinder-type triaxial
setup with a miniaturized borehole which allows fluid and particle transport at different fluid injection and
flow conditions. We further apply a novel large-scale high-pressure flow-through triaxial test system equipped
with µ-CT to evaluate soil failure modes and triggers relevant to gas hydrate production and slope stability. The
presentation will emphasize an in-depth evaluation of our experimental approach, and it is our concern to discuss
important issues of translating laboratory results to gas hydrate reservoirs in nature. We will present results
from high-pressure flow-through experiments which are designed to systematically compare soil mechanical
behaviour of gas hydrate-bearing sediments in relevant production scenarios focusing on depressurization and
CO2 injection. Experimental data sets are analyzed based on numerical models which are able to simulate coupled
process dynamics during gas hydrate formation and gas production.
1 INTRODUCTION permafrost (Schoderbek et al., 2013) and in the Nankai

Trough offshore Japan (Yamamoto, 2013). Both tests
In marine sediments and permafrost soils gas hydrates were accompanied by substantial sediment yielding
can be present in large amounts and at high relative sat- and sand production events, which were, however, fol-
urations. Gas hydrate formation can occur whenever lowing different patterns. In the onshore field test,
pressure-temperature (p-/T-) conditions are inside the natural gas production was initiated after N2 :CO2
gas hydrate stability region, and gas hydrate forming mixed gas injection and gas-hydrate exchange, and
components such as CH4 or CO2 are available in suf- further enhanced through stepwise depressurization.
ficient amounts. Dependent on gas hydrate saturation In this field test, sand production was initially very
and gas hydrate-sediment particle fabrics, the presence high. Substantial technical problems occurred, includ-
of gas hydrates contributes to marine sediment stiff- ing failure of technical components and extended
ness and strength, and it is important to understand production shutdown intervals. However, sand pro-
and quantify this contribution for reliably predicting duction rates decreased with ongoing gas production.
sediment mechanical behavior and assessing risks of In the first offshore field test in the Nankai Trough,
ground subsidence or slope failure. gas production was achieved through depressuriza-
The lack of knowledge and simulation tools to tion only. In contrast to the results from the onshore
predict dynamic mechanical behavior of gas hydrate- permafrost test, sand production was low during early
bearing sediments became apparent in recent gas production and could be managed with applied stan-
hydrate production field trials, onshore in the Alaska dard technical means. However, after a few days,
437
sudden and catastrophic sand production occurred, and and Yamamoto 2015, Santamarina 2015, Yoneda et al.
the production test had to be canceled pre-maturely. 2015). Numerical gas hydrate reservoir simulators
The reasons for this distinct test specific behavior are have been used to study coupled processes during
currently unknown and might be related to numer- and after gas production from gas hydrates, and there
ous factors, such as site and reservoir characteristics, is strong effort to build-in soil mechanical constitu-
geological heterogeneities, gas hydrate alteration and tive models and improve model couplings. Recently, a
dissociation dynamics, technical issues, or dynamic gas hydrate reservoir simulation algorithm was devel-
thermo-hydro-chemo-mechanical process coupling. oped and applied for modelling particle mobilization
Gas hydrates contribute to sediment or soil mechan- and sand production as observed in the field trial
ical behavior in various ways, and effects might be offshore Japan (Uchida et al. 2015). A novel hydro-
defined as primary and secondary, respectively. Pri- geo-mechanical gas hydrate simulator was developed
mary effects result from direct and quasi-static interac- recently at TU Munich (Gupta et al. 2015). It was cali-
tion of gas hydrates and soil particles, and define sed- brated and tested by matching experimental data from
iment behavior in terms of soil stiffness and strength. high-pressure flow-through triaxial experiments sim-
The understanding of geotechnical behavior of gas ulating gas hydrate formation and dissociation under
hydrate-bearing sediments in this sense has improved isotropic compression at variable total and effective
tremendously in recent years, and with our current stresses and with dynamically changing gas hydrate
understanding, gas hydrates contribute to sediment saturations.
strength by changing friction and dilatancy, Soil criti- Using high-pressure flow-through triaxial experi-
cal state models were successfully applied to simulate ments in combination with numerical simulation is
soil plastic yielding (Klar et al. 2013). Today, the avail- the most direct and promising approach to understand
ability of experimental tools and methods to analyze multiphysics coupling and to improve physical process
and visualize micro-structuring of gas hydrate-bearing knowledge. High-pressure flow-through experimental
sediments helps to develop physical understanding systems, and different online and in situ monitoring
and improve constitutive models of bulk sediment tools have been successfully applied for testing various
behavior. For example, recent experimental stud- strategies for gas production from gas hydrate-bearing
ies on microscale gas hydrate-sediment structures sediments (essentially depressurization, thermal stim-
have demonstrated the presence of a water layer ulation and gas hydrate exchange after injection of
between gas hydrates and quartz sand particles inde- CO2 or CO2 -rich mixed gases). We used our NESSI
pendent of gas hydrate formation methods (Chaouachi system (N atural Environment Simulator for Sub-
et al. 2015), which has implications for understand- seafloor I nteractions) to study different CO2 injection
ing strength of gas hydrate-bearing soil in terms schemes and to improve gas hydrate exchange and
of assumed cohesion and cementation. In contrast natural gas production from gas hydrate-bearing soils
to primary effects, secondary, non-direct mechanical (Deusner et al. 2012). The NESSI system is used in
effects result from strong and dynamic thermo-hydro- combination with Raman spectroscopy and IR-based
chemo-mechanical process coupling. This definition gas analysis. These tools allow time-resolved moni-
emphasizes that under relevant non-equilibrium condi- toring of multiphase fluid flow (e.g. water, gas, liquid
tions with dynamic gas hydrate formation, alteration or CO2 ) as well as phase compositions (e.g. pure or
dissociation, sediment mechanical stability becomes dissolved compounds, mutual solubilities). Tomogra-
eventually dominated by multiphysics parameters such phy techniques have been applied for disturbance-free
as hydraulic properties (e.g. absolute and relative and high-resolution analysis of process dynamics and
permeability changes) or flow dynamics (e.g. mul- influence of heterogeneities in gas-hydrate bearing
tiphase fluid flow, gas migration or holdup). Quan- soils. Magnetic resonance imaging (MRI) was applied
titative effects emerging from multiphysics coupling to study phase distributions and permeability changes
are extremely complicated to predict, and coupled during gas hydrate formation (Kossel et al. 2014)
secondary effects could easily overprint simplified and gas hydrate exchange after injection of CO2 .
sediment stability and failure predictions based on Electrical resistivity tomography (ERT) was applied
quasi-static primary effects. to map fluid and solid phase distributions (Prieg-
It is an important task to reveal and disentan- nitz et al. 2013). Further, X-ray CT was applied to
gle multiphysics effects, understand process coupling visualize gas hydrate sediment structures and analyze
and resolve mechanistic details, and this importance permeabilities (Kneafsey et al. 2014, Kneafsey et al.
is increasingly recognized. For example, large-strain 2011).
deformation under deviatoric loading, and dynamic To better understand dynamic and coupled thermo-
stress-strain behavior during depressurization or ther- hydro-chemo-mechanical processes relevant to gas
mal stimulation was studied in triaxial experiments hydrate-bearing soils, we have developed novel high-
(Hyodo et al. 2013, 2014, Song 2014, Ghiassian and pressure flow-through triaxial systems for large sam-
Grozic 2013). In the very recent past there have also ples. The novel systems are used in combination with
been the first attempts to carry out studies on undis- µCT or ERT, respectively, high-resolution local-strain
turbed pressure cores, which will clearly advance the measurement and continuous fluid composition mon-
field towards a much better understanding of the itoring with flow-through sensors, as described above.
mechanics of the gas hydrate-bearing soils (Inada One system is equipped with a miniaturized perforated
438
borehole, which allows passage and sampling of both de-pressurized to atmospheric pressure and remain-
fluids and solids. Major research objectives are to ing CH4 gas in the pore space was released. System
experimentally simulate coupled processes relevant to re-pressurization and water saturation of pore space
gas-charged or gas hydrate-bearing sediments in the was achieved by instant filling and re-pressurization
context of slope stability and natural gas production. with pre-cooled (−1◦ C) saltwater medium according
The experimental systems allow inducing and mon- to seawater composition. Hydrate dissociation during
itoring large-strain visco-elasto-plastic deformation, the brief period of depressurization was minimized by
particle fluidization and sand production on different taking advantage of the anomalous self-preservation
scales (pore-scale to bulk scale). Experimental data effect, which reaches an optimum close to the chosen
are used to develop and test numerical codes, and to temperature (Stern et al. 2003). After completion of
define constitutive models based on high-resolution gas – water fluid exchange, the sample temperature
micro-scale data. was re-adjusted to 2◦ C.
Here, we present details about the novel triaxial
flow-through systems which were developed for stud-
ies of gas hydrate-bearing or gas charged soils and sed- 2.3 Flow-through experiments
iments, but might also be suitable for a range of related
scientific topics in geomechanics and geotechnics. Experiments were carried out in the custom-made
We further present results from experimental and high pressure apparatus NESSI (N atural Environment
numerical studies on depressurization, CO2 - or CO2 - Simulator for Sub-seafloor I nteractions, Deusner et al.
mixed fluid injection and gas hydrate exchange from 2012), which was equipped with a high-pressure tri-
high-pressure flow-through experiments. We briefly axial cell mounted in a 40 L stainless steel vessel. All
define upcoming test cases. Preliminary test data with wetted parts of the setup are made of stainless steel.
triaxial-CT and triaxial-ERT systems will be presented Saltwater medium was supplied from reservoir bottles
during ICEGT 2016. (DURAN, Wertheim, Germany) using a HPLC pump
S1122 (SYKAM, Fürstenfeldbruck, Germany). Pres-
sure was adjusted with a back-pressure regulator valve
2 MATERIAL AND METHODS (TESCOM Europe, Selmsdorf, Germany). Experi-
ments were carried out in upflow mode with injection
2.1 Sample preparation and mounting of CH4 gas and seawater medium at the bottom of
the sample prior and after gas hydrate formation, and
Sediment samples were prepared from quartz sand fluid discharge at the top of the sample during depres-
(initial sample porosity 0.35, grain size 0.1–0.6 mm, surization. Axial and confining stresses, and sample
G20TEAS, Schlingmeier, Schwülper, Germany), and volume changes were monitored throughout the over-
mixed with deionized water. The partially water sat- all experimental period. Pore pressure was measured
urated and thoroughly homogenized sediments were in the influent and the effluent fluid streams close to
filled into the triaxial sample cell equipped with sample top and bottom. The experiment was carried
a Viton sleeve to obtain final sample dimensions out at constant temperature conditions. Temperature
of 380 mm in height and 80 mm in diameter. Sam- control was achieved with a thermostat system (T1200,
ple geometry was assured using a sample forming Lauda, Lauda-Königshofen, Germany). Produced gas
device. The sample was cooled to 2◦ C after the triax- mass flow was analyzed with mass flow controllers
ial cell was mounted inside the pressure vessel. Initial (EL FLOW, Bronkhorst, Kamen, Germany). For con-
water permeability of gas hydrate-free sediment was trol purposes, bulk effluent fluids were also collected
50 × 10−11 m2 . inside 100 L gas tight TEDLAR sampling bags (CEL
Scientific, Santa Fe Springs CA, USA). The sam-
pling bags were mounted inside water filled sampling
2.2 Gas hydrate formation containers. After expansion of the effluent fluids at
atmospheric pressure, overall volume was measured
Prior to gas hydrate formation the sediment sam- as volume of displaced water from the containers.
ple was isotropically consolidated to 2 MPa effec-
tive stress under drained conditions. The sample was
flushed with CH4 gas and, subsequently, pressurized
with CH4 gas. During pressurization with CH4 gas and 2.4 Numerical modelling
throughout the overall gas hydrate formation period, To simulate gas hydrate formation and dissociation
formation effective stress conditions were maintained processes in the lab-scale triaxial compression experi-
using an automated control algorithm. The formation ment described in Section 2, we use the mathematical
process was continuously monitored by logging CH4 model and the numerical simulator developed by
gas pressure. Mass balances and volume saturations Gupta et al. (2015). This model considers kinetic
were calculated based on CH4 gas pressure and ini- hydrate phase change and non-isothermal, multi-
tial mass and volume values. After completion of gas phase, multi-component flow through porous medium.
hydrate formation, the sample was cooled to −5◦ C and The model accounts for the effect of hydrate phase
stress control was switched to constant total isotropic change on the mechanical properties of the soil, and
stress control before the sample pore space was also for the effect of soil deformation on the fluid-solid
439
interaction properties relevant to reaction and transport
processes (e.g., reaction surface area, permeability,
capillary pressure).
3 NOVEL EXPERIMENTAL SYSTEMS
3.1 Triaxial-CT
3.1.1 CT system
In close collaboration APS and GEOMAR develop
a new kind of triaxial test system combined with
computer tomography. In contrast to existing triax-
ial CT systems, the triaxial cell is fixed and the
CT scanner itself moves around the sample. Due to Figure 1. 2D segmented images and volume rendered plot
a high-precision alignment system, high-resolution of a respective sample using unsupervised networks (after
tomography becomes feasible. As the cell is not in Chauhan et al. 2016).
motion, the accuracy of the permeability and shear
tests performed increases substantially compared to
“moving cell” solutions. The thermodynamic pro- 3.2 Triaxial-ERT
cesses of the gas hydrate formation and dissociation
can be observed, as well as CO2 injection and cap- The system combines ERT and local deformation mea-
turing. The system is designed for cell pressures up surements (Fig.2). The triaxial test unit is designed for
to 40 MPa. Therefore a special cell was constructed mounting of large samples (diameter 150 mm, height
that allows the application of the cell pressure while maximum 400 mm) and can be operated up to 40 MPa.
remaining transparent for the X-ray tomography. As Coupled fluid flow – mechanical loading tests can be
the system is currently in the patent process, further performed with a perforated central well. Thus, sce-
information and first testing results will follow soon. narios of flow- or load-induced deformation and sand
production can be simulated. The loss of sample solids
through visco-plastic yielding or particle migration
leads to local straining rather than homogeneous bulk
3.1.2 Evaluation Software deformation. Heterogeneous deformation is moni-
Besides the experimental part of the CT-tests, the tored at high-resolution using electromagnetic sensors.
evaluation of the acquired data is a crucial task of ERT is applied to simultaneously acquire information
the project. Commercial and open-source software about heterogeneous phase distributions (e.g. spatial
solutions are available, but they are limited in the gas hydrate saturation, zones of gas hydrate dissocia-
meaning of objectivity. For the evaluation a manual tion and gas release, gas migration pathways, fractures,
definition of the distinct phases (water, matrix, air, gas holdup regions, etc.). Combining tools, bulk sam-
gas hydrates) is needed. The user defines the phases ple behavior can be correlated with physically relevant
according to his experience and thus with eventual heterogeneous processes.
errors and mistakes. In order to ensure an objective
definition of the phases a new program was devel-
oped based on machine learning technology (Chauhan
et al. 2015, Chauhan et al. 2016). The algorithm learns
from the data sets itself which phases are tomographi-
4.1 Depressurization experiments
cally recorded. The phases are clustered and evaluated
by their volumetric content. As there are several pos- Depressurization of gas hydrate-bearing reservoirs is
sibilities of suitable solver-algorithms, a module is the most mature approach for natural gas produc-
incorporated that allows to select and to compare the tion from gas hydrates. Depressurization refers to
results of different solver types. Depending on the a technical decrease in hydrostatic or pore pressure
specimen structure, the results of the different solvers at a well using pumps. Obviously, depressurization
scatter, in some cases substantially. The results of the changes effective stresses, induces fluid flow and
machine learned analysis are visualized in 2D or 3D phase changes, and gas hydrate saturations are effected
(see Fig. 1). Applying the histogram method, the volu- relative to gas hydrate stability conditions. After shut-
metric distribution of the phases is calculated. Besides down of pumps, hydrostatic pressure will recover, with
the absolute value of phase volumes, an evaluation recovery rate being dependent on multiple factors
of the pore-size distribution is implemented as well. including geological and sedimentological settings,
The constructed pore model can be exported from the permeability and reservoir dimensions.
program for subsequent pore network and flow sim- First triaxial experiments on dynamic gas hydrate
ulations, as well as for further elaboration and finite formation and dissociation were done without allow-
element analysis. ing particle flow (Fig.3), i.e. sediment particles were
440
often irreversible blocking. For similar reasons, near-
well flow assurance might become a major issue in
field application, and flow management and filtra-
tion means must be carefully chosen. The triaxial
experiments were focused on dynamically altering gas
hydrate saturations during gas hydrate formation or
gas hydrate dissociation and dissociation induced gas
flow in a water saturated sample. Pressure and load-
ing constraints were chosen to mimic gas hydrate
formation in weakly consolidated sediment which
is confined by low-permeability layers at its upper
boundary. Thus, usually effective stress is controlled
during gas hydrate formation, and total stress is con-
trolled during depressurization (Fig.4). Experimental
data from isotropic compression experiments were
used to calibrate a fully coupled numerical simulator,
soil constitutive behavior was defined in the frame-
work of poro-elasticity. CompositeYoung’s modulus of
gas hydrate-bearing sediment was modeled with addi-
tive soil and gas hydrate contributions, and it was found
that composite modulus depends almost linearly on Sh
during gas hydrate formation, while during the hydrate
dissociation period the dependence of Esh on Sh is
smaller. Further experiments at different gas hydrate
saturations have been carried out (Fig.5), and experi-
mental data are now used to further develop the model
approach and parameterization.
4.2 Gas hydrate-exchange experiments

The application of gas-hydrate exchange for natural
gas production was discussed extensively in recent
years, advantages essentially seen as being an emission
neutral energy production technology and as a cure
of depressurization induced process problems such as
sediment cooling from endothermic gas hydrate for-
mation. It was considered that gas hydrate exchange
could also contribute to mechanical integrity during
gas hydrate production by keeping gas hydrate-soil
fabrics untouched. Although this certainly is a very
idealized assumption, results from the Ignik Sikumi
field test suggest, that gas hydrate exchange might at
least partly prevent excessive sand production. At the
moment, the mechanical consequences of CH4 -CO2 -
hydrate exchange, even under most ideal assumptions
are not understood, and different injection strategies
and reaction proceedings (gaseous, liquid or super-
critical CO2 injection, N2 :CO2 injection, continuous
or discontinuous injection, mixed gas hydrate forma-
tion, etc.) will influence mechanical behavior in a
very complicated and coupled way. The role of gas
hydrate conversion for sediment mechanics is a very
central topic for applying the new triaxial systems with
Figure 2. Scheme of high-pressure flow-through triaxial
tomography.
test system with electromagnetic sensors for local strain
analysis, ERT and perforated well.
4.3 Important test cases

retained by the use of filter plates. The obligatory
use of filter plates was problematic, since gas hydrate 4.3.1 Slope stability
formation, water freezing or multiphase fluid flow Sub-marine slope stability is affected by different
caused substantial problems with permeability and factors, and slope failure and sub-marine avalanches
441
Figure 3. Experimental scheme for depressurization exper-
iments.
Figure 4. Converted data from dynamic gas hydrate forma-

tion and dissociation experiments.
could be triggered by different mechanisms. The influ- Figure 5. Raw data from dynamic gas hydrate formation
ence of gas migration, gas holdup, and gas charging and dissociation experiments with different initial gas hydrate
of sediments is long recognized as an important fac- saturations.
tor, although mechanistically many unknowns remain.
The impact of gas hydrates in slope stability is less
well defined. As a primary effect, gas hydrates tend 4.3.2 Sand production
to increase sediment shear strength and could oppose The mechanisms and progress patterns of sand migra-
destabilization. However, in a dynamic marine setting tion observed in field trials are not understood, and the
with processes as active gas ascent and seeping, or gas phenomenon could potentially be explained by tran-
hydrate dissociation the presence of gas hydrates could sient particle fluidization or plastic flow. To improve
well be destabilizing (e.g. by enabling pore pressure the understanding of sand production, and to define
increase from acting as a low permeability barrier and criteria and triggers for catastrophic soil failure, sam-
causing sediment effective unloading, gas release from ples of gas hydrate-bearing sediments with different
gas hydrate dissociation or defining failure planes). To grain size distributions and gas hydrate saturations
improve the understanding of mechanical effects of will be investigated under different deviatoric load-
gas charging, gas migration and gas hydrate dissocia- ing and flow conditions. Tomographic systems will be
tion for slope stability, the new flow-through triaxial used to analyze phase distributions and soil-hydrate
systems with tomography will used to simulate rele- fabrics on the small (mm to cm) to micro-scale (few
vant gas flow and gas release dynamics and investigate µm) in order to define zones and progression of ini-
mechanical response on micro- and bulk scale. The tial disturbance, and to dynamically monitor transition
objective is to improve coupled process understanding of initial local disturbance to bulk failure. Certainly,
and constitutive laws used for mechanical modeling. studies to improve mechanistic understanding of sand
442
production will be extended to study technical means hydrate-bearing sand in deep seabed. Soils and Founda-
of sand production management (e.g. application of tions 53, 299–314.
sand screens or improved depressurization schemes Inada, N. & Yamamoto, K. 2015. Data report: Hybrid Pres-
avoiding peak shear loading on the particle scale). sure Coring System tool review and summary of recovery
result from gas-hydrate related coring in the Nankai
Also, the effect of different injection and gas hydrate Project. Marine and Petroleum Geology 66, 323–345.
formation and exchange schemes will be tested. Klar, A., Uchida, S., Soga, K., &Yamamoto, K. 2013. Explic-
itly Coupled Thermal Flow Mechanical Formulation for
Gas-Hydrate Sediments. Spe Journal 18, 196–206.
ACKNOWLEDGEMENTS Kneafsey, T. J. & Moridis, G. J. (2014). X-Ray computed
tomography examination and comparison of gas hydrate
This work was further funded by the German Fed- dissociation in NGHP-01 expedition (India) and Mount
eral Ministries of Economy (BMWi) and Education Elbert (Alaska) sediment cores: Experimental observa-
and Research (BMBF) through the SUGAR project tions and numerical modeling. Marine and Petroleum
Geology 58: 526–539.
(grant No. 03SX250, 03SX320A & 03G0856A), and Kneafsey, T. J., Seol, Y., Gupta, A., & Tomutsa, L., Perme-
by DEA Deutsche Erdoel AG. The research leading to ability of Laboratory-Formed Methane-Hydrate-Bearing
these results has received funding from the European Sand: Measurements and Observations Using X-Ray
Union Seventh Framework Programme (FP7/2007- Computed Tomography. Spe Journal 16: 78–94.
2013) under the MIDAS project, grant agreement n◦ Kossel E., Deusner C., Bigalke N., & Haeckel M. 2014.
603418. We gratefully acknowledge the support for Experimental investigation of water permeability in quartz
S. Gupta by the German Research Foundation (DFG), sand as function of gas hydrate saturation. Conference
through project no. WO 671/11-1. Proceedings of the 8th International Conference on Gas
Hydrates, Peking, China.
Priegnitz, M., Thaler, J., Spangenberg, E., Ruecker, C., &
Schicks, J. M. 2013. A cylindrical electrical resistivity
REFERENCES tomography array for three-dimensional monitoring of
hydrate formation and dissociation. Review of Scientific
Chaouachi, M., Falenty, A., Sell, K., Enzmann, F., Kersten, Instruments 84.
M., Haberthuer, D., & Kuhs, W. F. 2015. Microstruc- Santamarina, J. C., Dai, S., Terzariol, M., Jang, J., Waite, W. F.,
tural evolution of gas hydrates in sedimentary matrices Winters, W. J., Nagao, J., Yoneda, J., Konno, Y., Fujii, T., &
observed with synchrotron X-ray computed tomographic Suzuki, K. 2015. Hydro-bio-geomechanical properties of
microscopy. Geochemistry Geophysics Geosystems 16: hydrate-bearing sediments from Nankai Trough. Marine
1711–1722. and Petroleum Geology 66, 434–450.
Chauhan, S., Rühaak, W., Khan, F., Enzmann, F., Mielke, Schoderbek, D., Farrell, H., Hester, K., Howard, J., Rater-
P., Kersten, M., & Sass, I. 2016. Processing of rock core man, K.,Silpngarmlert, S., Martin, K., Smith, B. & Klein,
mi-crotomography images: Using seven different machine P. (October 1, 2008–June 30, 2013). ConocoPhillips gas
learning algorithms. Computers & Geosciences 86: 120– hydrate production test final technical report. Technical
128. report, ConocoPhillips, URL http://www.netl.doe.gov.
Chauhan, S., Rühaak, W., Sass, I., Anbergen, H., Kabdenov, Song, Y. C., Yu, F., Li, Y. H., Liu, W. G., & Zhao, J. F. 2014.
A., Freise, M. & Wille, T. 2016. A new software collec- Mechanical property of artificial methane hydrate under
tion for 3D processing of X-ray CT images. Proceedings triaxial compression. Journal of Natural Gas Chemistry
1st International Conference on Energy Geotechnics and 19: 246–250.
Geotechnics, 29th–31st August 2016, Kiel Stern, L. A., Circone, S., Kirby, S. H., and Durham, W. B.,
Deusner, C., Bigalke, N., Kossel, E., & Haeckel, M. 2012. 2003. Temperature, pressure, and compositional effects
Methane Production from Gas Hydrate Deposits through on anomalous or “self ” preservation of gas hydrates.
Injection of Supercritical CO2 . Energies 5: 2112–2140. Canadian Journal of Physics 81: 271–283.
Ghiassian, H. & Grozic, J. L. H. 2013. Strength behavior of Uchida, S., Klar, A., & Yamamoto, K. 2015. Abstract Sand
methane hydrate bearing sand in undrained triaxial testing. migration modeling in hydrate-bearing sediments pre-
Marine and Petroleum Geology 43: 310–319. sented at 2015 Fall Meeting AGU, San Francisco, Calif.,
Gupta, S., Helmig, R. & Wohlmuth, B. 2015. Non- USA, 14–18 Dec
isothermal, multi-phase, multi-component flows through Yamamoto, K. 2013. Japan completes first offshore methane
deformable methane hydrate reservoirs. Computa- hydrate production test – methane successfully produced
tional Geosciences, 1–26doi:10.1007/ s10596-015-9520- from deepwater hydrate layers. Fire in the ice: Depart-
9, URL http://dx.doi.org/10.1007/ s10596-015-9520-9. ment of energy, office of fossil energy, national energy
Hyodo, M., Li, Y., Yoneda, J., Nakata, Y., Yoshimoto, N., & technology laboratory. Methane Hydrate News Letter 13,
Nishimura, A. 2014. Effects of dissociation on the shear No. 1–2.
strength and deformation behavior of methane hydrate- Yoneda, J., Masui, A., Konno,Y., Jin,Y., Egawa, K., Kida, M.,
bearing sediments. Marine and Petroleum Geology 51: Ito, T., Nagao, J., & Tenma, N. 2015. Mechanical proper-
52–62. ties of hydrate-bearing turbidite reservoir in the first gas
Hyodo, M., Yoneda, J., Yoshimoto, N., & Nakata, Y. (2013). production test site of the Eastern Nankai Trough. Marine
Mechanical and dissociation properties of methane and Petroleum Geology 66: 471–486.
443
Influence particle size on strength of gas hydrate cemented

granular materials
B.N. Madhusudhan & C.R.I. Clayton

Faculty of Engineering and the Environment, University of Southampton, UK
ABSTRACT: Quantifying the effect of gas hydrates on engineering properties of sediments is essential to
assess its role as triggering mechanism for submarine slope instabilities, potential energy resource or accelerating
climate change. Previous studies show cementation due to presence of gas hydrate in deep ocean sediments or
permafrost influences the seismic wave velocities of host sediment. This study examines the effect of particle
size on the strengths of gas hydrate cemented granular materials using our recently developed gas hydrate triaxial
apparatus. Cylindrical specimens of given porosity, methane hydrate content were prepared using ‘excess gas
method’ and sheared undrained at constant effective stress and rate of shearing for sands. These tests were then
compared with their corresponding host sediments with no hydrates. The stress strain behaviour indicates the
host soils exhibits stiffer behaviour due to presence of hydrates similar to structured soils of sensitivity greater
than 10. However, the change in strength behaviour of the disseminated gas hydrate sediments is significantly
influenced by the particle size, in terms of their specific surface and grading of the granular material regardless
of similarity in the hydrate concentration.
Keywords: laboratory tests, sands, methane hydrate, fabric/structure of soils
1 INTRODUCTION which is dependent on location of water within the

pore space of the media. Hydrate may form within the
The role of methane hydrate as one of mechanisms pore water, acting as additional mineral grains with the
sought to understand triggering of huge underwater voids, termed pore filling or it could occur as cement
landslides that could trigger tsunami in oceanic sedi- between grains, formed via the initial capillary water
ments (Urlaub et al., 2013) has driven researchers to bridge between two grains, termed frame supporting
understand their occurrence, morphology and stress- or cementing by Clayton et al. (2010). Previous stud-
strain behaviour in order to model submarine slopes ies by Clayton et al. (2010) and Priest et al. (2009) on
and assess the future risks (Talling et al., 2014). Quan- measuring wave velocities by forming hydrates using
tifying the engineering properties of gas hydrate is ‘excess water method’ did not have significant effect
also essential to accurately assess its role as poten- on measured wave velocities in sands, whereas using
tial energy source and climate change accelerator ‘excess gas method’ for same target hydrate content
(Mascarelli, 2009). led to rapid increase in the wave velocity. Clayton et al.
Methane hydrates are ice like structures that natu- (2010) also demonstrated that, for a given hydrate sat-
rally occur in deep ocean sediments and permafrost uration, relative density and effective stress, coarser
that forms within stability zone, governed by spe- grained quartz sand were twice stiffer than the finer
cific thermobaric range (Fig. 2). The major factors grained ones.This paper aims to demonstrate the above
influencing their occurrence, distribution, and mor- reported effect of grain size on the strength behaviour
phology are sediment size and geological structure of disseminated gas hydrate bearing sediments.
(Brewer et al., 1997; Clennell et al., 1999), poros-
ity and permeability of the host sediment (Collett,
1993) and source of methane gas. Depending on these
factors, the gas hydrate could occur in disseminated, 2 EXPERIMENT METHODOLOGY
nodular and hydrate sheets/veins (Malone, 1985). Dis-
seminated gas hydrates are known to occur in high Gas hydrate triaxial apparatus was specially built by
permeability soils (Collett, 1993; Booth & O’Leary, the authors at University of Southampton as part of
1991), wherein the above factors govern the hydrate research assessing impact of Arctic climate change
saturation. Disseminated hydrates formed within pore on increasing the risk of slope instability. Figure 1
space of a granular media could be classified into two shows the schematic layout of the apparatus and asso-
different micro-morphologies (Clayton et al., 2010), ciated accessories to apply appropriate thermobaric
445
Figure 1. Schematic of the gas hydrate triaxial apparatus.
conditions to create and test gas hydrates bearing sed-

iments. Disseminated methane hydrate was formed Figure 2. Thermobaric path used to form methane hydrates
within pore space of selected granular materials using by ‘excess gas method’.
the thermobaric controllers under conditions of con-
stant host sediment void ratio, effective stress and
stress history. pressure of 10 MPa (point A to B, Fig 2) with an effec-
tive stress of 250 kPa, before cooling to 2◦ C to form
hydrate (point B to C, Fig. 2). Temperature, back pres-
2.1 Host materials sure and effective stress conditions were maintained at
In order to study the effect of particle size on strength point C for at least 48 h, to allow complete hydrate for-
behaviour of gas hydrate bearing sediments, uniform mation. Small cycles of stain controlled loading tests
Leighton Buzzard sand representing coarse (Leighton upto 0.5% axial strain was conducted at 12h, 24h, 36
Buzzard B or LBB, Clayton et al., 2004) and fine and 48h to determine the comparative strength evo-
(Leighton Buzzard E or LBE, Clayton et al., 2005) lution during hydrate formation process. The rate of
grained uniform sediment particle sizes were chosen strain used was 0.2% per hour. During the process
for this study. Details regarding the index properties of of hydrate formation and thereafter, the loadram was
the materials used in testing can be found in Clayton always kept in contact with the sample at a stress of
et. al (2010). 15 kPa.
2.2 Hydrate formation

2.3 Hydrate content
Cylindrical triaxial specimens of 70 mm diameter and
140 mm height moist sands, mixed with known amount Calculating the percentage pore space occupied by
of water and cured overnight in sealed bag, were pre- methane hydrate formed by using excess gas method
pared using moist-tamping method (Ladd 1978). A has been described by Clayton et al. (2005, 2010). The
known amount of water, ∼2.5% to induce 10% hydrate degree of hydrate saturation at the end of gas hydrate
saturation in the granular pore space was used for this formation under ‘drained’ conditions could be calcu-
comparative study. Care was taken to see no spillage lated on the assumption that all the water in the sand
of the sand particles occurred and sample was encased pore space had been converted to hydrate under the
in Butyl membranes in order to avoid any leakage at thermobaric condition (Sultaniya et al., 2015). Con-
high pressures. The volumes, dry densities and void sidering the molar ratio of methane to water to be 1 :
ratio for the specimens for LBB sand was 551 cm3, 5.75 (Sloan and Koh, 2007), mass of the added water
1706 kg/m3 and 0.553 respectively, whilst for the spec- and porosity of the sand specimen, the hydrate satu-
imens for LBE sand was 520 cm3, 1616 kg/m3 and ration (Sh) can be calculated by using the expression
0.640 respectively. After forming the specimen by proposed by Sultaniya et al. (2015):
moist tamping, the specimen it was taken through the
methane hydrate/water phase to initiate hydrate for-
mation, by injecting methane at 20◦ C to attain a back
446
where Mw = molar mass of water (18.015 g/mol);
Mhy = molar mass of methane hydrate (119.63 g/mol),
mw = mass of pore water; ρhy = mass density of
methane hydrate (917 kg/m3), n = specimen porosity,
and VT = total specimen volume.
2.4 Strength testing

Strain controlled triaxial undrained shearing test was
performed on host sands without hydrates and with
10% hydrate (after hydrate formation) at constant
temperature of 2◦ C. The samples were sheared at con-
stant shearing rate of 0.2% per hour until constant
strength was reached for all the samples. The speci-
men temperature and cell volume was also measured
and monitored during the shearing process.
3.1 Effect of hydrate formation

As mentioned in earlier, gas hydrate was formed over
a period of 48 hours in the pore space by supply-
ing constant methane back pressure, whist measuring
the specimen temperature at bottom (Fig1). The time
calculated for hydrate formation is time since the spec- Figure 3. Strength evolution of a) LBB sand and b) LBE
imen temperature reaches 2◦ C, even though the cell sand during hydrate formation.
fluid temperature reached the designated temperature
(Sultaniya et al., 2015). Figure 3 shows the evolution
The significant mobilisation of the strength beyond
of the strength due to hydrate formation in both LBB
0.3% strain for the hydrate bearing LBB sand and
and LBE sands by performing small cycles of strain
beyond 0.17% for the hydrate bearing LBE sand indi-
on to the sample (0.5%), tested at 24, 36 and 48 hour.
cates either that the contribution of the cementation
The figure clearly indicates the evolution of hydrate
between the grains to the global strength is attained
cementation within the pore space with time. In case of
only after there is initial yielding of the specimen or
LBB sand (coarse grained) there is significant increase
that the sample could have misaligned (Gasparre et al.,
in the strength at strain below 0.3% regardless of the
2014) due to hydrate growth in its pore spaces, even
time after hydrate begins to form, however since the
though the sample was prepared carefully to form per-
cementation is not yet complete after 24 hr, the strength
fect cylinder and the loadcell was in contact throughout
is lesser in comparison to that after 48hrs beyond 0.3%
the process of hydrate formation.
strain. At 0.4% strain the strength attained by the LBB
Previous studies on the effect of given hydrate
sand were 180 kPa (no hydrate), 1175 kPa (24hr) and
cement (10%) on stiffness of LBB and LBE sand
1770 kPa (48 hr), indicating that the strength of sand
(Clayton et al., 2010) show that at effective stress of
increases greater than 6 times within 24hr of hydrate
250 kPa, LBB sand was twice stiffer than that of LBE
formation and after 48 hr there is 10-fold increase in
sand.
the coarse sand sediment strength hosting 10% hydrate
in their pore space. LBE sand on other hand also show
significant increase in strength for 48 hr and 36 hr test-
ing before 0.2% strain, but it is interesting to note that 3.2 Effect of particle size on strength behaviour
the specimen behaves significantly stiffer when tested
after 36 and 48h in comparison to the 24 hr below The effect of particle size on the stress strain behaviour
0.17% strain. At 0.4% strain the LBE sand attained of gas hydrate cemented sediments was examined by
107 kPa, 300 kPa, 950 kPa and 1050 kPa strength when triaxial undrained shearing tests on hydrate bearing
tested with no hydrate, after 24 hr, 36 hr and 48 hr Leighton Buzzard B and E sands and comparing it
respectively. As noted earlier the strength of LBE sand with uncemented their counterparts at similar porosity
due to hydrate cementation increases 3 times after and stress levels. The strength behaviour is expressed
24hr of hydrate formation and greater than 5 times as stress ratio, deviatoric stress (q) to mean effective
after 36hr. There is not significant gain in strength stress (p ), variation with deformation, axial strain on
when sample is left for 48hr to form hydrate indicat- the sample in figure 4 for all the specimens tested
ing that most of the hydrate cementation occurs before with and without cementation due to hydrate. Both
36 hours. the coarse (LBB) and fine (LBE) grained sands show
447
sharp increase in strength at very early stage of shear-
ing, below 1% strain due to hydrate cementing the
grains and thereafter the deviatoric stress begins to
drop as the sample deforms further until the sample
reaches constant deviatoric stress beyond 10% axial
strain. Hydrate cemented Leighton Buzzard B sand
exhibited peak dilative behaviour similar to a dense
sand or heavily overconsolidated soil, whereas its
uncemented counterpart exhibit loose sand behaviour
indicating the significant effect of hydrate cement on
the sediment behaviour. Leighton Buzzard E sand, on
the other hand exhibited slight peak dilative behaviour
similar to a medium sand or lightly overconsoli-
dated soil, whereas its uncemented counterpart exhibit
loose sand behaviour. Both the sands used for this
Figure 4. Stress-strain behaviour of LBB and LBE sands
study have been extensively characterised and tested with and without 10% hydrate.
at the University of Southampton (Abbireddy et al.,
2009; Clayton et al., 2010). From these studies, it is
known that Leighton Buzzard B had a higher unce-
mented stiffness than Leighton Buzzard E at the same
stress level and relative density and for 10% hydrate
cemented Leighton Buzzard B had twice the stiff-
ness to that of Leighton Buzzard E. The mechanism
explained by Clayton et al., (2010) was that since the
hydrate replaces the water, assuming 100% hydrate
replacement (Priest et al., 2005), the contact area
between two particles, previously bridged by capillary
water is replaced by hydrate cement (Fig. 5). Since the
cemented surface area is greater for Leighton Buzzard
B sand in comparison to Leighton buzzard E sand, it
explains why the higher strength for a coarser particle
Figure 5. Disseminated hydrate formation between two
when sheared at same hydrate content, effective stress sand grains using ‘excess gas method’ after Clayton et al.
and relative density. (2010).
3.3 Effect of shearing disseminated hydrates

The stress paths of all the specimens tested with or
without hydrate, sheared undrained, are shown in Fig-
ure 6. The stress paths indicate that there is minimal
pore pressure developed during the undrained shearing
of cemented hydrate and the effective stress is sim-
ilar to the total stress. This indicates that the hydrate
cement simply begins to rupture beyond the peak strain
and hence there is gradual reduction in the strength
with increasing deformation on the sample.
4 CONCLUDING REMARKS
This paper has presented the results of experiments Figure 6. Stress path of the specimens tested.
conducted to investigate the strength behaviour of
hydrate cemented sands, tested using new gas hydrate
triaxial apparatus built for this specific purpose.
Using the excess gas method to form disseminated the hydrate formation process took upto 36 hours and
hydrates, the formation of hydrates induced cementa- beyond that there the gain in strength is marginal.
tion between the particles and thus they exhibit stress The difference in the strength behaviour of simi-
strain behaviour similar to over consolidated soils. For lar hydrate bearing sediments due to particle size and
a given hydrate content, relative density and effective perhaps morphology (future study) reveals that the par-
stress coarse grained quartz sand show higher strength ticle surface area available for the cementation along
in comparison to finer grained sand. Small cycles of with hydrate content and porosity can be quantified
strength tests performed at regular intervals reveal that and incorporated in modelling.
448
REFERENCES Ladd, R. 1978. Preparing test specimens using undercom-
paction. Geotechnical Testing Journal, 1 (1): 16–23.
Abbireddy, C. O. R., Clayton, C. R. I. & Huvenne, V. A. I. Malone, R. D. 1985. Gas hydrates. DOE/METC/SP-218.
2009.A method of estimating the form of fine particulates. Washington, DC: US Department of Energy.
Géotechnique 59(6): 503–511. Mascarelli, A. 2009. A sleeping giant? Nature Reports
Booth, J. S. & O’Leary, D. W. 1991. A statistical overview Climate Change, 3: 46–49.
of mass movement characteristcs on the North Ameri- Priest, J. A., Best, A. I. & Clayton, C. R. I. 2005. A laboratory
can Atlantic outer continental margin. Marine Geores. investigation into the seismic velocities of methane gas
Geotechnol. 10(1): 1–18. hydrate bearing sand. J. Geophys. Res. Solid Earth 110
Clayton, C. R. I., Theron, M. & Vermeulen, N. J. 2004. The (B4): B04102.
effect of particle shape on the behaviour of gold tailings. Priest, J.A.,. Rees, E.V. L and Clayton. C. R. I. 2009. Influence
In Advances in geotechnical engineering: the Skempton of gas hydrate morphology on the seismic velocities of
conference. London: Thomas Telford. sands. J. Geophys. Res., 114: B11205.
Clayton, C. R. I., Priest, J. A. & Best, A. I. 2005. The effects Sloan, E. D., and Koh, C. A. 2007. Clathrate Hydrates of
of disseminated methane hydrate on the dynamic stiffness Natural Gases, 3rd ed., CRC Press, Taylor and Francis
and damping of a sand. Géotechnique 55 (6): 423–434. Group, New York.
Clayton, C. R. I., Priest, J. A. & Best, E. V. L. 2005. The Sultaniya, A. K., Priest, J. A. and Clayton, C. R. I. 2015. Mea-
effects of hydrate cement on the stiffness of some sands. surements of the changing wave velocities of sand during
Géotechnique 60(6): 435–445. the formation and dissociation of disseminated methane
Clennell, B. M., Hovland, M., Booth, J., Henry, P. & Win- hydrate. J. Geophys. Res. Solid Earth, 120: 778–789.
ters, W. J. 1999. Formation of natural hydrates in marine Talling, P, Clare, M, Urlaub, M, Pope, E, Hunt, J.E, Sebas-
sediments: 1. Conceptual model of gas hydrate growth tian, W. 2014. Large Submarine Landslides on Continen-
conditioned by host sediment properties. J. Geophys. Res. tal Slopes: Geohazards, Methane Release, and Climate
104, (B10): 22 985–23 003. Change. Oceanography, 27 (2): 32–45.
Collett, T. S. 1993. Natural gas hydrates of the Prudhoe Bay Urlaub, M., Talling, P. J., Masson, D. G. 2013. Timing and
and Kuparuk River area, North Slope, Alaska. AAPG Bull. frequency of large submarine landslides: implications for
77 (5): 793–812. understanding triggers and future geohazard. Quaternary
Gasparre A, Hight DW, Coop MR, Jardine RJ. 2014. The Science Reviews, 72: 63–82.
laboratory measurement and interpretation of the small-
strain stiffness of stiff clays. Géotechnique 64 (12): 942–
953.
449
Sand production modeling of the 2013 Nankai offshore gas production test
S. Uchida
Department of Civil and Environmental Engineering, Rensselaer Polytechnic Institute, Troy, USA
A. Klar
Faculty of Civil and Environmental Engineering, Technion – Israel Institute of Technology, Haifa, Israel
K. Yamamoto
Methane Hydrate Research & Development Group, Japan Oil, Gas and Metals National Corporation, Chiba, Japan
ABSTRACT: A better understanding of the behavior of gas hydrate-bearing sediments during gas extraction
is a vital step towards realization of long-term gas production for the future. In March 2013, the world first
trial of gas production from offshore hydrate-bearing sediments by depressurization method was conducted at
the Eastern Nankai Trough site, Japan. While the operation was successful in producing gas, after six days it
suddenly encountered a large amount of sand migration into the well, a phenomenon known as sand production,
leading to a premature termination of the operation. This incident has highlighted the importance of development
of sand migration model within hydrate-bearing sediments and understanding of the geomechanical behavior
of hydrate-bearing sediments with the effect of sand migration during gas extraction. This paper presents the
overview of the recently developed thermo-hydro-mechanically coupled formulation that entails sand migration
in gas hydrate-bearing sediments. The formulation is then applied to simulate the 2013 Nankai production test
in wellbore scale, including history matching of produced water and gas. The amount of produced sand at the
end of the test is also matched and the effects of sand migration on geomechanical behavior are investigated.
1 INTRODUCTION sediments by 4) together with sand migration mod-

els (7). The highlight of the developed formulation is
Methane hydrate has been attracting international its coupled manner of incorporation of sand migra-
interest because of its potential as an abundant and tion features such that grain detachment causes stress
widespread source of natural gas that could meet many reduction, sediment shear deformation induces grain
decades worth of global energy demands. As of today, detachment and grain migration alters multiphase fluid
only a few short-term field trials have been success- pressure and temperature profiles.
ful in producing gas, namely at the Mallik site, Canada, This paper presents a brief summary of the devel-
in 2007 and 2008 (1), at the Mount Elbert site, USA, oped thermo-hydro-mechanical sand production for-
in 2007 (2), at the Ignik Sikumi site, USA, in 2012 mulation in gas hydrate-bearing sediments and its
(8) and at the Nankai Trough site, Japan, in 2013 (12). application to the field production test conducted at
Having learned lessons from the past field trials, it the offshore Eastern Nankai Trough in 2013. This work
was hoped that the 2013 Nankai gas production test is a part of Research Consortium for Methane Hydrate
operated by Japan Oil, Gas and Metals Corporation Resources in Japan (MH21), supported by JOGMEC.
(JOGMEC) would be a long-term success but it was on
the sixth day that sand production (excessive migration
of sand into the well) occurred despite the presence of 2 COUPLED THERMO HYDRO MECHANICAL
sand screen, leading to an untimely abandonment of FORMULATION FOR SAND MIGRATION IN
the well. Therefore, establishing a commercially viable GAS HYDRATE-BEARING SEDIMENTS
and sustainable method of methane gas production
from gas hydrate-bearing sediments still remains as This section briefly describes the derivation of the
a major challenge. developed coupled thermo-hydro-mechanical formu-
Motivated by the premature termination of the lation for sand production. The detailed explanations
test, the authors have recently developed a cou- can be found in 9).
pled thermo-hydro-mechanical sand production model
in gas hydrate-bearing sediments (9). The devel- 2.1 Solid mass balance
oped model is the extension of existing thermo- Since the solid volume is no longer constant when
hydro-mechanical simulator for gas hydrate-bearing considering sand migration, three states for solids are
451
where V is the control volume, v is the volumet-
ric strain (compression negative), n is the porosity
(=Vv /V ), βs is the thermal expansion coefficient of
soil grains and T is the temperature. This leads to a
change in the storage term equations of water, gas and
hydrate. Together with: [1] the capillary pressure equa-
tion Pc (Swe ) derived from the model by 11), where Swe
is the effective water saturation; and [2] the pore space
Figure 1. An elementary volumetric cube representing solid condition (i.e. dSw + dSg + dSh = 0 where S is the
states, mixture and concentrations. saturation), the incremental form of five unknowns,
introduced: [1] flowing solids (fs) that are currently Pw , Pg , Sw , Sg and Sh can be solved as simultaneous
flowing with fluid; [2] stable solids that are still part equations. For example, dPw is given by:
of the original soil skeleton and thus intact (ssi); and
[3] solids that are settled after flowing (sst). A simple
volumetric cube incorporating these states is shown in
Fig. 1 where the subscripts w, g and h are for water,
gas and hydrate, respectively. A grain is assumed to
flow with water or gas. Thus, the concept of mix-
ture is introduced as water-flowing solid mixture (wm)
and gas-flowing solid mixture (gm). The volumetric
concentrations of the flowing solids in the water-
mixture (fs|wm) and gas-mixture (fs|gm) are defined
V V
s
as cwm = Vw +fs|wm
Vfs|wm
s
and cgm = Vg +fs|gm
Vfs|gm
, which are also
illustrated in Fig. 1.
The mixtures are assumed to hold the same super-
ficial velocity as their corresponding fluids, that is,
qwm /Vwm = qw /Vw and qgm /Vgm = qg /Vg where q is
the discharge vector. Assuming the Darcy’s law is valid
for the sole fluid, the discharge of flowing solids can
be given by:
S S S P
where D = KSww + Kgg − Kw Kg (Sw +wSgg)(1−S
c
wr −Sgr )
, =1−
Sg Pc
where µ is the viscosity, Kh is the intrinsic permeabil- Kg (Sw + Sg )(1−Swr −Sgr )
,
Swr and Sgr is the degrees of
ity tensor with the effect of hydrate, k r is the relative residual water and gas saturation (discussed later),
permeability factor, P is the pressure, ρ is the density K is the bulk modulus, M is the molecular mass,
and g is the gravity vector. Nh is the number of water molecules to consti-
The mass of the solids in a control volume is not tute one hydrate molecule with one gas molecule
affected by the change in the solid state by itself. (i.e. Mh = Nh Mw + Mg ) and Rh is the rate of hydrate
Rather, it is altered only by mass divergence of the formation/dissociation (dissociation positive) deter-
flowing solids: mined by the current Pg and T such as the first-
order kinetic model of 3). Equation (4) clearly shows
how the variable is affected by five components:
where m is the mass, t is the time and the subscript s fluid flow, mechanical deformation, hydrate forma-
is for solids. tion/dissociation, temperature change and sand migra-
tion, which are represented by a large square bracket
in the right-hand side of the equation.
2.2 Hydro-sand migration coupling Considering hysteresis nature of the multiphase
The solid mass change results in the change in the void media and also the presence of hydrate, the effective
volume: water saturation is defined as:
452
It should be noted that, at in-situ condition, there
is no gas present as any residual gas should form
hydrate under the pressure-temperature condition. The
residual gas saturation shall then increase accord-
where is the strain vector, Dehs is the elastic stiff-
ing to developing gas in pores (without exceeding
ness matrix of hydrate-bearing soil continuum (i.e.
the gas saturation itself) and shall never decrease.
Dehs = Deh0 Sh + Des ), g is the plastic potential, f is the
This development of residual gas saturation can be
yield function, ξ is the hardening parameters in a vec-
modeled by:
tor form, δ = (1, 1, 1, 0, 0, 0)T and β∗ is the volumetric
mean of grain and hydrate thermal expansion coef-
ficient (i.e. β∗ = (1 − n)βs + nSh βh ). This expression
states that the effective stress change can be caused
by soil straining, hydrate dissociation, temperature
where Sgr0 is the maximum residual gas saturation and change and grain detachment, which are represented
Sg,max is the historical maximum of the varying gas by each large square bracket on the right hand side of
saturation during gas production operation. the equation.
2.3 Mechanical-grain detachment coupling 2.4 Thermo-sand migration coupling
The effective stresses are defined as the stresses that As flowing solids carry heat, the sand migration
are carried by the intact solids. When a grain is contributes to the convection term:
detached from the skeletal intact solids, the grain is
no longer part of the intact solid continuum. There-
fore, the part of the effective stresses can be released
upon grain detachment, leading to stress relaxation.
where cT ∗ is the volumetric mean of specific heats, that
The magnitude of the released stresses should depend
is, cT ∗ = (1 − n)ρs csT + n(Sw ρw cwT + Sg ρg cgT + Sh ρh chT ).
on both the current effective stresses and the fraction
of the detached grain to the intact solids. Thus, intro- Introducing the above term into the energy-balance
ducing a proportionality factor ω1 , the effective stress equation within gas hydrate-bearing sediments results
change due to the grain detachment is: in:
where σ is the effective stress vector and dmssi is

the change in the mass of intact solids (negative
denotes grain detachment) per control volume. Intro-
ducing the above expression into a hydrate-dependent
elastoplastic constitutive model results in:
where K T ∗ is the bulk thermal conductivity tensor

which can be obtained by a similar manner to cT ∗ ,
H is the enthalpy of methane hydrate (H > 0
and Rh is positive/negative depending on hydrate
dissociation/formation, respectively) and σ dp is the
mechanical energy dissipation because of plastic
deformation. The expression implies that the temper-
ature change is caused by conduction, convection of
water and gas, hydrate formation/dissociation, plas-
tic mechanical deformation and convection of sand
migration, each of which is represented by the large
square bracket on the right-hand side of the equation.
453
3 SOLID STATE CHANGE: DETACHMENT,
SETTLING AND LIFTING
This section summarizes the suggested mechanisms

and corresponding mathematical representations for
the state change of solids by detachment (from ssi to
fs); settling (from fs to sst); and lifting (from sst to fs)
by 9).
3.1 Grain detachment

For grain detachment, the following five mechanisms
are assumed. First, grains are detached from the intact
soil skeleton when hydraulic gradient exceeds its crit-
ical value. Second, when hydrate is present in pores,
grain detachment should be harder to occur. Third, the
flowing grain does not switch the media while flow-
ing. Forth, the active fluid (i.e. fluid that has greater
hydraulic gradient than its critical value) realizes the
detachability potential into an actual detachment in
proportion to relative pore occupancy to multiphase
fluids. Fifth, the detachability potential increases as a
result of shearing deformation. These assumptions can
be expressed together as:
Figure 2. Model geometry and boundary conditions.
3.2 Grain settling and lifting

where ω2 is the controlling parameter for the rate of
detachment (dimension of time−1 ), M dtc is the detach- Considering that the settled solids are no longer part of
ability potential, H (x) is the heaviside step function, i soil skeleton, they should be easier to initiate flowing
is the hydraulic gradient of a fluid (water or gas) and icri compared to the solids of intact skeleton. Thus, intro-
is the critical hydraulic gradient, which is influenced ducing another proportionality factor ω5 (0 ≤ ω5 ≤ 1),
by the hydrate saturation through: the mass of settled solids becomes:
where icri0 is the critical hydraulic gradient for grain

detachment without hydrate and ω3 is the increasing
factor of the critical gradient with hydrate. The first
term of the right-hand side of Eq. (12) is for water- i
where Hw = H ( ω5iwicri − 1) and Hg = H ( ω5gicri − 1). The
induced flowing solids mfs|wm and the second term is w g
for gas-induced flowing solids mfs|gm . The detacha- first two terms of Eq. (15) represent settling and the
bility potential increases with shearing deformation last term is for lifting. Unlike grain detachment, grain
but actual detachment itself reduces the potential such lifting does not directly affect soil effective stresses.
that:
4 NUMERICAL SIMULATIONS OF THE 2013
NANKAI GAS PRODUCTION TEST
where ω4 is the model parameter that gives the increase 4.1 Model geometry and in-situ conditions
in the potential due to shearing deformation and ms0
The model geometry and the boundary conditions
is the initial intact solid mass equal to the initial
used for the wellbore-scale simulations of the 2013
solid mass (i.e. mssi0 = ms0 ).It can be seen that grain Nankai offshore test are shown in Fig. 2. The wellbore
mssi
detachment ceases when ln ms0 becomes equal to is assumed to be constructed instantaneously in time
−ω4 d . without disturbing the in-situ stresses. The wellbore
454
boundary for displacement is fixed in the radial direc-
tion but smooth in the vertical direction, that for fluid
flow is only permeable inside the production zone and
that for heat is insulated. The seabed is assumed to
be at 1000 m from the mean sea level. The production
zone is between −278 and −308 m from the seabed.
Figure 3 provides (a) hydrate saturation profile
and (b) geometric mean permeability profiles with
(red line) and without (black line) hydrate. Between
Figs. 3a and b, the soil classification is shown. From
the in-situ measurement, it is found that the sedi-
ments are highly stratified with alternating sand and
clay layers. The measurement was conducted approxi-
mately every 50 cm deep interval but the model mesh is
slightly coarser (e.g. 90 cm around the production zone Figure 3. Hydrate saturation and permeability profiles used
and on average 400 cm above and below the produc- in the simulations.
tion zone). In order to incorporate the strata in terms of
fluid flow, the horizontal permeability without hydrate
for a given element is averaged by:
where α is the scaling factor for the overall permeabil-

ity to be used later for history matching, khr,j is the mea-
sured horizontal permeability without
hydrate, zj is
the measurement interval and j zj is the element
vertical size. In contrast, the vertical permeability
without hydrate is:
For an element level, therefore, the permeability tensor

is defined as:
Figure 4. Recorded depressurization conducted during the
2013 Nankai test.
This facilitates the presence of a thin clay layer. In

the analysis, the horizontal permeability is on average
found to be approximately 200 times greater than the
vertical permeability.
As a boundary condition, the well pressure
decreases according to the recorded value as shown
in Fig. 4. The starting time of the simulation is set at
4 am on 12 March, 2013. The thermocline is set to be
0.272 K/m and 278.8 K at the seabed based on the in-
situ measurement. The salinity is assumed to be 3%. Figure 5. Hydrate phase boundary at the Nankai site.
Accordingly, the profiles of the water pressure and the
phase equilibrium pressure at in-situ are obtained and
shown in Fig. 5. The lower boundary of the hydrate sta- the soil responses, both shear and volumetric, of the
bility zone coincides with the location where hydrate Nankai sand and clay, making the use of the model’s
is inferred non-existent albeit sand layers. versatility. For the Nankai sand, drained triaxial com-
pression test data by 5) while, for the Nankai clay,
undrained triaxial compression test (preceded by K0
4.2 Soil properties and other physical properties
consolidation) data by 6) are calibrated. The results
To capture the geomechanical behavior of the Nankai (normal lines) including the experimental data (dot-
sediments, the soil model by 10) is utilized to match ted lines) are shown in Fig. 6. The calibrated model
455
parameters are summarized in Table 1. Other phys- water production history is less affected by the relative
ical parameters used for the simulation such as the permeability. Subsequently, the gas production history
thermal conductivity of water, gas and hydrate are is matched by adjusting the relative permeability fac-
presented in 9). tor. Finally the sand production is matched by varying
sand migration related parameters.
The overall permeability is adjusted to match the
4.3 History matching of the 2013 Nankai water production history by changing the value of α
production test described in Eqs. (16) & (17). Thus, the anisotropic
nature in the permeability is unaffected. In other
Figure 7 illustrates the histories of (a) gas production, words, the value α swifts the permeability profiles
(b) water production and (c) sand production. The dot- illustrated in Fig. 3b leftwards when α < 1 and right-
ted lines represent the recorded values during the wards when α > 1. For the best matching, α is deter-
Nankai 2013 operation. The normal lines represent the mined to be 1/8.
best-fitted results, obtained from a series of recursive Based on α = 1/8, the gas production history is
analyses. Since production history is predominantly matched by altering the relative permeability curves,
influenced by the in-situ permeability, the water proof which the best matched case is shown in Fig. 8. The
duction is first equated. This is because a large fraction relations are derived based on vanGenuchten (1980)
of pore space is occupied by water. In other words, the formula and the parameters are found to be a = 0.55,
b = 3.5 and c = −0.05. In addition, the residual satura-
tions defined in Eq. (5) are determined to be Swr = 0.2
and Sgr0 = 0.012.
The recorded history of sand production is not avail-
able and thus this study seeks to match the overall
production at the end of six day operation, known to be
approximately 27 m3 . Grain detachment occurs when
the hydraulic gradient exceeds the critical gradient and
its amount depends on the potential which is cumula-
tive deviatoric strain as described in Eqs. (12) & (14),
respectively. Considering there are currently flowing
and settled solids, the detached solid volume is always
Table 1. Soil parameters used for the Nankai soils.
Parameter Sand Clay
Void ratio e0 = n/(1 − n) 0.82 0.72

Slope of critical state M 1.40 1.32
Compression line λ 0.16 0.20
Swelling line κ 0.025 0.05
Poisson’s ratio ν 0.28 0.23
Preconsolidation stress pcs 5.0 MPa −0.012z(m) +
0.29 MPa
Subloading factor u 10 960
Yield surface expansion pcd 45 Sh 0.8 MPa –
Sh dependent modulus Eh0 320 MPa –
Figure 6. Deviatoric and volumetric behaviors of (a) Nankai Degradation factor m 15 –
sand and (b) Nankai clay.
Figure 7. Production histories and simulation results.
456
Figure 8. Relative permeability and capillary pressure rela-
tions used in the simulation.
Table 2. Sand migration related parameters used for the

analyses. Figure 9. Thermo-hydro behavior of the sediments.
Parameter Value
Stress reduction factor ω1 1.0

Rate of detachment ω2 0.1 hour−1
Critical gradient factor ω3 1.0
Shear to detachability ω4 1.0
Grain settling factor ω5 0.5
Critical gradient icri0 5.0
greater than the produced solid volume. When d is

known, therefore, a rough estimation of produced vol-
ume is feasible. Table 2 lists the sand migration related
parameters employed for the best matching analysis.
4.4 Thermo-hydro-mechanical studies during

the 2013 Nankai production test
Figure 9 presents the thermo-hydro behavior of
the Nankai hydrate-bearing sediments during the
production test: (a) hydrate saturation profile; and (b)
temperature profile at the elapsed times of 3 and 6 Figure 10. Mechanical behavior of the sediments.
days. As can be seen, hydrate dissociation does not uni-
formly occur due to the differences in the permeability
and temperature. The area with lower temperature cor- ratio implies whether the sediment deformation is
responds to the area with greater hydrate dissociation. shearing-oriented (q) or volumetric-oriented (p ) (or
It is evident that a greater hydrate dissociation occurs consolidation-oriented). The sediments around the
in the layers adjacent to the low initial Sh (cf. Fig. 3), well inside the production zone appears to deform in
where adequate heat convection is available. volumetric manner, whereas the layers with relatively
Figure 10 presents the mechanical behavior of the high Sh tend to deform in shearing manner. This is
hydrate-bearing sediments in terms of (a) stress ratio because the hydrate dissociation occurred in the layers
q/p where q is the deviator stress and p is the with low Sh induces stress relaxation and subsequently
mean effective stress;and (b) displacement magni- the stiffer sediments (i.e. layers with high Sh ) carry
tudes, that is, |u| = ur 2 + uz 2 , with the arrows the redistributed stresses. From the arrows in Fig. 10b,
indicating its magnitudes and directions, at the elapsed large vertical deformation is evident just above the pro-
times of t = 3 and 6 days. The change in the stress duction zone and large radial deformation is noticeable
457
the stress redistribution from the low Sh layers. It is
also shown that a significant shearing deformation as
well as sand migration appeared to occur at the depth
of approximately −300 m.
REFERENCES
Dallimore, S. R., J. F. Wright, K. Yamamoto, & G. Belle-
fleur (2012). Proof of concept for gas hydrate production
using the depressurization technique, as established by
the jogmec/nrcan/aurora mallik 2007-2008 gas hydrate
production research well program, mackenzie delta, north-
west territories, canada. Bulletin of the Geological Survey
of Canada 601, 1–15.
Hunter, R. B., T. S. Collett, R. Boswell, B. J. Anderson,
S. A. Digert, G. Pospisil, R. Baker, & M. Weeks (2011).
Mount Elbert Gas Hydrate Stratigraphic Test Well, Alaska
North Slope: Coring operations, core sedimentology, and
lithostratigraphy. Marine and Petroleum Geology 28(2),
311–331.
Kim, H., P. Bishnoi, R. Heidemannn, & S. Rizvi (1987).
Kinetics of methane hydrate decomposition. Chemical
Engineering Science 42(7), 1645–1653.
Figure 11. Sand migration of the sediments. Klar, A., S. Uchida, K. Soga, & K. Yamamoto (2013).
Explicitly coupled thermal-flow-mechanical formulation
in the sediments away from the wellbore, particularly for gas hydrate sediments. Society of Petroleum Engineers
at r ≈ 10 m. Journal 18(2), 196–206.
Figure 11 presents: (a) the snapshot of the normal- Masui, A., H. Haneda, Y. Ogata, & K. Aoki (2007). Mechani-
ized flowing solid volume; and (b) the normalized cal Properties of Sandy Sediment Containing Marine Gas
detached solid volume (i.e. change in the intact) at the Hydrates in Deep Sea Offshore Japan Survey drilling in
elapsed times of 3 and 6 days. The values of the set- Nankai Trough. In Seventh ISOPE Ocean Mining Sympo-
sium, Lisbon, Portugal, pp. 53–56. International Society
tled solids Vsst /V are negligibly small and most of Vfs of Offshore and Polar Engineers.
are flowing with water. Comparing Figs. 11a & b, the Nishio, S., E. Ogisako, & A. Denda (2009). Geotechnical
values of Vfs /V away from the well are almost iden- properties of core samples recovered from seabed ground
tical to those of −Vssi /V . This implies that, even in East Nankai Trough. Journal of Geography 118(5),
though the solids are detached in the region, they 955–968.
have hardly moved towards the well due to the slow Papamichos, E., I. Vardoulakis, J. Tronvoll, & A. Skjaerstein
fluid velocity. The lowest possible value for Vssi /V (2001). Volumetric sand production model and experi-
is n − 1 = −0.65 when all the solids are detached. This ment. International Journal for Numerical and Analytical
implies that the amount of detached solids are around Methods in Geomechanics 25, 789–808.
Schoderbek, D., H. Farrell, K. Hester, J. Howard,
5% near the well. At the depth of −300 m, detach- K. Raterman, S. Silpngarmlert, K. L. Martin, B. Smith, &
ment occurs as far as 20 meters from the well and this P. Klein (2013). ConocoPhillips Gas Hydrate Production
location corresponds with the layers with high q/p , Test. Technical report, US Department of Energy.
implying that shearing deformation contributes to the Uchida, S., A. Klar, & K.Yamamoto (2016). Sand production
sand migration and eventually sand production. model in gas hydrate-bearing sediments. International
Journal of Rock Mechanics and Mining Sciences 86,
303–316.
5 SUMMARY Uchida, S., K. Soga, & K. Yamamoto (2012). Critical state
soil constitutive model for methane hydrate soil. Journal
of Geophysical Research 117, B03209.
This paper presented the overview of the cou- van Genuchten, M. (1980). A closed form equation for pre-
pled thermo-hydro-mechanical formulation for sand dicting the hydraulic conductivity of unsaturated soils.
migration in gas hydrate-bearing sediments and its Soil and Science Society of America Journal 44, 892–898.
application to the 2013 Nankai offshore gas produc- Yamamoto, K., Y. Terao, T. Fujii, I. Terumichi, M. Seki,
tion test. The formulation was able to achieve good M. Matsuzawa, & T. Kanno (2014, may). Operational
history matching of the 2013 Nankai test for gas, water overview of the first offshore production test of methane
and sand production. Through wellbore-scale thermo- hydrates in the Eastern Nankai Trough. In Offshore
hydro-mechanical studies, it is found that the layers Technology Conference, Huston, USA.
with high Sh deformed in shearing manner caused by
458
A small pore size effect on dissociation behavior of gas hydrates

in fine-grained sediments
Taehyung Park, Hak-Sung Kim & Tae-Hyuk Kwon

Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology (KAIST),
South Korea
ABSTRACT: As capillarity in small pores is known to affect dissociation of gas hydrates, the dissociation
behavior of gas hydrates in fine-grained sediments need to be clearly understood for successful practices of
resource recovery. Therefore, this study investigated the dissociation behavior of gas hydrates in different sized
pores of host sediments. Gas hydrates were synthesized from a partially water-saturated conditions in a sand
sample and a natural fine-grained sediment sample cored from Ulleung Basin (UB), offshore Korea, and these
hydrates were thermally dissociated under a constant volume condition while monitoring the pressure and
temperature. The dissociation of gas hydrate in the fine-grained sediments occurred at the lower temperature
than the bulk equilibrium by ∼1.5◦ C. When compared with the pore size distribution obtained by the mercury
intrusion porosimetry, gas hydrate was preferentially formed in small pores for a given range due to the initial
water locations in partially water-saturated conditions.
1 INTRODUCTION effect on hydrate equilibrium conditions (Handa and

Stupin, 1992; Uchida et al., 1993; Tohidi et al., 2001;
Gas hydrates are ice-like crystalline structure com- Anderson et al., 2003; Uchida et al., 2002).
prised of hydrogen bonded water cages with gas Despite of numerous gas hydrate studies in porous
molecules inside (Sloan, 1998). These are naturally media, there have been lack of experimental data on
formed under high pressure and low temperature con- the dissociation behavior of gas hydrates in natural
ditions (Sloan, 1998). In deep ocean sediments where fine-grained sediments as most of the previous studies
those conditions are satisfied, abundant amount of were conducted using porous silica gels. Therefore, in
natural gas hydrates exist in continental margins and this study, we investigated the dissociation behavior
permafrost regions (Sloan and Koh, 2008). As energy and the equilibrium conditions of CO2 hydrate in the
use and fossil fuel consumption increase from decade sediments with different pore sizes, including a fine
to decade, gas hydrate deposits in oceanic sediments sand and a natural clayey silt sample cored from the
have been spotlighted as prospective energy resource Ulleung Basin (UB), offshore South Korea.
for the large amount of methane (Lee et al., 2003).
Potential reserves of natural gas hydrates are over
1.5 × 1016 m3 all over the offshore earth areas (Mak- 2 EXPERIMENTAL PROGRAM
ago et al., 2007). Additionally, gas hydrates have been
considered as one of the possible solutions for car- 2.1 Materials
bon dioxide (CO2 ) mitigation. By the formation of CO2 gas (commercial-grade with the purity of 99.9%,
a CO2 hydrate layer in subsurface, CO2 hydrate can Sam-Oh Gas Co., South Korea) was used for CO2
also function as a cap-rock structure which provides hydrate formation in this study. Fine sand (Ottawa
the secondary shield to prevent the leakage of CO2 F110) was selected as representative coarse-grained
(Tohidi et al., 2010). sediments, as shown in Figure 1a. A natural fine-
For successful practices of methane recovery from grained sediment sample retrieved from Ulleung
hydrate deposits or CO2 storage using hydrates, bet- Basin, located in the southwestern part of the East Sea
ter understanding of the dissociation behavior and the of South Korea was used for the experiment, as shown
phase equilibrium of gas hydrates in natural sediments in Figure 1b. Detailed descriptions of the sediment
is required (Lee et al., 2007, Uchida et al., 2004). For samples are available in previous references (Kwon
these reasons, there have been numerous studies to et al., 2011; Kim et al., 2013). The basic properties of
understand the phase equilibrium of gas hydrates in the tested samples, including mean particle diameter
porous media; previous studies found that dissociation (D50 ), specific gravity (Gs ), specific surface area (Sa ),
pressure of gas hydrates were higher in small pores porosity, and water saturation (Sw ), are described in
than that in a bulk phase, which means the inhibition Table 1.
459
Table 1. Properties of the sediments
Fine sand (Ottawa F110) Natural UB sediment

(Kwon et al., 2008) (Kim et al., 2013)
Mean diameter, Mean diameter,

D50 = 120 µm D50 = 4.83 µm
Specific gravity, Gs = 2.65 Specific gravity, Gs = 2.52
Specific surface area, Specific surface area,
Sa = 0.019 m2 /g Sa = 32.09 m2 /g (dry method)
Porosity = 0.41 Porosity = 0.61
Water saturation, Sw 0.46 Water saturation, Sw = 0.52
Figure 3. Temperature and pressure conditions of CO2

hydrate in sand sample during 1st heating, 2nd cooling and
3rd heating procedures.
Figure 1. (a) Microscopic image of the fine sand (Ottawa 2.3 Experimental Procedures
F110) sample and (b) scanning electron microscope (SEM)
image of the natural UB sediment. Each sample was mixed with DIW and the mixture was
hand-tamped in the cell to achieve a partially saturated
condition. The porosity of 0.41 and 0.61 was achieved
for the fine sand and natural UB sediment, respec-
tively (Table 1).The cell was flushed with pure CO2 gas
several times to remove the residual air inside the reac-
tion cell. CO2 gas was then injected and pressurized
to 3 MPa. The temperature of the cell was lowered to
274.15 K. CO2 hydrate nucleation took place, and the
pressure and temperature was maintained for 24 h until
forming a sufficient amount of CO2 hydrate. Hydrate
nucleation was confirmed by appearance of a temper-
ature peak where an exothermal reaction takes place.
After 24 h, the temperature of the reaction cell was
increased in steps of 0.5◦ C every 6 h under a con-
Figure 2. Schematic diagram of experimental setup.
stant volume condition (i.e., isochoric heating). When
the temperature exceeded the equilibrium temperature,
A proper amount of deionized water (DIW, 18 M- hydrates started to melt due to the increased temper-
cm) was prepared to partially saturate the sediments ature. The pressure also increased as the dissociation
and form CO2 hydrates. of CO2 hydrate released free CO2 gas. Thereby, the
pressure and temperature were tracked to obtain the
2.2 Experimental Apparatus hydrate phase equilibrium. Such step-wise heating was
The experimental setup was designed to measure the continued until complete dissociation of CO2 hydrate
dissociation temperature and pressure of CO2 hydrate (i.e., 1st heating cycle). Then, the reaction cell was
in the sediments. These experimental setup is graphi- cooled again to 0◦ C for CO2 hydrate formation (i.e.,
cally described in Figure 2.A cylindrical, transparent 2nd cooling cycle). The aforementioned dissociation
high pressure cell made of polycarbonate with a procedure was repeated to make sure the experiment
volume of 3.18 cm3 was used for this study. results shows the repeatability (i.e., 3rd heating cycle).
During whole experimental procedures, the cell
was submerged in the temperature-controlled bath
(RW-2025G; Lab Companion, South Korea) to con-
trol the temperature of the cell. A platinum resistance
thermometer (Pt100; Hankook Electric Heater, South
3.1 Dissociation behaviors of CO2 hydrate in fine
Korea) was placed at the center of the sediment inside
sand and fine grained sediment
the cell and two pressure transducers (PX302; Omega,
United States) were connected to the cell. For real Figure 3 and Figure shows the pressure-temperature
time monitoring of the temperature and pressure con- trances during the experiments with the fine sand and
ditions, the data acquisition unit (34970A; Agilent, UB sediment samples, respectively. CO2 hydrate in
United States) was used. the fine sand sample started to dissociate at 5◦ C, 2.25
460
Figure 4. Temperature and pressure conditions of CO2 Figure 5. Shift of phase boundaries of CO2 hydrate in small
hydrate in UB natural sediment sample during 1st heating, pore sizes (calculated from Gibbs-Thomson equation) and
2nd cooling and 3rd heating procedures. dissociation behavior of CO2 hydrate in natural UB sediment.
MPa during 1st heating process, in which the pressure-

temperature trance (or PT trace) followed the phase
boundary of pure CO2 hydrate in a bulk condition
(or bulk phase boundary). As expected, CO2 hydrates
formed in find sand sample did not experience any
capillarity effect due to the large enough pore sizes
larger than 1 µm. This is consistent with with an ear-
lier work (Kwon et al., 2008), confirming the validity
of our experimental procedures and measurements.
While the dissociation conditions of CO2 hydrate
in fine sand sample was indistinguishable to those
of pure CO2 hydrate, the dissociation of CO2 hydrate
in the natural UB sediment started at 5.6◦ C, 2.8 MPa
which was at a lower temperature than the bulk phase
boundary. Natural UB sediment showed a pronounced
capillary effect depressing the melting point of CO2
hydrate in fine-grained sediments. Figure 6. Pore size distribution of natural UB sediment
analyzed by mercury intrusion porosimetry.
3.2 Capillary effect of fine-grained sediments
on dissociation behavior of CO2 hydrate ρh0 is the density of CO2 hydrate, and Lf is the latent
Association of water molecules with mineral surfaces heat of dissociation of CO2 hydrate. In this study,
in the fine-grained sediments becomes dominant in Tdep were calculated using previously reported data:
small pore sizes, which lowers the water activity and γhw = 0.030 N/m, mh = 174 g/mol, ρh0 = 1065 kg/m3 ,
leads to the depression of melting temperature of and Lf = 65.2 kJ/mol (Anderson et al., 2003; Kwon
ice crystals. Such melting temperature depression has et al., 2008).
been widely observed in ice melting or water freez- The phase boundaries considering the melting point
ing in fine-grained sediments. Likewise, due to the depression (Tdep ) by different pore sizes were super-
decreased water activity by capillarity in small pores, imposed with the bulk phase boundary in Figure 5. It
CO2 hydrate equilibrium condition was inhibited in appeared that the hydrate-containing pore size of the
fine-grained sediments with the melting temperature natural UB sediment was in the range of ∼50–70 nm.
depression, as shown in Figure 4. The depression of This implies that CO2 hydrate formation and dissocia-
melting temperature (Tdep ) in porous media from tion in fine-grained sediments is affected by capillarity
bulk conditions (Tbulk ) can be estimated by Gibbs- while those in coarse-grained sediments may not be
Thomson equation (e.g., Anderson et al., 2003; Kwon affected.
et al., 2008): Figure 6 shows the pore size distribution of the
natural UB sediment measured by the mercury intru-
sion porosimetry (MIP). The MIP results showed that
the pore size of the natural UB sediment ranged from
10 nm to 10 µm. Whereas, our experimental data (Fig-
where γhw is the surface tension between CO2 hydrate ure 5) indicated that the hydrate-containing pores sized
and water, mh is the molar mass of the CO2 hydrate, ∼50–70 nm. This implies the preferential formation of
461
CO2 hydrate in small pores of fine-grained sediments. Lee, J. H., Kim, H. & Choi, S. C. Effects of phase-equilibrium
This was possibly attributed to the initial location of temperature and pressure on the thickness decision of a
water when hydrates were nucleated, as water favors methane hydrate container. Key Engineering Materials,
small pores in partially water-saturated conditions due 2007. Trans Tech Publ, 2782–2785.
Lee, S., Liang, L., Riestenberg, D., West, O. R., Tsouris, C. &
to the capillarity. Adams, E. 2003. CO2 hydrate composite for ocean carbon
sequestration. Environmental Science & Technology, 37,
3701–3708.
Makogon, Y., Holditch, S. & Makogon, T. 2007. Natural gas-
ACKNOWLEDGMENTS hydrates—A potential energy source for the 21st Century.
Journal of Petroleum Science and Engineering, 56, 14–31.
This work was partially supported by This work was Shahnazar, S. & Hasan, N. 2014. Gas hydrate formation con-
supported by the Korea Institute of Energy Technology dition: Review on experimental and modeling approaches.
Evaluation and Planning (KETEP) and the Ministry of Fluid Phase Equilibria, 379, 72–85.
Trade, Industry & Energy (MOTIE) of the Republic of Sloan, E. D. 1998. Gas hydrates: review of physical/chemical
Korea (No.20152520100760). properties. Energy & Fuels, 12, 191–196.
Sloan JR, E. D. & Koh, C. 2007. Clathrate hydrates of natural
gases, CRC press.
Tohidi, B., Anderson, R., Clennell, M. B., Burgass, R. W. &
REFERENCES Biderkab, A. B. 2001. Visual observation of gas-hydrate
formation and dissociation in synthetic porous media by
Anderson, R., Llamedo, M.,Tohidi, B. & Burgass, R. W. 2003. means of glass micromodels. Geology, 29, 867–870.
Experimental measurement of methane and carbon diox- Tohidi, B., Yang, J., Salehabadi, M., Anderson, R. & Chapoy,
ide clathrate hydrate equilibria in mesoporous silica. The A. 2010. CO2 hydrates could provide secondary safety
Journal of Physical Chemistry B, 107, 3507–3514. factor in subsurface sequestration of CO2. Environmental
Bergman, P. D. & Winter, E. M. 1995. Disposal of carbon science & technology, 44, 1509–1514.
dioxide in aquifers in the US. Energy Conversion and Uchida, T., Ebinuma, T., Takeya, S., Nagao, J. & Narita,
Management, 36, 523–526. H. 2002. Effects of pore sizes on dissociation temper-
Kim, H.-S., Cho, G.-C. & Kwon, T.-H. 2013. Effect of CO2 atures and pressures of methane, carbon dioxide, and
hydrate formation on seismic wave velocities of fine- propane hydrates in porous media.The Journal of Physical
grained sediments. Geochemistry, Geophysics, Geosys- Chemistry B, 106, 820–826.
tems, 14, 1787–1799. Uchida, T., Takeya, S., Chuvilin, E. M., Ohmura, R., Nagao,
Kwon, T.-H., Kim, H.-S. & CHO, G.-C. 2008. Dissociation J.,Yakushev, V. S., Istomin, V. A., Minagawa, H., Ebinuma,
behavior of CO2 hydrate in sediments during isochoric T. & Narita, H. 2004. Decomposition of methane hydrates
heating. Environmental science & technology, 42, 8571– in sand, sandstone, clays, and glass beads. Journal of
8577. Geophysical Research: Solid Earth, 109.
462
Experimental and numerical studies on geomechanical behavior

of various gas hydrate-bearing sediments in China
X.G. Xie & Y.F. Leung

The Hong Kong Polytechnic University, Hong Kong, China
S. Uchida
Rensselaer Polytechnic Institute, New York, USA
J.S. Lu, D.L. Li & D.Q. Liang

Guangzhou Institute of Energy Conversion, Guangzhou, China
ABSTRACT: Gas hydrates exist in pores as a solid, bonding surrounding soil grains together and also densifying
the host sediments. As a result, hydrate-bearing sediments exhibit stiffer, stronger and more dilatant behavior
than hydrate-free sediments. This paper presents experimental and numerical studies to capture these features
of the geomechanical behavior of hydrate-bearing sediments. The experimental data includes triaxial tests on
reconstituted soils of South China Sea and synthetic samples of carbon dioxide hydrate-bearing soils, while the
critical state-based soil constitutive model is calibrated using an optimization-based technique. The results show
that the critical state-based model is capable of predicting shearing response of the hydrate-bearing sediments
under different confining stresses, drainage conditions and degrees of hydrate saturation. It is also found that
the influence of gas hydrates manifests itself mainly through enlargement of the initial yield surface. The
corresponding model parameters are presented for the host sediments in China, which can be adopted for the
simulation of future gas exploration and gas production in China.
1 INTRODUCTION the reserve amounts to half of the sum of petroleum and

natural gas in the mainland (Geng 2012). Cores had
Gas hydrate is a crystalline formed by the intru- been drilled in northern South China Sea in 2007 and
sion of gas (e.g. methane or carbon dioxide) into 2013, while in-situ methane hydrate-bearing samples
water under high pressure and low temperature. have been retrieved at Tibetan plateau in 2009.
These conditions are often found in deep offshore or The first part of this paper presents undrained tri-
permafrost regions, where gas hydrate-bearing sedi- axial compression tests on hydrate-free and methane
ments are formed. Compared with hydrate-free sed- hydrate-bearing samples of South China Sea sand.
iments, higher strength, stiffness and dilatancy can Subsequently, a brief description of the critical state-
be observed in hydrate-bearing samples due to grain- based soil model is presented. The model is incorpo-
bonding and/or pore-filling effects (Masui et al. 2005, rated into an optimization-based calibration approach,
Masui et al. 2007, Miyazaki et al. 2011, Hyodo which allows determination of the critical state param-
et al. 2013). As a result, it is difficult to accurately eters. The approach is also applied to tests conducted
predict the geomechanical behavior of gas hydrate- by other Chinese researchers, and the favorable com-
bearing sediments. The importance of investigating parisons between experimental data and simulation
the geomechanical behavior of the sediments has been results demonstrate that the critical state-based model
highlighted by the large deformations around the well is applicable to both methane and carbon dioxide
encountered in several field-scale gas production tests hydrate bearing sediments.
from gas hydrate-bearing sediments (Dallimore et al.
2008, Yamamoto et al. 2014, Schoderbek et al. 2013).
In recent years, China has invested heavily on gas 2 TRIAXIAL EXPERIMENT ON
hydrate-related researches. It is estimated that China RECONSTITUTED SAMPLES FROM SOUTH
has over 200 billion tonnes of oil equivalent (TOE) CHINA SEA
reserves of methane hydrate deposits, 35% of which
reside in South China Sea region, with the rest in per- 2.1 Apparatus
mafrost regions of Tibetan plateau and Heilongjiang The depth of the gas hydrate stability zone in South
Province (Wang et al. 2010). In South China Sea alone, China Sea extends to 1200–2300 m from sea surface
463
sample preparation process is introduced as follows.
First, the sand was mixed with a predetermined amount
of water to achieve targeted degrees of hydrate satura-
tion (Sh ). Second, the moist sand was put into a mold,
50 mm in diameter and 100 mm in height, covered
by 1 mm thick rubber membrane. Third, in order to
let the hydrate-free sample stand by itself, the sam-
ple was vacuumed from the top and covered. Fourth,
the cell pressure and the back pressure were gradually
increased as pressurized methane gas was injected into
the inner cell. The cell pressure was kept 3 MPa higher
than the back pressure (i.e. gas pressure) during the gas
injection process, and the back pressure was eventually
increased to 12 MPa. Fifth, when pressurization was
Figure 1. Triaxial apparatus used in this study. completed, temperature of the confining fluid in the
cell was lowered to 275 K so that the desired tempera-
ture and pressure conditions to form methane hydrate
were fulfilled. Finally, the back pressure and temper-
ature were kept constant for 24 hours to ensure that
all the water in pores to react with the methane gas.
In other words, after 24 hours the pores of the sam-
ple was filled with either formed hydrate or gas (no
water). The degree of hydrate saturation was calculated
assuming that one floating CH4 molecule is caged in
6 H2O molecules (i.e. n = 6 in CH4·nH2O). Three
samples were prepared: one is hydrate-free specimen
and the two hydrate-bearing specimens with hydrate
saturations of 36% and 42%.
Figure 2. Schematic diagram of the apparatus used in this
study. (1) computer; (2) confining fluid thermostatic tank; (3)
2.3 Triaxial compression test results
data logger; (4) vacuum pump; (5) circulation pump; (6) load
cell; (7) specimen; (8) buffer tank; (9) methane gas; (10) water In the three tests, cell pressure was 15 MPa while the
pump for inner cell; (11) gas pump; (12) confining pressure gas pressure was 12 MPa, keeping the effective con-
pump; (13) hydraulic oil pump for axial load. fining pressure σc to be 3 MPa. The temperature inside
the cell was kept at 275 K during shearing. A constant
and approximately 100 m from seabed (Trung 2012, strain rate of 0.2%/min was adopted.
Wu et al. 2008), where the pressure is over 10 MPa, Figure 3 presents the development of deviator stress
temperature is around 275 K (Wu et al. 2005), and the with axial strain. It can be seen that the existence
effective stresses is in the range of 1 MPa. In order to of hydrate strengthened the sediments by approxi-
simulate these pressure and temperature conditions, mately three times for Sh = 36 % and four times
a triaxial apparatus that could sustain high pressure for Sh = 42 % sample, respectively. The hydrate-
and provide low temperature during shearing is assem- bearing samples also showed increase in the stiffness.
bled at the Guangzhou Institute of Energy Conversion, The hydrate-bearing specimens are associated with
China. The cell of this apparatus could resist a pres- more dilatant behavior upon shearing, as shown by the
sure up to 30 MPa while the loading arm capacity is strain-softening response post-peak. These corrobo-
250 kN. The apparatus is also equipped with a ther- rate the previous findings of geomechanical behavior
mostatic tank which controls the temperature of the of gas hydrate-bearing sediments by Masui et al.
confining fluid from 243 K to 323 K with an accuracy (2005), Masui et al. (2007), Miyazaki et al. (2011),
of 0.5 K, by circulating 50% ethylene alcohol solution. Hyodo et al. (2013), etc.
Synthesized hydrate-bearing samples can be formed
and sheared within the same chamber. Possible dis-
sociation or depressurization during transportation is 3 METHANE HYDRATE CRITICAL STATE
thus avoided. Figure 1 shows the apparatus and Figure MODEL
2 shows the schematic diagram of the test setup.
Uchida et al. (2012) developed the methane hydrate
critical state (MHCS) model based on the critical
2.2 Sample preparation
state framework (Roscoe et al. 1958, Roscoe &
Sand from gas hydrate-bearing layer in South China Burland 1968) and the model is able to capture the
Sea was used as host sediments in this study and essence of geomechanical behavior of hydrate-bearing
partial water saturation (PWS) method was adopted sediments. The main features of the model are: (a)
to form methane hydrate within sediment pores. The to incorporate the strength increase due to hydrates
464
where u is the material parameter that incorporates
pre-yield plasticity. The fourth feature (d) leads to
addition of the hydrate effect on the elastic stiffness as
where K is the soil bulk modulus which is func-

tion of the mean effective stress in the critical state
framework, ν is the Poisson’s ratio, Eh0 is the mate-
rial property that determines the effect of hydrate on
stiffness.
Therefore, in addition to the conventional crit-
ical state model, the MHCS model employs four
hydrate-dependent parameters (α, β, m andEh0 ) and
one hydrate-free parameter (u).
4 MODEL CALIBRATION
4.1 Testing database

Figure 3. Triaxial test result on methane hydrate bearing In order to simulate the experiments on hydrate-
South China Sea sand.
bearing sediments in China, a database is established
consisting of the results shown in Figure 3 and tri-
by enlarging the yield surface of the hydrate-free
axial compression tests by other Chinese researchers
sediments, which implicitly leads to the increase in
reported in the literature. The database includes tests
dilatancy; (b) to capture shearing-induced degradation
on both methane and carbon dioxide hydrate-bearing
effect of mechanical contribution from hydrates; (c) to
soils. Details of these experiments are listed in Table 1.
exhibit smooth transition from linear elastic to plastic
behavior; and (d) to include the effect of hydrates on
the elastic stiffness. The first and third features lead 4.2 Calibration method
to modification of the conventional critical state yield
As stated in Section 3, the parameters of MHCS
function (Roscoe & Burland 1968) such that:
model can be classified into two groups, hydrate-free
parameters {M , λ, κ, ν, pcs , u} and hydrate-dependent
parameters {m, Eh , α, β}. For each study listed in Table
1, hydrate-free parameters are calibrated with hydrate-
where f is the yield function, q is the deviator stress, free specimens (Sh = 0) before hydrate-dependent
M is the stress ratio at the critical state, p is the mean parameters are derived based on hydrate-bearing spec-
effective stress, pcs is the yield stress for hydrate free imens.
specimens, R is the subloading ratio (described later) The calibration process is essentially an optimiza-
and pcd is the strength and dilatancy enhancement due tion problem, whose target is to obtain the ‘optimal’
to hydrates, which is defined as: set of MHCS model parameters that would lead to
minimum differences between the experimental data
and the simulated curve. In this study, one of the
most efficient optimization methods, Differential Evo-
where α and β are model parameters and Shmec is lution (DE) (Storn & Price 1997), is adopted to find
the mechanical hydrate saturation. This mechanical this best set of parameters. DE is a derivative-free
hydrate saturation is utilized to represent the second optimization technique and is conceptually similar to
feature (b) such that: other evolutionary algorithms such as genetic algo-
rithm. In the DE optimization process, a population
of N candidate solutions is first generated randomly
(Section 4.2.1). The candidate solutions are vectors
p
of MHCS parameters, known as trial vectors. The
where d is the plastic deviatoric strain and m is the algorithm then explores the search space by vector
material parameter that determines the relationship difference of the various candidate solutions. At each
between the degradation effect and the plastic devi- iteration or generation, ‘mutant vectors’ are formed
atoric strain. The subloading ratio is introduced in Eq. by linear interpolation or extrapolation of trial vectors
(1) for the third feature (c) , and its form was first sug- randomly selected from the population. A new gener-
gested by Hashiguchi (1989), with its evolution given ation of trial vectors is then formed by the ’crossover’
by: process, whereby the components of mutant vectors
are mixed with those of the trial vectors in the previous
generation (Section 4.2.2).
465
Table 1. Conducted triaxial compression tests on hydrate bearing sediments in China
Hydrate type σc (MPa) Sh Test Specimen type Reference
Methane 3 0, 36, 42% Undrained Gas saturated This study

1, 2, 4 0, 16, 36, 56% Water saturated Sun et al. (2013)
Carbon dioxide 1, 1.5, 2 0, 20, 37, 51% Drained Water saturated Wei et al. (2011)
Table 2. Parameter range for the optimization A new generation of trial vectors (U) is formed by
the crossover of old trial vectors (X) and the mutant
Type Parameter Min Max vectors (V):
Hydrate M 1.0 1.7
free λ 0.1 0.4
κ 0.001 0.3
u 1.0 5.0
ν 0.10 0.40 where Cr (0 < Cr < 1) is the crossover probability. jrand
pcs (MPa) 3 20 is a value from 1 to 6 (the number of parameters) which
Hydrate m 1 50 ensures at least one element is from the mutant vector.
dependent Eh0 (GPa) 0.01 10.
α (MPa) 1 200 4.2.3 Selection
β 0.1 5.0
The fitness of vectors Xi,G and Ui,G+1 are compared
through a cost function to determine their survivabil-
ity. In simple words, the set of parameters that lead
to closer match with the experimental data is the bet-
ter (fitter) solution, and will remain in the population.
Fitness of Xi,G (parent, in generation G) and Ui,G+1 Since only deviator stress data is available, the cost
(child, in generation G + 1) are evaluated and com- function is defined in the current study as:
pared through the ‘selection’ process. The fitness
determines the survivability of the particular solution
– the fitter solutions stay in the population, while the
weaker ones will be discarded. In the current context,
the fitness is defined by a cost function (Section 4.2.3).
The comparison is performed for each candidate solu-
tion (i from 1 to N ), and the processes are iterated until
the population converges to a global optimum solution
(Section 4.2.4). The following subsections describe the
details of the optimization procedures.
4.2.1 Initialization where c is the number of cases (with different confin-

The population size of candidate solution, N , is set to ing stresses or hydrate saturations) considered. C1 (X)
be 10 times the number of parameters to be optimized and C2 (X) indicate the differences between experi-
(Storn & Price 1997). In this study, 6 hydrate-free mental and simulation curves. C1 (X) is the error of
parameters are considered, and therefore a popula- deviator stress curve and C2 (X) is the deviation of the
tion consisting of 60 trial vectors {X1 , X2 , . . ., X60 } slope of deviator stress curve concerned. C(X) is com-
is randomly generated. In this process, the upper and pared with C(U) and the vector that leads to the lower
lower bounds must be specified and their values are cost C will stay in the population while the other one
summarized in 4.2.2. is discarded.
4.2.2 Mutation and crossover 4.2.4 Convergence

To form a mutant vector (Vi ), three different trial vec- In this study, when the populations have converged
tors need to be randomly selected from the population to the same solution (i.e. X1 = X2 = · · · = X60 ), the
(Xr0 , Xr1 and Xr2 ). The difference between Xr1 and optimization is complete. The resulting solution is
Xr2 is then scaled by a factor F through interpolation the calibrated parameter set that produces the closest
or extrapolation, and the scaled difference is added to match with experimental data.
Xr0 to become the mutant vector. Mathematically, this
process is expressed as: 4.3 Calibration results and discussions
Table 3 summarizes the calibrated parameters obtained
from the optimization. The average value of each
466
Table 3. MHCS parameters from optimization
This Sun et al. Yan et al. Wei et al.

Parameters study (2013) (2012) (2011)
M 1.7 1.7 1.15 1.23

λ 0.37 0.4 0.1 0.4
κ 0.007 0.017 0.002 0.009
u 1.8 4.6 2.3 3.7
v 0.10 0.1 0.1 0.4
pcs (σ3 ) 14.5 9.0 (1) 3.0 (1) 3.0 (1)
(MPa) 14.1 (2) 3.1 (2) 4.6 (1.5)
15.0 (4) 3.2 (3) 7.3 (2)
m 2 30 4 1
Eh0 (GPa) 10 4 10 1
α (MPa) 183 29 32 11
β 1.1 0.9 0.6 0.9
∧
C1F 2% 5% 3% 2%
#
C1B 16% 13% 9% 9%
ˆ Average C1 for hydrate free specimens

# Average C1 for hydrate bearing specimens
Figure 4. Triaxial compression test on South China Sea sand

(this study).
C1 is also presented. Compared to the experimental Figure 5. Triaxial compression test by Sun et al. (2013) and
data, the model results in less than 5% deviation for MHCS simulation.
hydrate-free specimens and less than 16% deviation
for hydrate-bearing specimens. Figure 4 to Figure 7
show the comparisons between experimental data and 5 CONCLUSIONS
model simulations. The results clearly demonstrate
that the MHCS model, which is simple extension of This paper presents triaxial tests conducted on methane
the critical state model, can capture the essence of the hydrate-bearing soil, synthesized with sands recov-
geomechanical behavior of methane hydrate-bearing ered from South China Sea. Under the same effective
sediments. It is also worth noting that although the confining pressure, the stiffness and peak strength are
MHCS model was developed for methane hydrate- found to increase with increasing degrees of hydrate
bearing sediments, it is also applicable to carbon saturation. Strain-softening behavior is also observed
dioxide hydrate-bearing sediments. in the gas hydrate-bearing specimens.
467
Figure 6. Triaxial compression test by Yan et al. (2012) and Figure 7. Triaxial compression test by Wei et al. (2011) and
MHCS simulation. MHCS simulation
bearing-sediments. The good comparisons between

A global optimization method, DE, was adopted model simulations and tests from various sources
in the study to verify the applicability of the MHCS demonstrate the generality of the model.
model. Favorable matching was achieved, even though
the model only introduces 4 hydrate-dependent param-
eters and 1 hydrate-free parameter in addition to the ACKNOWLEDGEMENTS
conventional critical state model parameters. More-
over, the study shows that the model can be applied The authors would like to acknowledge the finan-
to represent the behavior of carbon dioxide hydrate cial supports from the Hong Kong PhD Fellowship
468
Scheme, Rensselaer Polytechnic Institute Junior Fac- Schoderbek, D., Farrell, H., Hester, K., Howard, J., Raterman,
ulty Startup Fund, Chinese Academy of Science (Grant K., Sipngarmlert, S., Martin K., Smith, B., & Klein, P.
No. KGZD-EW-301), National Oceanic Geological (2013). ConocoPhillips gas hydrate production test final
Special Project (Project No. GHZ2012006003) and technical report. United States Department of Energy.
Storn, R. & Price, K. (1997). Differential evolution – a simple
National Natural Science Foundation of China (Grant and efficient heuristic for global optimization over con-
No. 41276043, 51474197). tinuous spaces. Journal of Global Optimization, 11(4),
341–359.
Sun, Z.M., Zhang, J., Liu, C.L., Zhao, S.J., & Ye, Y.G. (2013).
REFERENCES Experimental study on the in situ mechanical properties of
methane hydrate-bearing sediments. Applied Mechanics
Dallimore, S.R., Wright, J.F., Nixon, F.M., & Schlumberger, and Materials, 275, 326–331.
K.K. (2008). Geologic and porous media factors affecting Trung, N.N. (2012). The gas hydrate potential in the South
the 2007 production response characteristics of the JOG- China Sea. Journal of Petroleum Science and Engineering,
MEC/NRCAN/AURORA Mallik Gas Hydrate Production 88, 41–47.
Research Well. In Proceedings of the 6th International Uchida, S., Soga, K., & Yamamoto, K. (2012). Critical state
Conference on Gas Hydrates. Vancouver. soil constitutive model for methane hydrate soil. Journal
Geng, W.H. (2012). Combustible ice’s commercial explo- of Geophysical Research: Solid Earth, 117(B3), B03209.
ration timetable. Land and Resources Information, Wang, Z.M., Qu, H.L., & Jian, Z.J. (2010). Combustible
2012(7), 23–25. ice’s distribution and development status in China. Energy
Hashiguchi, K. (1989). Subloading surface model in uncon- Conversion, 334(5), 4–5.
ventional plasticity. International Journal of Solids and Wei, H.Z., Yan, R.T., Chen, P., Tian, H.H., Wu, E.L., & Wei,
Structures, 25(8), 917–945. C.F. (2011). Deformation and failure behavior of carbon
Hyodo, M., Yoneda, J., Yoshimoto, N., & Nakata, Y. (2013). dioxide hydrate-bearing sands with different hydrate con-
Mechanical and dissociation properties of methane tents under triaxial shear tests. Rock and Soil Mechanics,
hydrate-bearing sand in deep seabed. Soils and Founda- 32(Supp. 2), 198–203.
tions, 53(2), 299–314. Wu, N.,Yang, S., Zhang, H., Liang, J., Wang, H., Su, X., & Fu,
Masui, A., Haneda, H., Ogata, Y., & Aoki, K. (2005). Effects S. (2008). Preliminary discussion on gas hydrate reservoir
of methane hydrate formation on shear strength of syn- system of Shenhu Area, North Slope of South China Sea.
thetic methane hydrate sediments. In Proceedings of In Proceedings of the 6 th International Conference on Gas
the Fifth International Offshore and Polar Engineering Hydrates. Vancouver.
Conference ISOPE. Seoul. Wu, S. G., Zhang, G. X., Huang, Y. Y., Liang, J., & Wong, H.
Masui,A., Haneda, H., Ogata,Y., &Aoki, K. (2007). Mechan- K. (2005). Gas hydrate occurrence on the continental slope
ical properties of sandy sediment containing marine gas of the northern South China Sea. Marine and Petroleum
hydrates in deep sea offshore Japan. In Proceedings of Geology, 22(3), 403–412.
the seventh International Offshore and Polar Engineering Yamamoto, K., Terao, Y., Fujii, T., Terumichi, I., Seki, M.,
Conference ISOPE Ocean Mining Symposium. Lisbon. Matsuzawa, M., & Kanno, T. (2014). Operation overview
Miyazaki, K., Masui, A., Sakamoto, Y., Aoki, K., of the first offshore production test of methane hydrates
Tenma, N., & Yamaguchi, T. (2011). Triaxial compres- in the Eastern Nankai Trough. In Offshore Technology
sive properties of artificial methane-hydrate-bearing sed- Conference. Houston.
iment. Journal of Geophysical Research: Solid Earth, Yan, R.T., Wei, C.F., Wei, H.Z., Tian, H.H., & Wu, E.L. (2012).
116(B6), B06102. Effect of hydrate formation on mechanical strength of
Roscoe, K.H., Schofield, A.N., & Wroth, C.P. (1958). On the hydrate-bearing sand. Chinese Journal of Geotechnical
yielding of soils. Geotechnique, 8(1), 22–53. Engineering, 34(7), 1234–1240.
Roscoe, K.H. & Burland, J.B. (1968). On the generalized
stress-strain behavior of wet clays. Engineering Plasticity
(p. 535-609). Cambridge University Press.
469
A new software collection for 3D processing of X-ray CT images
S. Chauhan, W. Rühaak & I. Sass

Technische Universität Darmstadt, Institut für Angewandte Geowissenschaft, Fachgebiet,
Angewandte Geothermie, Darmstadt Deutschland
H. Anbergen, A. Kabdenov, M. Freise & T. Wille

APS Antriebs-, Prüf- und Steuertecknik GmbH, Rosdorf, Germany
ABSTRACT: The identification of accurately segmented phases in images observed through X-ray microcom-
puter tomography (XCT) is vital towards characterizing uncertainties involved in determining the geometries
of pore network. Currently popular methods such as histogram, thresholding which are commonly used for
XCT image segmentation exhibit a number of shortfalls. In this paper a new software is proposed, which is
based on machine learning (ML) techniques, for the 2D/3D visualization of XCT data. The segmentation and
classification of different phases are based on feature vector selection. Hence relative porosities and trends
in pore size distribution can be computed. In this study, the computational performance is optimised using
correlation-based feature vector selection, demonstrated using unsupervised, supervised and ensemble ML tech-
niques. Furthermore, accuracies of ML techniques are accessed based on entropy, purity, and receiver operation
characteristics.
1 INTRODUCTION accuracy and therefore decreasing the magnitudes of

the uncertainties involved in determining the geome-
The APS GmbH is developing a new triaxial testing tries of pore networks (Chauhan et al., 2016).
system for methane hydrate research. Embedded in Machine learning (ML) techniques provide a
the latest research project “SUGAR 3” a high-pressure promising alternative to segment and classify different
triaxial system is equipped with high-resulution voxel phases from XCT rock images. The software tool
X-ray computer tomography (XCT). In order to get we propose used unsupervised, supervised and ensem-
most accurate results, a new software solution was ble classifier techniques to extract relative porosity and
developed that analyses objectively obtained XCT- pore size distribution 3D XCT gray scale images. The
images. This software solution was developed in coop- study validated the accuracy and performance of these
eration between APS and the Institute of Applied ML techniques.
Geothermal Science of the Technische Universität
Darmstadt.
Accurate segmentation of the phases from the XCT 2 EXPERIMENTAL APPROACH
rock images are the main building block for a reliable
prognosis of the transport processes, elastic and elec- For this study we used a volcanic rock Andesite and
tric properties; simulated using digital rock physics Rotliegend Sandstone taken from Tongariro National
(DRP) models (Fusseis et al., 2014; Chauhan et al., Park, New Zealand and Rotliegend Germany Shown in
2016;). The segmentation problem is reduced to the Figure. 1.Andesite has an effective porosity of 17±2 %
need to quantify the binary solid–void phase distri- and Sandstone of 14 ± 2% measured using a GeoPyc
bution (i.e., a binarization problem) when modelling pycnometer (Micromeritics Instrument Corporation
fluid transport at the pore scale. However, Leu et al. Norcross, GA, USA). Andesite has a porphyritic tex-
(2014) recently performed a sensitivity study in which ture with large plagioclase crystals (up to 3 mm in
they showed that even a small bias in the accuracy of diameter), pyroxene in a cryptocrystalline matrix, and
the binarization may lead to a significant error in the isolated vesicles up to 6 mm in diameter (Chauhan
calculated permeability. Binarization is an essential et al., 2016). Whereas, Rotligend Sandstone had dif-
prerequisite of DRP studies, but there are few accurate ferent grain size (between 0.5 to 5 mm) of fine
and fast binarization algorithms that are not biased by sand and gravel, with monocrystalline quartz 26%,
manual (subjective) interventions by the user. Choos- poly-crystalline quartz up to 35% Feldspate 8%, sedi-
ing an appropriate scheme to binarize an image is key mentary volcanic lithoclast grains 9% along with and
to characterizing a porous space with a good degree of 13% cement.(Aretz et al ., 2013).
471
Figure 1. Top panel shows the rock sample Andesite and Sandstone which were used for XCT. Middle panel and the bottom
panel shows the raw images and histogram plots of the raw image of the respective samples. The mineral composition of
Andesite and Sandstone was determined by thin sections.
The samples were imaged using a high resolu-

tion XCT scanner (Chauhan et al 2016), applying
X-ray energy of 110 keV and using a pre-filter
of 0.3 copper. During the reconstruction of the
projections no noise filter was used. The projec-
tions were Radon-transformed in sinograms, there-
after converted through back-projection into tomo-
grams. These stacked tomograms resulted in a 16-bit
3D imagery, with a resulting voxel resolution of
13 µm and 21 µm for Andesite and Sandstone respec-
tively.Andesite required no beam hardening correction
(BHC), whereas BHC for Sandstone was done using
based on regression analysis using 2D paraboloid Figure 2. Schematic illustration of our proposed method
fitting. Finally, the tomograms are saved in raw format (Fig.2 has been modified after Chauhan et al., 2016).
3.1 Unsupervised techniques

3 MACHINE LEARNING In the unsupervised technique K-means (MacQueen
1967), Fuzzy C-means (FCM) (Dunn 1973) and self-
The graphical user interface (GUI) uses machine learn- organized Maps (SOM) (Kohonen 1990) were used
ing (ML) algorithms to map pixels of similar values in for segmentation pore, mineral and matrix phases.
to respective classes. The ML algorithms implemented The unsupervised algorithms were configured to per-
fall in the catogories of unsupervised, supervised and form segmentation of three to seven classes. These
ensemble classifiers. Figure. 2. illustrates the pro- classes in one-dimensional feature space are the non-
cessing scheme implemented in the GUI. The image overlapping segments of pixel bins in a histogram.
pre-processing was done according to Chauhan et al. Filter based feature vector selection (Euclidian and
(2016). Manhattan distance function) were used to initialize
472
centroids for K-means, FCM and SOM. In the case weak classfiers. The main difference between Brag-
of FCM different degree of membership values [1.10 ging and RUSBoost is the way they train their weak
to 1.85] were tested to ‘loosely’ or ‘tightly’ segregate classifiers. Bragtree in an iterative scheme, train its
pixel values between mineral and matirx phase. grid classifiers with randomly chosen samples from the
topology was chosen in the case of SOM. training data set, in the second step collects the mis-
classified instances and retrains its classifiers until
the misclassification error is minimized (Chau-han et
3.2 Supervised techniques
al 2016). Whereas, RUSBoost sequentially trains its
In the supervised category feed forward artificial neu- classifiers using the whole training later, essentially
ral Network (FFANN) (Jain et al., 1999) and least focusing on retraining inaccurate classifiers with the
square support vector machine (LS-SVM) (Suykens large data set until its misclassification error is min-
and Vandewalle. 1999) were used to classify pore, minimized. The ensemble classifiers where trained using
eral and matrix phases (Chauhan et al., 2016). The same feature vector (FV) used for LS-SVM, with a
supervised algorithms rely on a classification model minimum leaf size of five and learning rate of 0.1.
which has to be trained using example set of data that
represent each class. In the case FFANN the classifica- 4 FEATURE REDUCTION
tion model was trained using segmented dataset from
K-means, FCM, and SOM. For LS-SVM a training In a practical rock CT segmentation/classification task
data set was created, which contained range of pixel a set of apriori information in the form of most useful
values which best represented pore, mineral, matrix pixel values is given to ML algorithms for segmenta-
and noise regions, these pixel ranges where further tion or training the classification model. This dataset
labelled in to different classes, which ranged from containing apriori information is termed as feature
one to seven. For FFANN and LS-SVM the models vectors (FV). For unsupervised K-means, FCM, SOM
were tuned using ten-fold cross-validation function a set of ten XCT images were used to develop the FV.
(repeated training and testing) and misclassification For FFANN five images out of ten were used to train
rate was determined using mean square root error the net-work and LSSVM and ensemble based classi-
(MSE) in the case of FFANN. Once the classifica- fiers a set pixel values which best represented the pore,
tion model reached an optimal performance threshold mineral, matrix and noise regions were used a fea-
it was tested on rest of the XCT slices. ture vectors. The unsupervised m ML techniques were
tested on ten slices of XCT data 31,577,290 pixels.
3.3 Ensemble classifier techniques FFANN was trained using 15,788,645 feature vectors
and tested on 31,57,290 pixels. For LS-SVM brag-
In the supervised category feed forward artificial neu- ging and boosting the classification model was trained
ral Network (FFANN) (Jain et al., 1999) and least using 2,007 feature vectors and tested on unknown data
square support vector machine (LS-SVM) (Suykens sample of 31,57,290 pixels.
and Vandewalle. 1999) were used to classify pore, min-
eral and matrix phases (Chauhan et al., 2016). The
supervised algorithm relies on a classification model 5 PERFORMANCE AND ACCURACY
which has to be trained using example set of data that
represent each class. In the case FFANN the classifica- Computational performance was measured in terms
tion model was trained using segmented dataset from of the segmentation and classification speed of the
K-means, FCM, and SOM. For LS-SVM a training ML algorithms shown in table 2. Test were performed
data set was created, which contained range of pixel on Windows Server 2008 R2 Standard 64-bit Operat-
values which best represented pore, mineral, matrix ing System, with two processor Intel(R) Xenon (R),
and noise regions, these pixel ranges where further CPU: E645 2.40 GHz and Installed memory (RAM)
labelled in to different classes, which ranged from of 48.0 GB. For unsupervised techniques accuracy
one to seven. For FFANN and LS-SVM the models or cluster validation was performed to identify ideal
were tuned using ten-fold cross-validation function class (es), representing the ‘best’ porosity values and
(repeated training and testing) and misclassification to compare the clustering approaches. External valida-
rate was determined using mean square root error tion measures ‘Purity’ and ‘Entropy’ was performed
(MSE) in the case of FFANN. Once the classifica- on all the pixels corresponding to the classes three
tion model reached an optimal performance threshold to seven. The Purity and Entropy measure the ability
it was tested on rest of the XCT slices. of the clustering method to recover the know classes,
despite number of classes are different from number
of segementable classes (Jain et al., 1999). Purity is
a real number between [0,1], larger the purity val-
3.4 Ensemble classifier techniques
ues, so better is the clustering method. Conversely, the
In the ensemble classifier technique RUSBoost and lower the Entropy value better is the clustering per-
Bragtree algorithms are used (Seif-fert et al., 2008; formance. In the case of FFANN, an objective method
Breiman, 1996) to classify pore, mineral and matrix to determine the classification criterial is by calcu-
phases (Chauhan et al., 2016). In general ensamble lating the mean square root error (MSE) between the
classifiers is a ‘bootstrap aggregation’ of different output and the targets. Lower the MSE value better is
473
the classification, zero corresponds to no misclassifi- the ROC curve represents the accuracy of the clas-
cation. For LS-SVM receiver operation characteristic sification model. The area of 1 represents a prefect
(ROC) curve was plotted to compute the accuracy, classification; an area of 0.5 represents worthless clas-
ROC curve give the quality of the classification model. sification (Khan et al., 2016). In case of Bragging and
It shows a tradeoff between the sensitivity of the clas- Boosting misclassification cost of the weak classifiers
sification model; with respect to the specificity with gave an estimate of the this was performed using Kfold
which it can classify unknown data set. The area under cross-validation technique.
6 RESULTS AND ANALYSIS

Table 1. Computation time for processing ten slices
CPU: Time (hrs:min:sec)

6.1 Esitmated porosity and pore size distribution
Machine learning The porosities was determined from the stack of ten
Techniques Andesite Sandstone Musli XCT slices for three to seven classes using differ-
ent ML techniques are shown in the left panel of the
K-means 00:15:35 00:12:04 0:10:59 Figure 3 and the segmented images are shown in Fig-
FCM 00:29:19 00:56:03 00:42:21
SOM 01:07:06 17:35:08 01:11:23
ure 4. The estimated porosity is the ratio between the
FFANN(traning 08:58:18 pore phase voxels and entire sample volume multi-
using K-means) plied by 100. In case of Sandstone and Musli estimated
FFANN(traning 06:35:43 porosity was determined using unsupervised tech-
using FCM) niques whereas for Andesite all the ML techniques
LS-SVM 63:29:35 were used. In general, the porosity using unsupervised
Bragging 05:57:05 ML techniques agree well with each other for all the
Boosting 07:47:05 three samples. For Andesite, Sandstone and Musli, the
Figure 3. The right panel shows relative porosity using ma-chine learning algorithms for different rock samples. Middle
panel shows the volume fraction of different phases quantified using machine learning techniques and the right panel show
the pore size distribution of different sample using watershed technique.
474
averaged estimated porosity sum over all classes is
15.8 ± 2.5%, 14.8 ± 8.8% and 50.9 ± 13.3% respec-
tively. This is in good agreement to the experimental
porosity values obtained for Andesite and Sandstone
using GeoPycpynometer. The large standard deviation
in the case of Sandstone and Musli is caused by FCM
segmentation scheme. When the membership func-
tion is tightly constrained [1.10, 1.35] the segregation
between pore phase voxels and pore throat voxels is
underestimated contributing to the increase in poros-
ity. Conversely, when membership function loosely
constrained [1.60, 1.85] pore throat and micro pores
are segmented as matrix phases resulting in decrease
in porosity and increase in matrix phase, which is
clearly visible in volume fraction plot of Sandstone
and Musli in the middle panel. The low standard devi-
ation in the estimated porosity values of Andesite is
due to the absence of micro porosity and intercon-
nected pores. The pore, mineral and matrix phases are
distinct from each other therefore the ML techniques
have less difficult in segmentation and classification
Pore size distribution (PSD) of Andesite, Sandstone
and Muli was computed using the method suggested
by Rabbani et al., (2014). The segmented gray scale
images where first converted to binary images using
thresholding technique. Morphological and filtering
operations were performed based on the complex-
ity of the segmented images. Distance transform to
convert the bright area into catchment basin and
later watershed transformation was performed to seg-
ment the pore boundaries. The bottom panel in the
Figure 3. Show the PSD and average pore radius
of Andesite, Sandstone and Musli from K-means
segmented images.
6.2 Performance and accuracy

The performance was evaluated based on processing
time, where, K-means was the fastest and LS-SVM the
slowest. Accuracy was computed for Andesite sample
only shown in Figure 5. Entropy values of unsuper-
vised techniques, means square root error (MSE) and Figure 4. The top, middle and last panel shows the 2D
receiver operational characteristics (ROC) of super- segmented images and volume rendered plot of respective
vised techniques (FFANN and LS-SVM respectively) samples using unsupervised networks (Andesite figure has
are shown in figure.5. Lowest entropy values are been modified after Chauhan et al 2016).
shown by K-means and SOM for classes three and
four. This agrees well with the histogram plot in fig- 7 CONCLUSIONS
ure 1. Therefore, the porosity values obtained from
class three and four are best representative porosity We propose a software tool to visualize and analy-
values of Andesite. In the case of FFANN the mis- ses 3D high resolution x-ray computer tomographic
classification rate can be estimated by MSE values. (XCT) rock images. The software tool is based
The lowest MSE value implies better is the classifi- on unsupervised, supervised and ensemble machine
cation accuracy. FFANN was trained using segmented learning (ML) techniques. In this study we have eval-
image of K-means and FCM and then was tested on uated the performance and accuracies of the ML
raw images of Andesite sample. The best classification techniques to compute relative porosity and pore
rate was obtained when FFANN was trained with K- size distribution in Sandstone (inter connected pores),
means (0.02; for class three) and worst when trained Musli (micro porosity) and Andesite (alterated min-
using FCM (1.35; for class seven). ROC values for erals) rock samples. The relative porosity values
LS-SVM show that, LS-SVM is able to classify pore of 15.8 ± 2.5%, 14.8 ± 8.8% and 50.9 ± 13.3% for
phases and matrix with 100% accuracy. Whereas, it is Andesite, Sandstone and Musli are in very good agree-
able to differentiate between mineral and matrix phase ment with the experimental values of 17 ± 2 and
up to 70%. 14 ± 2 obtained using GeoPycpynometer.
475
Figure 5. top left show the entropy values obtained for Andesite sample segmented using unsupervised techniques. Top right
show mean square root error obtained for Andesite sample classified using feed forward artificial neural network (FFANN).
The FFANN was trained using k-means, Fuzzy C-means with membership function [1.10, 1.85]. The bottom panel shows the
receiver operational characteristics of LS-SVM for classified class four.
In terms of computational time, K-means outper- REFERENCES

forms all the other ML techniques. Fuzzy c.means
can distinguish well between pore and pore-throat Aretz, A., Bär, K., Götz, A. E., & Sass, I. (2015). Out-
boundaries- when the membership function is loosely crop analogue study of Permocarboniferous geothermal
sandstone reservoir formations (northern Upper Rhine
constrained between 1.60–1.85. Unsupervised ML Graben, Germany): impact of mineral content, depo-
techniques perform well with filter based feature sitional environment and diagenesis on petrophysical
extraction techniques. We found that different tun- properties. International Journal of Earth Sciences, 1–22.
ing parameters (such as different FCM membership Breiman, L. (1996). Bagging predictors. Machine learning,
criteria and different SOM topologies and distance 24(2), 123–140.
functions) need to be tested for the unsupervised tech- Chauhan, S., Rühaak, W., Khan, F., Enzmann, F., Mielke, P.,
niques. A SOM topology “gridtop” layout (neurons Kersten, M., & Sass, I. (2016). Processing of rock core
arranged in a grid format), and a SOM Manhat- microtomography images: Using seven different machine
tan distant function (sum of the absolute difference) learning algorithms. Computers & Geosciences, 86, 120–
128
gave consistent results and FCM membership func- Dunn, J. C. (1973). A Fuzzy Relative of the ISODATA Pro-
tion between [1.35–1.85] gave consistent results. In cess and Its Use in Detecting Compact Well-Separated
the case of supervised techniques feed forward neural Clusters. Journal of Cybernetics, 3, 32–57.
networks when trained with K-means feature vec- Jain, A. K., Murty, M. N., Flynn, P. J. (1999). Data clustering:
tors contributed to less misclassification error of a review. ACM computing surveys, 31(3), 264–323.
0.16 compared to using FCM feature vectors; which Kohonen, T. (1990). The self-organizing map. Proceedings
gave a misclassification error of 0.36. The LS-SVM, of the IEEE, 78(9), 1464–1480.
bragging, and boosting, are driven by feature vectors MacQueen, J. (Ed.). (1967). Some Methods for classification
that are manually selected, and performed extremely and Analysis of Multivariate Observations. University of
California Press.
well despite a small subset of feature vectors where Suykens, J. A., Vandewalle, J. (1999). Least Squares Support
chossen to train the classification model. But in terms Vector Machine Classifiers. Neural Pro-cessing Letters,
of computational time LS-SVM performed the worst. 9(3), 293–300.
ACKNOWLEDGEMENTS
The project was funded by the Federal Ministry of Edu-

cation and Research (grant no. 03SX381H). The sole
responsibility for the contents lies with the authors
476
Predicting deformation of MHBS during hydrate dissociation

using a state-dependent critical state model
J. Shen, C.F. Chiu, G.H. Lei & J. Xu

Key Laboratory of Geomechanics and Embankment Engineering of the Ministry of Education,
Hohai University, China
C.W.W. Ng
Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology,
HKSAR, China
ABSTRACT: Methane Hydrate-Bearing Sand (MHBS) is a natural soil deposit containing methane hydrate
in its pores, which occurs in abundance in deep water marine sediments and permafrost regions. Submarine
landslides may be induced by methane hydrate extraction in deep water. Thus, it is crucial to evaluate the
mechanical behaviour of MHBS during hydrate dissociation. In this paper, a recently developed state-dependent
critical state model for MHBS was further extended to predict the effect of hydrate dissociation on volume
change and axial deformation of MHBS. In the formulation, the volume change of MHBS was derived as a
combined action of the change in stress, hydrate saturation and temperature. However, the axial deformation
is only affected by the change in stress and hydrate saturation. The proposed model was used to predict the
laboratory hydrate dissociation tests induced by the thermal recovery method reported in Hyodo et al. (2013).
It is found that despite some discrepancies the model predictions can capture some key deformation features of
MHBS during hydrate dissociation. In particular, shear failure of MHBS can be predicted for specimen subjected
to a high shear stress level.
1 INTRODUCTION demonstrated that the model can adequately capture

the stress-strain and volume change behaviours of
Methane Hydrate-Bearing Sand (MHBS) is a natural MHBS over a wide range of hydrate saturations, con-
soil deposit containing methane hydrate in its pores, fining pressures and densities using a unified set
which occurs in abundance in deep water marine sedi- of parameters. In this paper, the model is further
ments and permafrost regions. MHBS is considered as extended to predict the effect of hydrate dissociation
a potential energy source for the future (Kvenvolden on mechanical behaviour of MHBS.
1999). Depressurisation and thermal recovery are the
two most feasible methods to extract methane from
MHBS resulting in hydrate dissociation. Past stud- 2 MODEL FORMULATION
ies (Masui et al. 2005, Miyazaki et al. 2011, Hyodo
et al. 2013, Clayton et al. 2010) have shown that both 2.1 State-dependent critical state model for MHBS
the strength and stiffness of MHBS are greater than
those of clean sand because hydrate may either fill the The state-dependent critical state model is proposed
pores or form part of sand skeleton, as well as cement under the frameworks of elasto-plasticity, critical state
sand grains at inter-granular contacts. As the strength and state-dependent dilatancy. The yield criterion for
and stiffness enhancement of hydrate vanish during its MHBS in the mean effective stress (p) – deviatoric
dissociation in methane extraction, significant defor- stress (q) space (illustrated in Figure 1) is presented as
mation of MHBS may occur leading to submarine follows:
landslide and seabed subsidence (Mienert et al. 2005,
Nixon & Grozic 2007). Thus, it is crucial to evaluate
the deformation of MHBS during hydrate dissociation
in deep water methane extraction. where η∗ is the stress ratio at yielding, defined as
A state-dependent critical state model has recently η∗ = q/(p + pb ) and pb is a bonding strength parame-
developed for MHBS (Shen et al. 2016). It has been ter that quantifies the degree of hydrate cementing at
the inter-granular contacts.
477
defined as ψ(Sh ) = e − ec (Sh ), e and ec (Sh ) are the
current void ratio and void ratio at the critical state,
respectively for a given confining pressure and Sh , d0 ,
m and mb are three model parameters.
According to the studies of cemented sand (Abdulla
and Kiousis, 1997), the initial bonding strength pb0 of
MHBS is assumed to be the following power function
of Sh :
where α and β are two positive model parameters.

Considering the degradation of hydrate cementing
Figure 1. Yield surface for MHBS in the p − q space.
with respect to the deformation (as shown in Figure
1), a simple linear relation between the logarithmic
Based on the theory of plasticity, the incremen- of degradation rate of the bonding strength and the
tal stress-strain relationships in the p − q space are increment in the plastic shear strain is assumed as
expressed as follows: follows:
where kd is a positive model parameter.
2.2 Constitutive relations considering hydrate

dissociation
where K and G are the elastic bulk and shear moduli
of MHBS, respectively, dεev and dεeq are elastic volu- Klar et al. (2010) suggested that in the natural envi-
ronment methane hydrate is often formed in the sand
metric and shear strain increments, respectively, Kp is
deposit after it has been consolidated to a given in-situ
the plastic hardening modulus, and D is the dilatancy
p p p p stresses (p0 , q0 ). They further postulated that these in-
defined as the ratio dεv dεq (dεv and dεq are plastic
situ stresses should be considered in the stress-strain
volumetric and shear strain increments, respectively).
formulations of MHBS. The formulations in the p − q
According to the resonant column test results for
stress space can be expressed as follows:
MHBS reported in Clayton et al. (2010), the elastic
shear modulus of MHBS is assumed linearly related
to hydrate saturation Sh , expressed as follows
In the derivation of stress, Klar et al. (2013) sug-

where G0 and ξ are two positive model parameters, e gested that the thermo-mechanical effect should be
is void ratio and pa is the atmospheric pressure. taken into account because of the possible temper-
Based on elasticity theory, the elastic bulk modulus ature change involved in hydrate dissociation. It is
K is equal to assumed that temperature change will only result in
elastic volume change, and thus only mean effective
stress is influenced by temperature change. This is
approximately true for MHBS because experimental
observations (Graham et al. 2004) showed that no sig-
where ν is the Poisson’s ratio, which is assumed to be nificant effects of temperature were encountered in
independent of density, pressure and hydrate saturation either isotropic compression or triaxial shear for sand.
in the model. Under this assumption, the incremental form of Eqs.
It is evident that D of MHBS depends on both pres- (9) and (10) can be written as
sure and density (Hyodo et al. 2013). The following
state-dependent dilatancy equation modified from Li
and Dafalias (2000) is proposed:
where T is temperature. It can be seen that the change

where M (Sh ) is the critical stress ratio for MHBS for in stress can be caused by change in stiffness, strain
a given Sh ; ψ(Sh ) is the state parameter for MHBS, and temperature.
478
The first term on the right hand side of Eq. (11) The following expression is obtained by substituting
shows the change in mean effective stress resulted Eqs. (13), (17) and (18) into Eq. (11):
from the change in elastic bulk modulus, which could
be caused by hydrate dissociation. As the elastic bulk
modulus is a function of Sh , this term can be further
expanded as follows:
The second term on the right hand side of Eq. Similarly, Eq. (12) can be rewritten as
(11) is the conventional incremental stress-strain rela-
tionship. Combining Eq. (2), this term can be further
expanded as follows:
It can be seen that not only the effective stress incre- Eqs. (19) and (20) show that the deformation of MHBS
ments (dp , dq) but also the internal bonding strength can be caused by three factors: effective stress, hydrate
increment dpb are involved in the calculation of plastic saturation and temperature. During hydrate dissocia-
volumetric strain. According to Eqs. (7) and (8), pb is a tion, hydrate saturation decreases, coupling with the
function of hydrate saturation and plastic shear strain. increase in temperature and effective stress for thermal
Considering hydrate dissociation, by combining Eqs. recovery and depressurisation methods, respectively.
(7) and (8), the incremental form of bonding strength Therefore, the above two equations give the extended
pb can be rewritten as constitutive relations of MHBS considering hydrate
dissociation. When the changes of the three influ-
encing factors have been known, the deformation of
MHBS during hydrate dissociation can be predicted.
Substituting Eq. (15) into Eq. (3) yields

3 NUMERICAL PREDICTION
The hydrate dissociation test results reported in Hyodo

et al. (2013) were used to verify the extended consti-
tutive relations of MHBS. In Hyodo’s experiments,
Combining Eqs. (16) and (2), Eq. (14) can be MHBS were synthesised by injecting pressurised
rewritten as methane gas into unsaturated sand specimens. Then,
consolidated drained triaxial tests were conducted on
three MHBS specimens. All specimens were first
isotropically compressed to an effective confining
pressure of 5 MPa. After consolidation three differ-
ent shear stresses were applied as illustrated in Figure
2. In Case 1 no shear stress was applied. In Cases
The third term on the right hand side of Eq. (11)rep- 2 and 3, initial shear stress of 8 and 12 MPa were
resents the influence of thermal effect on stress change. applied under drained condition, respectively. After
The elastic volume change of MHBS under thermal shearing to different stress levels, hydrate dissocia-
effect is assumed because of the expansion of sand tion was commenced under drained condition using
and hydrate grains with the increase of temperature. thermal recovery method. The temperature was grad-
This term can be given by ually increased from 5◦ to 20◦ , as shown in Figure 3.
The axial and volumetric strains of the specimens were
measured. The released methane gas was collected to
calculate the change of hydrate saturation with elapsed
time as shown in Figure 4.
where βs and βh are thermal expansion coefficients of When using thermal recovery method to extract
sand and hydrate, respectively. methane gas from MHBS, both the effective stress
479
Table 1. Model parameters.
Parameter Symbol Value
Elastic moduli G0 300

ν 0.05
ξ 1200
Critical state M 1.18
e 1.116
λc 0.129
A 0.35
B 1.4
a 0.6
b 2.0
Hardening rule h1 1.98
h2 2.19
Figure 2. Loading condition for three different cases (test n 0.8
data from Hyodo et al. 2013). Bonding α(MPa) 0.7
β 1.6
and temperature are under control. Since the effec- Debonding kd 8
Dilatancy d0 1.6
tive stress and temperature has been determined, the
m 1.3
‘measured’ change of hydrate saturation during tests mb 0.04
was treated as an input in our procedures to calculate Temperature βs (K −1 ) 1.77 × 10−6
the deformation of the specimens. Table 1 summarises βh (K −1 ) 4.6 × 10−4
the model parameters used in the numerical predic-
tion of hydrate dissociation tests. Shen et al. (2016)
presented the procedures to calibrate the first nineteen
model parameters based on the triaxial compression
test results reported in Hyodo et al. (2013). For the
last two thermal expansion coefficients (βs and βh ),
the values reported in the coupled numerical analysis
for MHBS by Klar et al. (2013) were adopted.
For the sake of simplicity, the following piecewise
linear function was used to represent the variation of
temperature with time during the hydrate dissociation
tests as shown in Figure 3:
Figure 3. Variation of temperature with elapsed time (test

data from Hyodo et al. 2013).
where t is time in hours.
Likewise, as shown in Figure 4, the measured
change of hydrate saturation with elasped time was
also best-fitted by the following piecewise linear
function:
An alternative approach to predict the dissociation

rate of methane hydrate is given by the following Kim-
Bishnoi equation (Kim et al. 1987):
Figure 4. Variation of hydrate saturation with elapsed time
(test data from Hyodo et al. 2013).
moles of hydrate in the initial state, PF is the aver-

age pore pressure and Peq is an equilibrium pressure at
where DH is the dissociation parameter, NH is the temperatre T , defined as follows: Peq = exp (39.13 −
moles of hydrate in a given volume V , NH 0 is the −8533.8/T ).
480
used as an input in the calculation of the deforma-
tion caused by hydrate dissociation. Figure 5 shows
the predicted and measured results of hydrate dissoci-
ation tests. Through comparisons it can be found that
the predictions can well describe the trend of the defor-
mation process for all three cases. In Case 1 (no shear
stress) both the measured and predicted deformations
of MHBS are negligible, implying little disturbance
on the sand skeleton for MHBS due to hydrate dis-
sociation. In Case 2, the deformation of MHBS is
negligible at the beginning of hydrate dissociation.
As the hydrate saturation decreases below a threshold
value, the deformation increases substantially. Such
deformation characteristics can be well captured by
the predicted results. In Case 3 (the highest applied
shear stress) shear failure was observed during hydrate
dissociation. It should be noted that the strength of
MHBS is controlled by the hydrate saturation. When
the hydrate saturation reduces to a certain value, with
which the strength of MHBS is smaller than the applied
shear stress, the specimen will fail.
4 CONCLUSIONS
In this paper, a state-dependent critical state model

for MHBS developed by Shen et al. (2016) was
extended to predict the effects of hydrate dissociation
on mechanical behaviour of MHBS. In the formula-
tion, the volume change of MHBS is derived as a
combined action of the change in stress, hydrate satura-
Figure 5. Predicted and measured variation of axial and vol- tion and temperature. However, the shear strain is only
umetric strains with elapsed time (test data from Hyodo et al. affected by the change in stress and hydrate saturation.
2013).
The extended constitutive relations were used to pre-
The dissociation rate expressed by Eq. (23) can be dict the hydrate dissociation tests reported in Hyodo
written in terms of hydrate saturation as follows: et al. (2013). The model parameters were calibrated
based on the triaxial compression tests. It is demon-
strated that despite some discrepancies, the predictions
can capture the key deformation features of MHBS
during hydrate dissociation. In particular, shear fail-
ure of MHBS can be predicted under certain high shear
stress level.
According to Eq. (24), when thermal recovery
method is used to induce hydrate dissociation, temper-
ature is the unique variable controlling the dissociation
rate. Thus, the change of hydrate saturation induced ACKNOWLEDGEMENTS
by temperature recovery method can be estimated by
combining Eqs. (21) and (24). The suggested value of The research is financially supported by GRF
DH for methane hydrate is 5.85 × 109 . The estimated 617213 and M-HKUST603/13 provided by Research
change of hydrate saturation is shown in Figure 4. It Grants Council (RGC) of HKSAR and FP204 by
can be seen that the estimated rate of hydrate satura- HKUST. Grant B13024 by the 111 Project of China,
tion change using the suggested value is substantially grants 51578213, 51308190 and 41530637 by the
faster than the measured results. It is found that the National Natural Science Foundation of China, grants
estimated change of hydrate saturation with time is 2015B06014 and 2015B25914 by the Fundamental
consistent with the measurements by decreasing DH Research Funds for the Central Universities of China
two orders of magnitude, i.e., 5.85 × 107 . Several rea- are also acknowledged.
sons may contribute to the discrepancy, such as the
inhibiting effect of soil grains on the hydrate dissocia-
REFERENCES
tion, as well as the measurement delay due to the low
discharge capacity of the drainage line. Abdulla, A.A. & Kiousis, P.D. 1997. Behavior of cemented
To avoid the above uncertainties, in this study the sands-I. Testing. International Journal for Numerical &
best-fit lines of measurements using Eq. (22) were Analytical Methods in Geomechanics 21(8): 533–47.
481
Clayton, C.R.I. Priest, J.A. & Rees, E.V.L. 2010. The effects Masui, A. Haneda, H. Ogata, Y. & Aoki, K. 2005. Effects of
of hydrate cement on the stiffness of some sands. Geotech- methane hydrate formation on shear strength of synthetic
nique 60(6): 435–45. methane hydrate sediments. In: Proceedings of the 15th
Graham, J. Alfaro, M. & Ferris, J. 2004. Compression a International Offshore and Polar Engineering Conference
strength of dense sand at high pressures and elevated p. 364–69.
temperatures. Canadian Geotechnical Journal 41(6): Mienert, J. Vanneste, M. Bünz, S. Andreassen, K. Haflidason,
1206–1212. H. & Sejrup H.P. 2005. Ocean warming and gas hydrate
Hyodo, M. Yoneda, J. Yoshimoto, N. & Nakata, Y. 2013. stability on the mid-Norwegian margin at the Storegga
Mechanical and dissociation properties of methane Slide. Marine and Petroleum Geology 22(1): 233–44.
hydrate-bearing sand in deep seabed. Soils and Founda- Miyazaki, K. Masui, A. Sakamoto, Y. Aoki, K. Tenma, N.
tions 53(2): 299–314. & Yamaguchi, T. 2011. Triaxial compressive properties of
Kim, H.C. Bishnoi, P.R. Heidemann, R.A. & Rizvi, S.S.H. artificial methane-hydrate-bearing sediment. Journal of
1987. Kinetics of methane hydrate decomposition. Chem- Geophysical Research 116(B6).
ical Engineering Science 42(7):1645–53. Nixon, M.F. & Grozic, J.L.H. 2007. Submarine slope failure
Klar, A. Soga, K. & Ng, M.Y.A. 2010. Coupled deformation– due to gas hydrate dissociation: a preliminary quantifica-
flow analysis for methane hydrate extraction. Geotech- tion. Canadian Geotechnical Journal 44(3): 314–25.
nique 60(10):765–76. Shen, J. Chiu, C.F. Ng, C.W.W. Lei, G.H. & Xu, J. 2016. A
Klar, A. Uchida, S. Soga, K. & Yamamoto, K. 2013. Explic- state-dependent critical state model for methane hydrate-
itly coupled thermal flow mechanical formulation for bearing sand. Computers and Geotechnics 75:1–11.
gas-hydrate sediments. Spe Journal 18(2):196–206. Uchida, S. Soga, K. & Yamamoto, K. 2012. Critical state soil
Kvenvolden, K.A. 1999. Potential effects of gas hydrate on constitutive model for methane hydrate soil. Journal of
human welfare. Proceedings of the National Academy of Geophysical Research 117(B3).
Sciences of the United States of America 96(7): 3420–6.
Li, X.S. & Dafalias, Y.F. 2000. Dilatancy for cohesionless
soils. Geotechnique 50(4): 449–60.
482
Composite model to reproduce the mechanical behaviour

of Methane Hydrate Bearing Sediments
M. De La Fuente
Geology & Geophysics Department, National Oceanography Centre (University of Southampton
and Natural Environment Research Council), Southampton, UK
J. Vaunat
Department of Civil and Environmental Engineering, Universitat Politècnica de Catalunya, Barcelona, Spain
H. Marín-Moreno
National Oceanography Centre, European Way, Southampton, UK
ABSTRACT: This paper describes the fundamental hypothesis and the formulation of a new constitutive model
to cope with the mechanical behaviour of methane hydrate-bearing sediments. The model is based on a composite
approach that defines the mechanical response of mineral skeleton and hydrate bond network independently.
The proposed model represents the transitional behaviour of MHBS by coupling two independent constitutive
models under certain strain compatibility conditions and mass and energy considerations. CASM elasto-plastic
critical state model has been chosen to describe the matrix response and the bonding structure follows a damage
model. In order to overcome the technical challenges of natural gas production and the geotechnical problems
related with its complex rheological response, the constitutive model simulates the global stress-strain response
of the composite medium and allows to assess methane hydrate dissociation on the basis of the stress state and
temperature prevailing locally within the hydrate component.
1 INTRODUCTION presence of hydrate growing in the pores as a struc-

turing element (e.g., Soga et al. (2006) and Waite et al.
Gas hydrates are ice-like crystalline compounds com- (2009)). However, it has been shown that this effect
posed of low molecular weight gases trapped in a disappears when the sediment is excessively com-
lattice of hydrogen-bonded water molecules (Sloan pressed or the bonds are destroyed due to shearing
1998). Current estimates indicate that gas hydrates (Uchida et al., 2012). Furthermore, recent experimen-
trap more than half of the world’s total organic car- tal results show that the mechanical behavior of MHBS
bon mass and twice as much as all other fossil fuels is also strongly controlled by sediment type, stress
combined (Kvenvolden, 1993), turning out to be an level, hydrate morphology, and sediment history, as
attractive potential future energy resource (Collett, well as highly dependent on thermo, hydraulic and
2002; Dallimore & Collett, 2005; Ruppel, 2007). geo-chemical coupled interactions (e.g., Clayton et al.,
Methane hydrate-bearing sediments (MHBS) are 2010; Priest et al., 2009; Sanchez et al., 2014).
natural deposits that host methane hydrate inside its This paper describes a new constitutive model,
pore space. Methane hydrates are the most common inspired in frozen soils behavior, specially developed
occurring gas hydrate in nature that form under moder- to analyze sandy methane hydrate reservoirs as these
ately high pressure and low temperature conditions. Its may be the first target for production in the near
naturally occurrence is primarily found in permafrost future (Boswell & Collet, 2006). In this work and from
regions onshore and in ocean-bottom sediments at a soil mechanics perspective, MHBS will be under-
many continental margins and slopes (Collett, 2002; stood as a complex multi-phase bonded geomaterial
Kvenvolden, 1999). characterized by a highly non-linear, irreversible and
One of the main challenges in gas hydrate prospect- path-dependent response to applied loads. The model
ing and deposit characterization is obtaining the is based on a composite approach that allows defin-
geomechanical properties and controlling the stabil- ing the mechanical response of the mineral skeleton
ity of the well during production (Moridis, Collett, and hydrate bond network independently and includes
Pooladi-Darvish, et al., 2011). MHBS tend to exhibit the effect of hydrate morphology on the composite
greater stiffness, strength and dilatancy due to the response.
483
where:
Equation 5 gives the restriction on volumetric

strains and illustrates that, in a composite medium,
the external volumetric strain (εext vol ) is not equal to
Figure 1. Schematic arrangement considered in the MHBS the change in porosity (εint vol ) because of the bond
composite medium model and graphical definition of the
compressibility (Cb εbvol ).
volumetric variables.
2.1.2 Strain partition
2 METHANE HYDRATE BEARING SEDIMENT
Compatibility between external and local strains is
COMPOSITE CONSTITUTIVE MODEL
derived from the conservation of volume defined by
Pinyol et al. (2007). They propose a coupling coef-
2.1 Conceptual framework
ficient χb that relates the volumetric response of the
In this model, MHBS are considered as a compos- bond with the composite volumetric response and it is
ite medium made of a sandy matrix interlocked by a defined as:
homogeneously distributed network of hydrate bonds.
The soil matrix is associated to the characteristics of
the original pedologic formation, and its behavior is
represented by a typical elastoplastic soil model (Yu,
1998). The bonding hydrate network is modeled by
an elastic, suction and temperature dependent damage
law, typical of quasi-brittle materials such as ice (Carol
et al. 2001). Following this strain partition the external strain
increment applied to the MHBS can be rewritten from
2.1.1 Topologic considerations the equation 5 as:
In a reference volume, the following volumetric vari-
ables may be defined as (Fig. 1).
Vt Total volume
Vsm Solid volume of matrix skeleton
Vb Solid volume of bond material
Vm Volume of macropores (pore space between matrix
grains)
Vv Volume of available porosity (open voids available
for fluid transfer)
where the total volume is given by: When χb = 0, bonds are not considered in the load-
ing framework and the behaviour of the composite
material recovers that of a sandy soil. However, when
with the volume of solid phases expressed as: χb increases, bond behaviour starts to dominate the
composite mechanical response, and for high χb , it
behaves like a quasi-brittle material.
and the volume of macropores as:

2.2 Matrix behaviour
The hydrate resource pyramid (Boswell & Collet,
2006) clearly illustrates that marine sand reservoirs
Conventionally, the application of an external com- are the major target for long-term development of gas
pressive volumetric strain creates a decrease on the hydrates as a future energy supply. Consequently, an
available pore space and on bond volume, whose extension of the CASM (Clay and Sand model; Yu,
changes are given by: 1998) for unsaturated soils (González, 2011) has been
choose as the constitutive model to describe the matrix
response.
484
preconsolidation pressure which acts as the hardening-
softening parameter controlling the size of the yield
surface.
2.2.2 Elastic behaviour

The behaviour inside the yield surface is assumed to
be isotropic and elastic. The elastic stress-strain rela-
tionship is described based on the Bishop effective
stress concept (Bishop, 1959) where the bulk modu-
lus K incorporates the effect of suction on the global
volumetric elastic strain as:
Figure 2. CASM yield surfaces normalized by preconsol-

idation pressure. For a fixed spacing ratio r variations in n
value can display local peaks before approaching the critical
state condition. After González (2011).
where e is the macroscopic void ratio, κ is the slope
of the isotropic swelling line (in υ − ln p space), Sh
CASM model is a unified critical state model based is the hydrate saturation and dp∗ represents the incre-
on the critical state theory and is formulated in terms of ment in pressure due to mean stress (p) and suction (s)
the state parameter concept (ξ) introduced by Been & variations.
Jefferies (1985). Its election as a matrix constitutive
model is due to its simplicity and flexibility in describ- 2.2.3 Hardening parameter and flow rule
ing the shape of the yield surface as well as its proven On the basis that the yield surface located at zero suc-
ability to predict the mechanical behaviour of reconsti- tion defines the hardening variable p0 , an isotropic
tuted clays and specially sands. A total of seven model volumetric hardening-softening law controlled by the
parameters are needed to define CASM, five of which plastic volumetric strains is assumed and given by:
are the same as those in the standard Cam-Clay model,
the most popular model used to represent the mechan-
ical behaviour of MHBS (e.g., Uchida 2012). The two
extra model parameters are denoted by n and r, where
n is used to specify the shape of the yield surface and where λ∗0 = λ0 /(1 + e) and κ∗ = κ/(1 + e) are the mod-
r is a spacing ratio introduced to control the point ified slope of the saturated virgin consolidation line
of intersection of the critical state line and the yield and the unloading-reloading line, respectively.
surface. The CASM plastic potential is integrated from
Figure 2 shows that the intersection point between the stress-dilatancy relationship proposed by Rowe,
the critical state line and the yield surface in the CASM (1962) and it is defined as:
model does not necessarily occur at the maximum
deviatoric stress (as happens in the original and modi-
fied Cam-Clay models). This reproduces an important
feature observed in sands in which the deviatoric stress
often reaches a local peak before approaching the where ϕ is a size parameter controlling the size of
critical state condition (González, 2011). the plastic potential which passes through the current
stress state.
2.2.1 Yield function
The original CASM model is extended to establish 2.3 Hydrate behaviour
a suction dependency of the preconsolidation pres- MHBS are highly compacted-stable sediments that
sure and the hardening law following the proposal by behave as a bonded soil with hydrate acting as bond-
Alonso et al. (1990) for the Barcelona Basic Model. ing agent (e.g. Masui et al., 2005, Yun et al., 2007,
In the triaxial stress space the CASM yield function is Miyazaki, K. et al., 2008). However, once the hydrate
defined by equation 14: dissociates due to a decrease in pressure and/or
increase in temperature, or is damaged by a mechan-
ical strain, the bonded structure disappears and the
soil may behave as unconsolidated material (Brugada
et al., 2010). Accordingly, our composite model sim-
where q is the deviatoric stress, p is the mean net stress, ulates the hydrate bond behaviour by a scalar damage
ps is a tensile strength due to suction, M is the slope model proposed by Carol et al. (2001) that allows to
of the critical state line in the q-p space, and p0 is the represent the medium destructuralization.
485
The damage model is based on thermodynamical
principles that prevent any energy dissipation dur-
ing load cycles. Besides, it assumes a linear elastic
response for the bond in an undamaged state, with the
shear and bulk moduli decreasing progressively during
loading as micro-cracks develop inside the material.
The damage model is defined as:
Figure 3. Pore-scale distribution of methane hydrate: a

Pore-Filling, b Load-bearing, c Cementation. After (Brugada
where Kb0 and Gb0 are the bulk and shear moduli of et al., 2010).
the hydrate bond in an intact state, and Kb and Gb the
same moduli for a damage level given by parameter D
which is equal to the ratio of fissured bond area over
whole bond area.
As proposed by Carol et al. (2001), an alternative
damage variable is used:
where dbσ is given by:
This variable can be simply related to the degra-

dation of bond Young modulus by the following
expression: Equation (28) shows the constitutive law of the com-
posite medium, where χb is assumed to depend only
on the bond damage variable L by:
where E0 and E are the undamaged and damaged bond

Young’s moduli, respectively. The evolution of L is also
linked in the model with the increments of energy where χb0 is the value of χb in the intact state (L = 0).
stored in the bonds per unit of volume through sev- This dependency indicates that part of the load carried
eral damage evolution laws and allows to update the by the bond in a given state is transferred to the soil
elastic modules for a given damage: matrix when bonds degrade.
2.4.1 Effect of hydrate morphology in composite

behaviour
Recent works (e.g, Waite et al., 2009;Yun et al., 2007),
provide evidences that the type of sediment, confining
2.4 Coupling stress applied, hydrate saturation and hydrate morphol-
ogy govern the load-deformation response of MHBS.
Since hydrate bonds and soil matrix are endowed with Specifically, the effect of hydrate habit on seismic
their own independent behaviour, local strains and wave velocity, attenuation and small strain stiffness
stresses have to be defined for each component. Local has been reported extensively (e.g, Waite et al., 2009;
strains are compatible with the external strains by the Best et al., 2013). However the understanding on how
strain partition presented in equation (9). hydrate habit influences the intermediate and large
The stress partition is obtained by stating that the strain behavior, and the failure condition is still limited.
virtual work done by the medium under any compat- Three main hydrate habits or growth patterns are
ible external strain increment is equal to the sum of commonly described in the literature (Fig. 3): pore
the works performed by the matrix and the bonds. filling, load-bearing and cementation.
Defining σ ext , σ M and σ b as the stresses acting on the Pore-filling hydrate (Fig. 3a) nucleates and grows
medium, matrix and bonds, respectively, and assum- freely in the pore space, without bridging particles
ing that (i) the incremental external strain dεext vol is together and affecting essentially the pore fluid bulk
associated to the matrix stress σ M through the con- stiffness and fluid conduction properties (Ecker et al.,
stitutive law of the soil matrix and (ii) the incremental 1998; Helgerud et al., 1999). This habit likely evolves
strain of the bond dεbvol is associated to the bond stress to load-bearing hydrate (Fig. 3b) by pore invasion
σ b through the constitutive law of the hydrates, the leading to hydrate saturation (Sh ) values that exceeds
following equations are obtained: 25%–40% (Berge, 1999; Yun et al., 2005, 2007). This
hydrate pattern bridges nearby grains and contributes
to the mechanical stability of the granular skeleton
486
stress ph0 at hydrate formation time. This reference
value can be obtained from the equilibrium of the
chemical potentials between liquid and hydrate phases
existing in the sediment pore space by the Clausius–
Clapeyron equation (equation 30). Following the work
for frozen soils presented by Nishimura et al. (2009),
the thermodynamic equilibrium that needs to be sat-
isfied by the hydrate pressure (ph0 , which is assumed
equal to ice pressure due to its chemical similarities),
methane-saturated water pressure (pw ) and temprature
(T ) is given by:
where l is the specific latent heat of hydrate fusion and

ρ (=1/v) is the mass density of each phase.
b) Pore-filling hydrate is considered as a densifier
compound, not involved on the load supporting frame-
work. Its effect over the composite strength is only
considered by the suction generated on hydrate-water
Figure 4. Stress–strain results from drained triaxial tests on interface (ph0 − pw ).
Toyoura sand observed by Masui et al. (2005). c) An approximately linear evolution of the hard-
ening parameter value (p∗0 ) with hydrate saturation for
load-bearing and cementing hydrate habits (Fig. 5).
3 CONSTITUTIVE MODEL FLOW CHART
Figure 6 outlines the flow chart followed to compute

the stress increment from the strain increment applied
to the composite model. For the sake of clarity, the ideal
case of infinitesimal increments is considered. When
dealing with discrete increments, the tangent rela-
tionships indicated in the flow chart dσijM = Dijkl M
dεM
kl
and dσij = Dijkl dεkl should be replaced by convenient
b b b
iterative procedures.
The first step consists in calculating the increment
of hydrate strain dεbkl from the increment of external
strain dεext
vol , and the current value of χb and bond con-
Figure 5. Hardening effect of each hydrate habit on the centration Cb . The increment of hydrate stress dσijb is
matrix preconsolidation pressure (p0 ). obtained by integration of the damage model and arises
as the product of the damaged hydrate secant elastic
by becoming part of the load-bearing framework. stiffness matrix Db with dεbkl . The total hydrate stress
Finally, cementing hydrates (Fig. 3c) nucleate at soil is then updated by adding the stress increment dσijb to
grain contacts and act as a bonding agent that signif- the previous value.
icantly increases the sediment shear and bulk moduli Using the mean value of the pressure and the
(Helgerud, M. B., J. Dvorkin, A. Nur, A. Sakai 1999). temperature prevailing locally within the hydrate com-
Experimental results obtained from drained triax- ponent (ph , T), the condition for possible methane
ial compression tests performed on synthetic hydrate hydrate dissociation can then be assessed.
sands (Fig. 4) show how the hydrate habit exerts a However, the bond-damage model is restricted to
strong control on the macroscale mechanical proper- the pore-filling hydrate habit until the hydrate satura-
ties of the sediment. tion reaches values close to 35% and hydrate passes to
Figure 4 clearly shows that cementing samples form part of the load supporting framework.
exhibit a greater enhancement in stiffness, strength and Secondly, the change in matrix stress is assessed by
dilatancy than the load-bearing or pore filling cases. applying the elastoplastic model CASM to the matrix
To reproduce the effect caused by the different strain increment dσijM = DijklM
dεMkl . The effect of suc-
hydrate habits on the composite mechanical proper- tion generated by water-hydrate interfaces is taken into
ties, the model assumes three initial hypotheses. account similarly to the proposal by González (2011).
a) The hydrate is considered formed once the matrix Finally, the total stresses of the composite medium
is stressed, taking as a reference zero state the initial (equation 28) are obtained from the updated hydrate
487
effectives stress acting over the matrix skeleton by
introducing the role of the suction generated at the
hydrate-water interface.
The framework developed in this paper requires
further validation by comparison with experimental
and field data. In this perspective, the model is now
being implemented into the 3D finite element code
Code_Bright with the objective to model field prob-
lems involving MHBS, including the assessment of
the stability of offshore infrastructures, cables and
pipelines sitting on these sediments.
REFERENCES
Alonso, E. E., Gens, A. & Josa, A., 1990. A constitutive model
for partially saturated soils. Geotechnique 40, No. 3,
405–430.
Been, K. & Jefferies, M.G., 1985. A state parameter for sands.
Géotechnique, 35(2), 99–112.
Berge, L. I., Jacobsen, K. A., and Solstad, A., 1999. Measured
acoustic wave velocities of R11 (CCl3 F) hydrate samples
with and without sand as a function of hydrate con-
Figure 6. Flow chart for numerical algorithm. centration. Journal of Geophysical Research-Solid Earth,
104(B7), 15415–15424, July (10).
Best, A.I., Priest, J. A., Clayton, C.R.I., Rees, E.V.L., 2013.
and matrix stress. The constitutive law defined in this The effect of methane hydrate morphology and water
way is able to recover a pure soil behaviour for null saturation on seismic wave attenuation in sand under shal-
hydrate concentration (Cb = 0) or when total damage low sub-seafloor conditions. Earth and Planetary Science
of the bonding network occurs (χb = 0). Letters, 368, pp. 78–87.
Bishop,A.W., 1959. The Principle of Effective Stress. Teknisk
Ukeblad 39 (October): 859–863.
Boswell, R. and Collett T.S., 2006. The gas hydrates resource
4 SUMMARY pyramid: Fire inthe ice: Methane hydrate newsletter, Fall
issue, pp. 5–7.
The experimental study of MHBS has been hampered Brugada, J., Cheng,YP., Soga, K., Santamarina, J., 2010. Dis-
by sampling difficulties related with depressurization crete element modelling of geomechanical behaviour of
and thermal changes during core extraction and the methane hydrate soils with pore-filling hydrate distribu-
low solubility of methane in water in laboratory testing tion. Granular Matter, 12(5), pp. 517–525.
Carol, I., Rizzi, E. & Willam, K., 2001. On the formulation of
(Sanchez et al., 2014). These restrictions and the strong
anisotropic elastic degradation. I. Theory based on a pseu-
scientific and engineering interest regarding this geo- dologarithmic damage tensor rate. Int. J. Solids Struct. 38,
material have motivated the development of numerical No. 4, 91–518.
models to advance our current understanding of their Clayton, C.R.I., Priest, J.A., Rees, E.V., 2010. The effects of
behaviour. hydrate cement on the stiffness of some sands. Géotech-
In this paper a new composite constitutive model to nique, 60(6), pp. 435–445.
reproduce the evolving nature of MHBS mechanical Collett, T.S., 2002. Energy resource potential of natural gas
properties has been presented. The model results from hydrates, AAPG Bull., 86(11), 1971–1992.
the integration of two constitutive models, an elasto- Dallimore, S. R., and Collett, T.S., 2005. Summary and
implications of the Mallik 2002 gas hydrate production
plastic model for the matrix and a damage model for
research well program, in Scientific Results From the
the bond material, into a common composite frame- Mallik 2002 Gas Hydrate Production Research Well Pro-
work. Their integration follows imposed internal strain gram, Mackenzie Delta, Northwest Territories, Canada,
compatibility conditions, and the stress partitioning vol. 585. Geological Survey of Canada Bulletin 585.
between the two constituent materials is achieved by González, N., 2011. Development of a family of constitutive
means of the virtual work principle. models for geotechnical applications. (May), pp. 47–84.
One specificity of the model is to be based on a crit- Helgerud, M. B., J. Dvorkin, A. Nur, A. Sakai, and T.C.,
ical state constitutive law specially developed to cope 1999. Elastic-wave velocity in marine sediments with
with the mechanical behavior both clay and sands, gas hydrates: Effective medium modeling, Geophys. Res.
Lett., 26, 2021–2024.
instead of the Cam-Clay framework.This will allow for
Kvenvolden, K.A., 1993. Gas hydrates—geological perspec-
a more precise modelling of sandy reservoirs contain- tive and global change. Reviews of Geophysics, 31(2),
ing hydrate, which are the major target for near-future p. 173. Available at: http://doi.wiley.com/10.1029/93RG
development of gas hydrates as an energy supply. 00268.
Besides, this model considers the non-mechanical Kvenvolden, K.A., 1999. Potential effects of gas hydrate on
effect of pore-filling hydrate in the estimation of the human welfare. Proceedings of the National Academy
488
of Sciences of the United States of America, 96(7), Ruppel, C., 2007. Tapping methane hydrates for unconven-
pp. 3420–3426. tional natural gas. , Elements, 3, 193–199, doi:10.2113/
Masui, A., H. Haneda,Y. Ogata, and K.A., 2005. The effect of gselements. 3.3.193., pp. 1–19.
saturation degree of methane hydrate on the shear strength Sanchez, M.J., Gai, X., Shastri, A., Santamarina, J.C., 2014.
of synthetic methane hydrate sediments, Proceedings Coupled THCM Modeling of Gas Hydrate Bearing Sedi-
of the 5th International Conference on Gas Hydrates, ments. American Geophysical Union, Fall Meeting 2014,
Trondheim, Norway, vol. 2037, pp. 657–663. abstract #B11B-0020.
Miyazaki, K., A. Masui, H. Haneda, Y. Ogata, K. Aoki, and Sloan, E.D., 1998. Clathrate Hydrates of Natural Gases
T.Y., 2008. Variable-compliance-type constitutive model 2nd ed., New York.
for methane hydrate bearing sediment, paper presented at Soga, K., S. L. Lee, M.Y.A. Ng, andA.K., 2006. Characterisa-
6th International Conference on Gas Hydrates, Univ. B. tion and engineering properties of methane hydrate soils,
C., Vancouver, B. C., Canada. in Characterization and Engineering Properties of Natu-
Moridis G.J, Collett T.S, Pooladi-Darvish M, Hancock ral Soils, vol. 4, edited by K. Soga et al., pp. 2591–2642,
S, Santamarina J, Boswell R, Kneafsey T, Rutqvist J, Taylor and Francis, London, U.K.
Kovalsky M, Reagan M.T, Sloan E.D, Sum A.K, K.C., Uchida, S., Soga, K. & Yamamoto, K., 2012. Critical state
2011. Challenges, Uncertainties and Issues Facing Gas soil constitutive model for methane hydrate soil. Journal
Production From Gas Hydrate Deposits. SPE Res Eval of Geophysical Research: Solid Earth, 117(B3).
Eng 14(1):76–112, (March 2010). Waite, W.F. et al., 2009. Physical Properties of Hydrate-
Nishimura, S., Gens, A., Jardine, R., J. Olivella, S., 2009. Bearing Sediments. Change, 47(2008), pp. 1–38.
THM-coupled finite element analysis of frozen soil: Yu, H.S., 1998. CASM: a unified state parameter model for
formulation and application. Géotechnique, 59(3), clay and sand. International Journal for Numerical and
pp. 159–171. Analytical Methods in Geomechanics, 22(8), pp. 621–653.
Pinyol, N., Alonso, E.E. & Vaunat, J., 2007. A constitu- Yun, T. S., F. M. Francisca, J. C. Santamarina, and C.R., 2005.
tive model for soft clayey rocks that includes weathering Compressional and shear wave velocities in uncemented
effects. Géotechnique, 57(2), pp. 137–151. sediment containing gas hydrate, Geophys. Res. Lett., 32,
Priest, J.A., Rees, E.V.L., Clayton, C.R.I., 2009. Influence L10609.
of gas hydrate morphology on the seismic velocities of Yun, T.S., Santamarina, J.C. & Ruppel, C., 2007. Mechanical
sands, J. Geophys. Res., 114, B11205. properties of sand, silt, and clay containing tetrahydrofu-
Rowe, P.W., 1962. The stress-dilatancy relation for static equi- ran hydrate. Journal of Geophysical Research, 112(B4),
librium of an assembly of particles in contact. Proc. Roy. p. B04106.
Soc., 267, 500–527.
489
Shallow geothermal systems – geophysical support

for geothermal planning maps
R. Kirsch, A. Wolf, K. Baezner & C. Thomsen

LLUR Geological Survey of Schleswig-Holstein, Flintbek, Germany
H. Wiederhold
LIAG Leibniz Institute of Applied Geophysics, Hannover, Germany
ABSTRACT: Shallow geothermal energy using ground heat exchangers is an effective technique to produce
heat from (renewable) electrical energy. The geological surveys are concerned with ground heat exchangers
mainly due to two reasons: a) efficiency of the geothermal system, b) groundwater protection: ground heat
exchanger can penetrate groundwater protective layers leading to pathways for potentially contaminated surface
waters into the aquifer. Underground information, especially on thermal conductivity, is required for the proper
design of the Ground Heat Exchangers (GHE). Therefore, plannig maps showing the effective thermal conductiv-
ity or the specific thermal power of the ground for defined GHE lengths are provided by the geological surveys.
These maps are based on drilling results. In areas with low coverage of drillings geophysical data can be used
to fill the data gaps. An approach to use geotechnical techniques for an assessment of thermal conductivities is
demonstrated and discussed for an area in Northern Schleswig-Holstein.
1 TECHNIQUES OF SHALLOW Main components of this heating system are a

GEOTHERMAL ENERGY USE heatpump and ground heat exchangers (boreholes con-
taining 1 or 2 U-shaped plastic tubes filled with heat
Shallow geothermal energy using ground heat transfer fluid (water and antifreeze), Fig. 1). To pro-
exchangers is an effective technique to convert elec- vide a good thermal contact from the ground to the
trical energy to heat, about 4 times more effective than plastic tubes and to avoid the formation of pathways
conventional power to heat techniques.
Figure 1. Left: base components of a shallow geothermal heating system, right: schematic view of a heat pump (A, B see
text).
493
for contaminated surface water to the groundwater the
borehole is grouted with special cement.
The heat transfer fluid is circulating through the
plastic tube in the borehole and through the heat
exchanger of the heat pump. In the heat exchanger
thermal energy is passing from the transfer fluid to the
internal cycle of the heatpump, therefore the temper-
ature of the transfer fluid is reduced by 3–5 K and can
fall below 0◦ C. When the cooled transfer fluid is flow-
ing again through the plastic tubes, thermal energy is
passing from the ground to the transfer fluid and the
temperature recovers (Bernier 2006).
Figure 2. Mean temperature of the heat transfer fluid

related to the thermal conductivity of the underground
2 DESIGN OF GROUND HEAT (length of deep ground heat exchanger 100 m, annual heat
EXCHANGERS – ENERGY EFFICIENCY load 16.200 MWh, borehole cementation with high thermal
AND GROUNDWATER PROTECTION conductivity).
Ground heat exchangers are designed to fulfill two

The key underground parameter for the design of
demands: energy efficiency of the heat pump and
ground heat exchangers is the thermal conductivity
groundwater protection (LLUR 2011).
(Dietlefsen et al. 2013). Figure 2 shows the mean tem-
The efficiency of the heat pump is quantified by the
perature of the transfer fluid related to the thermal
coefficient of performance (COP), defined by the ratio
conductivity of the underground (results are modelled
of the thermal power provided by the heat pump and the
using the computer program Earth Energy Designer).
electrical power required to drive the heat pump. Since
For the planning of ground heat exchanger maps show-
the production of electrical power is still an important
ing the thermal conductivity of the underground are
source of CO2 emissions, a COP of at least 4 is neces-
required.
sary to be competitive with a heating system based on
natural gas in the context of CO2 reduction. The COP
depends on the difference between the temperature of
the transfer fluid entering the heat exchanger of the 3 THERMAL AND ELECTRICAL
heat pump (e.g. 5◦ C) and the temperature of the heating CHARACTERISTICS OF UNCONSOLIDATED
system (e.g. 30◦ C). Therefore, to enable a high effi- SEDIMENTS
ciency of the heat pump the temperature of the transfer
fluid should be as high as possible at the entrance of the The thermal conductivity is defined by the heat flow
heat pump (see Fig. 1). However, the balance between through the underground material driven by the tem-
heat extraction and recovery of the temperature field in perature gradient. For unconsolidated sediments the
the surrounding of the ground heat exchangers during thermal conductivity is mainly governed by the poros-
the warm seasons must be considered when ground ity because the thermal conductivity of the pore fillings
heat exchangers are planned. air (0.05 W/mK) or water (0.5 W/mK) are low com-
For the protection of groundwater resources the for- pared with the thermal conductivity of the mineral
mation of pathways for contaminated surface water grains. Average values of thermal conductivites for
must be excluded when groundwater protecting lay- water saturated material are:
ers (till or clay) were penetrated by the drillings for sand: 2.4 W/mK
the ground heat exchangers. The borehole must be till: 2.4 W/mK
filled and sealed with cement after the plastic tubes clay: 1.5 W/mK.
are installed. However, the reliance of the cement
against freezing and thawing is not secured. Lab- The thermal conductivity of dry/unsaturated sand
oratory experiments showed that cracks leading to is low, about 0.8 W/mK. For till and clay we assume
enhanced hydraulic conductivity of the cement can that both materials contain bound water even above the
be formed after repeated freezing – thawing cycles. groundwater table and so the thermal conductivity of
Therefore, freezing of the cement must be avoided and unsaturated clayey material is higher than the thermal
the temperature of the heat transfer fluid should not be conductivity of unsaturated sand. So, thermal conduc-
lower than –3◦ C when leaving the heat pump. tivities of 1.7 W/mK for unsaturated till and 1.0 W/mK
So, properly designed ground heat exchangers must for unsaturated clay were postulated.
assure that the two temperature requirements of the The definition of the electrical conductivity of
transfer fluid are fulfilled: a) sufficient temperature at underground material is similar to the definition of
the entrance of the heatpump (A in Fig. 1) to enable thermal conductivity: by the electrical current density
a high COP of the heat pump, b) temperature should driven by an electrical potential gradient. Commonly,
not fall below –3◦ C when leaving the heat pump (B in the reciprocal of the electrical conductivity, the spe-
Fig. 1) to avoid freezing of the cementation. cific electrical resistivity in ohm x meter (m) is used
494
Figure 4. Detail of the geothermal planning map based on
drilling results showing the mean thermal conductivity of the
100 m depth range.
Figure 3. Airborne electromagnetic mapping system

SkyTEM.
for the characterisation of the underground material.

Typical values of specific electrical resistivities for
water saturated material are:
sand: 80–200 m, till: 40–60 m
clay: below 30 m.
Clayey layers with low thermal conductivities are
clearly separated from sand and till layers with high
thermal conductivities by low specific electrical resis-
tivities. Therefore, the specific electrical resistivity of Figure 5. Distribution of drillings used for the geothermal
the underground can be used for an assessment of the planning map.
thermal conductivity.
A number of geophysical techniques can be applied
for the mapping of the electrical resistivity struc- Netherlands, Germany and Denmark are covered by
ture of the underground. Conventional 1D and 2D airborne electromagnetic surveys.
geoelectrical measurements use electrodes to inject
a current into the ground and to measure the result-
ing electrical potential at the surface. Data inversion 4 GEOTHERMAL PLANNING MAPS
reveals the 1D or 2D resistivity-depth structure. Elec-
tromagnetic techniques make use of eddy currents Geothermal planning maps are launched by the Geo-
in the underground induced by time variable mag- logical Surveys to give a quick overview of the under-
netic fields to obtain the resistivity structure of the ground conditions as an aid for the design of ground
underground. No direct contact to the soil is neces- heat exchangers. Commonly these maps display the
sary, so electromagnetic can be used as an airborne specific thermal power (W/m) following the technical
technique (HEM, SkyTEM) for the mapping of large guideline VDI 4640 (2001) for different depth ranges,
areas (Fig. 3). Increasing areas of the lowlands in the e.g., down to 50 m or 100 m below ground.
495
Figure 6. Assessment of thermal conductivities based on
SkyTEM resistivity results: from top to bottom: location of
the flight line, resistivity distribution below the flight line,
resistivity-depth distribution for a selected location (7,500 m
along the flightline), allocation of thermal resistivities of the
resolved layers, calculated effective thermal conductivity for
this location for the depth range down to 100 m.
A new set of geothermal planning maps for

Schleswig-Holstein show the thermal conductivity for
different depth ranges because this ground property Figure 7. Detail of the geothermal planning map showing
is required for thermal simulation programs like, e.g., the effective thermal conductivity for the 50 m depth range.
Earth Energy Designer (Hellström 1991, Hellström & Top: location map with boreholes and SkyTEM sounding
Sanner 2000). This map is based on drilling results. points, middle: map based on drilling results alone, bot-
For each drilling, the weighted mean thermal conduc- tom: map including additional data points from the SkyTEM
tivity is calculated from the layer sequence and the survey.
above mentioned tabular values of thermal conductiv-
ities for the saturated and the unsaturated zone. The (Fig. 5). Especially in the northern and western part
saturated and the unsaturated thermal conductivities of Schleswig-Holstein data gaps are obvious.
are interpolated and, using the depth to groundwater
head, thermal conductivities for different depth ranges
are calculated and displayed in maps (Fig. 4). A total 5 CLOSING DATA GAPS WITH RESISTIVITY
of 28,742 drillings are used for these maps, of which TECHNIQUES
4,069 drillings are exceeding 100 m depth penetration.
Although a large number of drillings is used for As shown before, the specific electrical resistivity is
the maps, the distribution of drillings is irregular closely related to the thermal conductivity. For an area
496
at the North Sea coast in Northern Schleswig-Holstein is attributed, then the effective thermal conductivity
with low drilling density the data base is improved by of this site is calculated by the weighted mean of the
resistivity data from a SkyTEM survey. thermal conductivities of the layers down to a defined
The conversion of resistivity data to thermal con- depth, e.g. 50 m or 100 m. If the drilling density is
ductivities is shown in Fig. 6. Data inversion of poor, the layer sequence of resistivity measurements
SkyTEM measurements leads to a sequence of 1D (VES or airborne EM) can be used to complement the
resistivity-depth distributions along the flight line. data base. This is demonstrated for a selected location
This resistivity-depth distribution reflects the layer in Northern Schleswig-Holstein where the SkyTEM
sequence of the underground. In Fig. 6 the 2 low results were weighted equally to the drilling results.
resistivity layers (12.4 and 4.6 m) are geologically
interpreted as clay, while the top layer with a spe-
cific electrical resistivity of 53.6 m is interpreted REFERENCES
as saturated sand. Using the above-quoted thermal
conductivities for sand and clay an effective thermal Bernier, M.A. 2006. Closed-Loop Ground-Coupled Heat
Pump Systems. ASHARE Journal: 13–19.
conductivity down to 100 m for this location can be
CLIWAT working group 2011. Groundwater in a future cli-
calculated to 1.9 W/mK. The uppermost (unsaturated) mate – the CLIWAT handbook. Central Denmark Region,
layer is not sufficiently resolved by SkyTEM measure- Vejle.
ments, this can be compensated by additional electrical Ditlefsen, C., Vangkilde-Pedersen, T., Sørensen, I., Bjørn, H.,
Schlumberger soundings at those locations. Lajer Højberg, A., Møller, I. 2013. GeoEnergy – a national
An example from an area in Northern Schleswig- shallow geothermal research project. Proceedings Euro-
Holstein is shown in Fig. 7. Here data from a SkyTEM pean Geothermal Congress 2013, Pisa, Italy.
survey in the German – Danish border region flown Hellström, G. 1991. Ground heat storage – Thermal Analyses
in cooperation with LIAG in the scope of the INTER- of Duct StorageSystems. University of Lund, Sweden.
Hellström, G., Sanner B 2000. Earth Energy Designer – users
REG project CLIWAT (CLIWAT working group 2011)
manual.
were used to produce additional sampling points for LLUR 2011. Leitfaden zur geothermischen Nutzung des
the thermal conductivity map. The use of SkyTEM oberflächennahen Untergrundes. Landesamt für Land-
data points leads to a more detailed picture and an wirtschaft, Umwelt und ländliche Räume Schleswig-
enhanced reliability of the thermal conductivity map. Holstein, Geologischer Dienst, Flintbek.
VDI 4640 2001. Thermische Nutzung des Untergrundes-
Erdgekoppelte Wärmepumpenanlagen, Teil 1. Verein
6 CONCLUSIONS Deutscher Ingenieure, Düsseldorf.
For geothermal planning maps showing the lateral dis-

tribution of thermal conductivities in the underground
a sufficient data base is required. Normally these maps
are based on drilling results. To each layer reached by
the drilling a material specific thermal conductivity
497
Validation of a numerical model for the design of borehole heat exchangers

in the presence of groundwater flow
J. Van Steenwinkel, D. Simpson, M. Degros & W. Vienne

AGT – Advanced Groundwater Techniques, Aartselaar, Belgium
IFTech Belgium
G. Van Lysebetten
WTCB – Scientific and Technical Centre for the Construction Industry, Brussels, Belgium
M. Müller-Petke
LIAG – Leibniz Institute for Applied Geophysics, Hannover, Germany
ABSTRACT: Groundwater flow is often disregarded in the design of Borehole Heat Exchangers. Without
groundwater flow, the transport of heat from the carrier fluid to the surrounding ground happens through con-
duction, which is a relatively slow process. The temperature distribution around the loop is axially symmetric,
assuming horizontally homogeneous layers. In the presence of groundwater flow, the thermal energy is carried
away by advection. The goal of this research was to validate a finite-difference numerical model with a controlled
field experiment, in order to develop a design strategy for borehole heat exchangers in the presence of ground-
water flow. A test field was installed in the north-east of Belgium in order to create a controlled field experiment
with an induced groundwater flow. A single Borehole Heat Exchanger (double-U pipe) was installed down to 50
m depth. A pumping well was drilled at 5 m distance and 6 piezometer wells were installed in two orthogonal
directions at different distances from the pumping well. The Borehole Heat Exchanger was equipped with a fibre
optic cable inside and outside the loop, to measure the temperature distribution in depth. First, a pumping test
was carried out in order to deduct the hydraulic parameters of the formation. Second, a thermal response test was
carried out to determine the thermal parameters. Finally, a second thermal response test was conducted while
pumping the well, to measure the effect of the groundwater flow. A finite-difference model was constructed with
MODFLOW2000 and MT3DMS, in order to simulate the groundwater flow and heat transport, respectively.
The field measurements could be accurately simulated with the numerical model, with minor adjustments of the
model parameters. Therefore, the model looks a promising tool for the design of borehole heat exchangers in
the presence of groundwater flow.
1 INTRODUCTION controlled field experiments where both the ground-

water flow and the thermal transport are well known.
Groundwater flow is often disregarded in the design Controlled field experiments are not often used, since
of Borehole Heat Exchangers. Without groundwater it is quite costly to install. However, they have the
flow, the transport of heat from the carrier fluid to advantage to eliminate a lot of variables that influence
the surrounding ground happens through conduction, the flow and heat transport, enabling to benchmark
which is a relatively slow process. The temperature numerical models in a rigorous manner.
distribution around the loop is axially symmetric, The goal of this research was to validate a finite-
assuming horizontally homogeneous layers. In the difference numerical model with a controlled field
presence of groundwater flow, the thermal energy experiment, in order to develop a design strategy
is carried away by advection creating an asymmet- for borehole heat exchangers in the presence of a
ric heat plume downstream of the BHE. Recently, groundwater flow.
there has been an increased interest in the effect of
groundwater flow in the design of Borehole Heat
Exchangers. Both analytic, such as the moving finite 2 EXPERIMENTAL SITE SETUP AND
line source model (Carslaw & Jaeger, 1959), and MEASUREMENTS
numerical models (finite-element or finite difference,
Capozza & Zarrella, 2013; Gehlin & Hellström, 2003; The experiments carried out on the site had the
Bauer, 2009) were used to simulate the influence following objectives:
of groundwater flow. The models can be validated • Determine the local hydrogeology with borehole
against existing BHE-fields (Bauer et al., 2012) or in descriptions and logging.
499
• Estimate the hydraulic parameters with a pumping
test and flow test.
• Estimate the thermal transport parameters based on
a thermal response test, once without and once with
an induced groundwater flow by pumping.
• Monitor the temperature distribution during the
tests to validate the numerical model.
2.1 Local hydrogeology

An experimental site was selected near the village
Hechtel-Eksel in the north-east of Belgium, based on
the following criteria:
• Presence of a sufficiently thick aquifer, with a high
transmissivity to allow a significant groundwater Figure 1. Horizontal position of the pumping well (PW) and
piezometer wells (PZ) on an aerial photo. BHE indicates the
flow. Preferably a confined aquifer to simplify the position of the borehole heat exchanger.
analysis.
• Sufficiently distant from existing abstraction wells,
rivers or other boundary conditions that could (Figure 1). The boreholes were drilled with the reverse
locally disturb the groundwater flow. rotary flush method. A borehole heat exchanger was
installed at 5 m from the pumping well down to 50 m
The local hydrogeology was determined from bore- depth. The pumping well was screened in Aquifer
hole drill cuttings description and geophysical bore- 2 between 13.5 and 50 mbgl. The piezometer wells
hole logging: each contained multiple piezometer screens at differ-
• AQUIFER 1: phreatic: ent depths (Figure 2). Piezometer PZ1 had a screen at
75 m depth (below the base of the pumping filter), to
◦ 0–1 mbgl: coarse sand (Quaternary); be able to estimate the vertical hydraulic conductivity
◦ 1–7 mbgl: fine sand, Formation of Kasterlee (Ter- of Aquifer 2.
tiary); Before pumping, the groundwater levels were mea-
• AQUITARD 1: sured in the piezometers to determine the natural
gradient of the groundwater flow. The gradient in
◦ 7–13.5 mbgl: clay layers intercalated with fine aquifer 2 was equal to 0.36%.
sand, rich in glauconite, Formation of Kasterlee A pumping test was carried out, by first conducting
and Diest (Tertiary); a step-drawdown test with four steps of one hour and
• AQUIFER 2a: confined, Miocene aquifer system: then continuing the fourth step for 18 hours. The dis-
charges were increased in steps of 7-12-16-21 m3 /h.
◦ 13.5–24.5 mbgl: fine sand with thin clay lenses, The drawdowns in all the piezometers were recorded
rich in glauconite, Formation of Diest (Tertiary); with automatic pressure sensors. The time-drawdown
• AQUITARD 2: curves were fitted with the multi-layer aquifer mod-
elling software MLU (http://microfem.nl/), of which
◦ 24.5–26.5 mbgl: clay layer, Formation of Diest the following hydraulic parameters were obtained:
(Tertiary);
• Aquifer 2: Kh/Kv = 3.9/0.77 m/d; Ss = 9.0E-6 1/m.
• AQUIFER 2b: confined, Miocene aquifer system: • Aquitard 1: Kv = 0.012 m/d.
◦ 26.5–205 mbgl: fine sand with thin clay lenses, Apart from the pumping test, a flow test was con-
rich in glauconite, Formations of Diest and ducted with a propeller flow logger while pumping at a
Bolderberg (Tertiary); constant discharge of 21 m3 /h. The flow test indicated
• AQUITARD 3: a very even distribution of the transmissivity over the
whole thickness of the screened section of Aquifer 2.
◦ 205–280 mbgl: stiff clay, Formation of Rupel
(Tertiary).
The Miocene aquifer system is well known and has 2.3 Thermal response tests (TRT)
a regional average, horizontal hydraulic conductivity
Two thermal response tests were carried out, a first one
of 12 m/d (VMM, 2008).
without and a second one with an induced groundwa-
ter flow by pumping in the pumping well. The first
test was used to determine the parameters governing
2.2 Pumping test and flow test
the conductive transport, being the thermal conductiv-
A pumping well was installed and 6 piezometer wells ity and borehole heat resistance. The second test was
along two perpendicular transects at different disused to add the advective process to the conductive
tances between 5 and 40 m from the pumping well transport.
500
Figure 2. Schematic vertical section along transect
PW-PZ6, indicating the depth of the pumping well screen
and piezometer screens below ground level and the hydroge-
ological layers.
The borehole heat exchanger was installed in a

borehole of 130 mm diameter, drilled with a rotary
flush drilling rig. A PE, double U-pipe was installed
in the borehole. Alongside the U-pipes, a U-shaped,
fibre optic cable was installed to be able to measure a
complete temperature profile (Figure 3). The annular
spacing was filled up with a thermal grout (Thermo-
Cem plus of Heidelberg Cement, λ = 2.0 W/(m.◦ K)).
Figure 3. installation of the fibre optic cables alongside
A second fibre optic cable was installed inside one (black) and inside (red) the heat carrier pipes, which are
of the U-pipes in order to measure the heat-carrier connected to the TRT-unit in the back.
fluid temperature inside the pipe at regular depths. The
fibre optic cables were connected to a measuring unit Before the last step, the TRT-test was stopped, in
(GeoDTS of AP sensing, Germany) to read and record order to measure the recovery of the temperatures.
the temperature depth profiles in time. Based on the temperature differences between ingo-
Both thermal response tests were carried out with an ing and outgoing fluid temperatures, the bulk thermal
in-house built unit (IFTech, Belgium), which consists conductivity and borehole heat resistance were cal-
of: culated with the classical TRT analysis based on the
• a heater to add heat to the heat-carrier fluid, line-source theory at 2.85 W/(m.◦ K) and 0.07 m.◦ K/W.
• circulation pumps, These are acceptable values for a typical BHE in a
• pressure control, saturated sand.
• sensors to measure incoming and outgoing temper-
atures, flow and energy.
3 NUMERICAL SIMULATION OF THE TRT
The system is regulated such that the heat carrier
TESTS
fluid circulates with a constant discharge through the
pipes while a constant power is added to fluid. This
3.1 Design of the numerical model
results in a constant temperature difference (deltaT)
between ingoing and outflowing fluid. The TRT’s were A finite-difference numerical model was constructed
carried out with the following settings: to simulate both the groundwater flow and heat trans-
port. The open-source package Modflow2000 (USGS)
• Circulation discharge: 0.92 m3 /h.
was used for the flow calculations and MT3DMSv5.3
• Power injected: 4.23 kW or 84.6 W/m borehole
(University of Alabama, US) for the heat transport,
length.
applying the mass-heat transport analogy (Langevin
• Duration 5.79 days (TRT without flow) and 18.72
et al., 2008). According to this analogy, mass is
days (TRT with flow).
equivalent to thermal energy and mass concentration
During the second TRT, a groundwater flow was to groundwater temperature. The same mass trans-
induced by pumping the pumping well. The pumping port processes of diffusion, dispersion, advection and
discharge was increased in three steps as follows: retardation can be applied to heat transport modelling.
The finite-difference grid was composed as follows:
• 0–2.71 days: no pumping, start TRT heating;
• 2.71–6.72 days: Q = 0.93 m3 /h, TRT heating; • Horizontal extension: 4 × 4 km. The model needs
• 6.72–10.05 days: Q = 2.91 m3 /h, TRT heating; to be this large to avoid that the drawdown of the
• 10.05–18.72 days: Q = 8.63 m3 /h, no heating. pumping well reaches the borders of the model.
501
• Horizontal cell dimensions: 100 × 100 m at the bor-
der, refined to 20 × 20 cm around the well field and
to 2.32 × 2.32 cm at the borehole heat exchanger.
The first refinement serves to accurately simulate
the groundwater pressures and temperatures and the
second refinement is to simulate the borehole fluid
flow through the pipe. The inner surface of the pipe
corresponds to surface of the finest cell.
• Number of layers: 13, partly to capture the hydro-
geological layers and partly to simulate the bottom
of the borehole heat exchanger.
The pumping well was inserted as a constant flux
boundary condition.
Figure 4. Temperature evolution of the heat carrier fluid
The flow of the heat-carrier fluid through the during the two TRT-tests, at 33 m depth below ground sur-
borehole heat exchanger was numerically reproduced face: measured data and modelled values for ingoing and
by attributing a high vertical hydraulic conductivity outcoming pipes.
inside the pipes and a zone of very low hydraulic
conductivity around the pipes. The heat-carrier fluid
was circulated in the model by introducing a constant
flux boundary condition (Well-package in Modflow)
at the top of the model. The temperature of the out-
flowing fluid was accepted as temperature for the
inflowing fluid, a feature which is incorporated in the
latest MT3DMS version (Zheng, 2010). A constant
heat flux was introduced with the mass-loading option
of the Well-package, just after the ingoing constant
flux boundary condition.
Both the top and bottom of the model were assumed
to be no-flux boundary conditions, both for groundwa-
ter flow and heat transport.
The hydraulic properties were derived from the
pumping test, while the thermal properties were esti-
mated based on the TRT without groundwater flow and
literature values as initial estimates. The parameters
thermal conductivity, total porosity and heat capac-
ity of the solid matrix were varied until a best, visual
match was obtained between measured and modelled
temperatures:
• Initial temperature of the groundwater and solid
matrix: 11.5◦ C.
• Bulk thermal conductivity KT_bulk = 2.75 W/(m.◦ K).
• Total porosity θ = 0.2 m3 /m3 .
• Heat capacity of the dry solid matrix Cp_solid =
824 J/(kg.◦ C)
• Dry bulk density of the solid matrix ρb = 1590 kg/m3 .
Figure 5. Measured and simulated heat carrier fluid tem-
peratures at different depths and times.
3.2 Simulation of the TRT tests
The two TRT-tests were simulated one after the other,
as in the real situation. The best match between mea-
sured and modelled temperatures is shown in Figure had the same heat input as the first, so it is not clear
4 and Figure 5. The fibre optic measurements allow yet why the temperatures rise higher than in the first
to evaluate the temperature profiles in depth and time. TRT-test. One hypothesis is the increased air temper-
The general evolution of the heat carrier fluid tempera- ature during the second test, which was not taken into
tures was reproduced well. The decline in temperatures account. Diurnal fluctuations in the measured temper-
due to the advective groundwater flow in the second atures at shallow depths indicates there is an influence
TRT-test was modelled similar in magnitude and shape of air temperature on the carrier fluid temperatures.
as the measured data. However, the simulated, absolute Furthermore, the recovery after the second TRT-test is
values of the temperatures of the second TRT-test were slower than the measured data. These differences are
lower than the measured values. The second TRT-test still being investigated.
502
4 CONCLUSIONS REFERENCES
The influence of advection on the thermal transport Bauer, D., Heidemann, W., Müller-Steinhagen, H., Diersch,
around a BHE was successfully induced by pumping H.-J.G. 2009. Modelling and simulation of groundwater
influence on borehole thermal energy stores. Proceed-
in a nearby groundwater well. The strong advection
ings Effstock – 11th International Conference on Energy
provoked a significant decrease in temperature in the Storage, Stockhom, Sweden, 2009.
BHE carrier fluid as most of the heat is exported Bauer, D., Heidemann, W., Drück, H., 2012. Validation of
rapidly. The effect of the advection could be well repro- groundwater flow and transport modeling tool for bore-
duced with a finite-difference, numerical groundwater hole thermal energy stores based on FEFLOW. Innostock
model. Some issues still need to be investigated, such 2012 – The 12th International Conference on Energy
as the influence of air temperature variation on the Storage, Lleida, Spain.
BHE, the vertical distribution of thermal conductivi- Carslaw H.S., Jaeger J.C. 1959. Conduction of heat in solids.
ties and the effect of different numerical discretization Claremore Press, Oxford.
Gehlin, S.E.A., Hellström, G., Influence of thermal response
schemes.
test by groundwater flow in vertical fractures in hard rock.
Renewable Energy 28, 2221–2238.
Langevin, C.D., Thorne, D.T., Jr., Dausman, A.M., Sukop,
ACKNOWLEDGEMENTS M.C., Guo, Weixing 2008. SEAWAT Version 4: A Com-
puter Program for Simulation of Multi-Species Solute and
This research was funded by the Flemish Govern- Heat Transport: U.S. Geological Survey Techniques and
ment, Department of Environment, Nature and Energy Methods Book 6, Chapter A22, 39 p.
(LNE). The authors wish to thank Benvitec and its VMM, 2008. Grondwater in Vlaanderen: Het Centraal Kemp-
subsidiary Kempische Metaalwerken for their cooper- isch Systeem. Vlaamse Milieumaatschappij. Aalst. 110 p.
Zheng, C. 2010. MT3DMS v5.3 a modular three-dimensional
ation and access to the experimental site. Also special
multispecies transport model for simulation of advec-
thanks to drilling company Verheyden Putboringen for tion, dispersion and chemical reactions of contaminants
their cooperation in the drilling and installation of the in groundwater systems – Supplemental user’s guide.
borehole heat exchangers. University of Alabama, 56 p.
503
Fiber-optic temperature measurements in closed-loop geothermal

systems: A case study in heterogeneous bedrock
G. Radioti, S. Delvoie, K. Sartor, F. Nguyen & R. Charlier

University of Liege, Liege, Belgium
ABSTRACT: In order to study the behaviour of shallow closed-loop geothermal systems, four borehole heat
exchangers equipped with fiber optics were installed on the campus of the University of Liege (Liege, Belgium)
over a surface area of 32 m2 . This paper presents continuous, high-resolution temperature profiles measured along
the boreholes length at different phases: at the undisturbed state, during hardening of the grouting material, during
the recovery phase of a Distributed Thermal Response Test (DTRT) and during a DTRT of a long duration
(7 months). The undisturbed ground temperature is affected by the heat loss from ground structures located
close to the boreholes, as also indicated by a 3D numerical model. Temperature profiles during hardening of
the grouting material indicate extended fractured zones in the rock mass. Temperature measurements during
the recovery phase can be correlated to rock layers with different mineral content. The results are in good
agreement with those of the borehole televiewer logging method. The long duration DTRT allow us to follow the
thermal plume in the heterogeneous rock mass. Moreover the effect of the duration of the test to the calculated
mean thermal conductivity and borehole thermal resistance is investigated. The presented analysis could provide
information on bedrock heterogeneity, on the anisotropic thermal behaviour of the rock mass and on the ground
temperature variations due to heat loss from ground structures. These information could significantly contribute
to the long-term behaviour prediction of the geothermal system and the geothermal reservoir potential.
1 INRODUCTION The site geology is characterised by deposits of sand

and gravel until a depth of approximately 8 m. The
In order to study the behaviour of closed-loop geother- bedrock follows which consists mainly of siltstone and
mal systems in heterogeneous bedrock, four Borehole shale interbedded with sandstone. Fractured zones are
Heat Exchangers (BHEs), namely B1 to B4, were detected in the rock mass mainly until a depth of 35 m.
installed on the campus of the University of Liege The mean layer dip angle is approximately 45◦ SE.
(Liege, Belgium) over a surface area of 32 m2 (Radioti Moreover, azimuth and deviation were measured by
et al. 2013). The four BHEs, equipped with double- magnetometers and inclinometers (Monier-Williams
U geothermal pipes of 100 m long are located at a et al. 2009). Based on these data, the horizontal dis-
distance of approximately 15 m to a building (SEGI) tance through depth between B2 and the other three
and 6.6 m to an underground structure (feeder pipe) boreholes was estimated and presented in Figure 2. B1
(Fig. 1). and B2 are characterised by roughly the same lithos-
After drilling the boreholes, a borehole televiewer tratigraphy, which is observed at different depths in
was lowered inside the four boreholes. A detailed B3 and B4, due to the layer dip angle orientation. The
bedrock characterisation was conducted based on distance through depth between B1 and B2 oscillates
acoustic borehole imaging data, gamma-ray logging around 4.1 m. The distance between B2 and B3 or B4
data and cuttings observation (Radioti et al. 2015a). decreases through depth, becoming almost the half at
the bottom of the boreholes.
During the installation of the geothermal pipes,
fiber optic cables were attached along the pipe loops.
Fiber optics allows us to obtain continous, high-
resolution temperature profiles along the pipes length
by applying the DTS technique (Hermans et al. 2014).
This technique is based on Raman optical time domain
reflectometry. The fiber ends are connected to the
DTS instrument, which injects a laser pulse. The
light travels into the optical fiber and is scattered
Figure 1. Site location on the campus of the University of and reemitted from the observed point. The backscat-
Liege (Liege, Belgium). tered light is spread across a range of wavelengths.
505
Figure 2. Horizontal distance between B2 and the other
three boreholes.
Figure 3. Undisturbed ground temperature measured by

The Raman backscatter signal is temperature sensitive. fiber optics in B2.
The temperature along the fiber is determined by the
intensity of Raman Stokes and anti-Stokes signals.
The position of the temperature reading is deter- approximately 18 m correspond to the thermally unsta-
mined by the arrival time of the reemitted light pulse. ble zone, where ground temperature is influenced by
The temperature resolution (standard deviation) of the the air temperature. Below 18 m, measured tempera-
DTS instrument used in this study is in the order of tures appear invariant to time in the two-years period
0.05◦ C. Temperature was recorded every 20 cm (sam- and the temperature decreases through depth at a mean
pling interval) with a spatial resolution of 2 m. Spatial rate of approximately 0.25◦ C/10 m, opposite to the
resolution determines the slope width of a measured geothermal gradient effect.
temperature change and is an important parameter for A 3D numerical model was developed, by using
the temperature accuracy of local hotspots. If the width the finite element code LAGAMINE (Charlier et al.
of the hotspot is lower than the spatial resolution, the 2001, Collin et al. 2002), to investigate if the mea-
measured temperature is reduced by approximately the sured ground temperature profiles could be a result
ratio of hotspot width to spatial resolution (Hoffman of the heating of the ground due to structures heat
et al. 2007). loss into the subsurface (Radioti et al. 2015b). This
The boreholes were backfilled with the following model takes into account the heat loss through the
grouting materials: B1 and B3 with a silica sand-based basement of the SEGI building and through the shell of
commercial material (Geosolid, λ = 2.35 W/mK), B2 the feeder pipe. The feeder is simulated with a surface
with a bentonite-based commercial material (Füll- heating element. The heat loss (150 W/m length) is
binder, λ = 0.95 W/mK) and B4 with a homemade calculated based on temperature measurements inside
admixture with graphite (λ = 2.5 W/mK) (Erol & the feeder pipes (Sartor et al. 2014). The heat loss
François 2014). through the foundations of the SEGI building is simu-
Temperature was measured at the undisturbed state, lated by imposing a constant temperature through time
during hardening of the grouting material and dur- (17.7◦ C) at the whole building surface, as measured
ing Distributed Thermal Response Tests (DTRT) (Fujii by temperature data loggers at the basement of the
et al. 2006) conducted in situ in the four BHEs. At each building. The initial ground temperature (before the
depth position two temperature measurements were presence of engineering structures, in 1970) is calcu-
available, since the fiber optic cables were attached in lated analytically based on air temperature data and
a loop along the pipes. The temperature profiles pre- on the ground thermal properties (Tinti 2012). The
sented hereafter correspond to the average of the these air temperature parameters used in this calculation are
two measurements. based on statistical data for the Sart-Tilman area for
a period of 20 years (climate-data.org) and the mean
ground thermal conductivity on Thermal Response
2 UNDISTURBED GROUND TEMPERATURE Tests (TRTs) conducted in situ in the four boreholes.
In this model the simplified assumptions are made
Temperature was measured at the undisturbed state that at bare ground the temperature is constant through
several times in a period of 2 years (Fig. 3). The first time, equal to the average annual air temperature, and
506
Figure 4. Numerical results of ground temperature at the Figure 5. Temperature profiles measured by the fiber optics
location of the boreholes. in the four boreholes in 2014.
that the ground surface under the pavement can be

simulated by a no-heat-flux boundary condition.
The numerical results of the ground temperature are
presented in Figure 4.The geothermal gradient effect
is evident in the temperature profile for the year 1970,
which corresponds to the ground temperature before
the presence of engineering structures. The heating
of the ground, due to the feeder and the SEGI build-
ing operation, modifies the temperature gradient at the
boreholes location until a depth of 100 m after 20 years
(1990). The heating effect becomes progressively evi-
dent at greater depth, reaching a depth of 130 m after
45 years (2015). Moreover, the numerical results for
the first 100 m are in good agreement with the tem-
perature profiles measured by the fiber optics for the
measurement period (2013–2015).
Figures 5 shows the temperature profiles in 2014
measured by the fiber optics in the four boreholes. The
temperature profiles in B2, B3 and B4 coincide with
each other and display a higher temperature in the first
20 m compared to B1. This could be attributed to the
distance of each borehole from the feeder pipe, which
is located at an average depth of 2.5 m. Given that B2,
Figure 6. Numerical results of ground temperature at the
B3 and B4 are 4 m closer to the feeder pipe than B1, location of the four boreholes for the year 2014.
the heat loss effect from the feeder pipe will be more
enhanced in theese three boreholes. This effect is also
observed in the numerical analysis results presented in the grouting material, which results in a temperature
Figure 6. increase along the borehole length. In the following
days temperature retrieves its initial undisturbed pro-
file. In the first 18 m temperature is influenced by
3 DURING HARDENING OF THE GROUTING the air temperature. Below 18 m, the profile during
MATERIAL hardening of the grouting material is characterised
by a local maxima, of a significantly increased tem-
Figure 7 shows the temperature profile measured one perature value, at 29 m. This location corresponds
day after injecting the grouting material in B4. Heat to an extended fractured zone, more than one meter
is generated during the first hours of hardening of thick, based on the borehole televiewer data analysis.
507
Figure 7. Temperature profiles at the undisturbed state and
Figure 8. Temperature difference after 4h of recovery and
during hardening of the grouting material in B4.
natural gamma radioactivity data for B3.
sandstone/siltstone is evident in the in-situ measure-

The local maxima could be attributed to a local larger ments despite the relatively small thickness of these
quantity of grouting material and/or local lower ther- layers.
mal diffusivity due to gathering of fractures. A local Sandstone/siltstone layers thinner than 1.2 m are not
maxima is also observed in the profile of B1, at a depth always detectable in the recovery profile. The width of
corresponding to the center of an extended fractured hotspots corresponding to these layers is lower than the
zone (Radioti et al. 2015a). measurements spatial resolution (2 m) and the sam-
Random large fractures (more than 10 cm wide) or pling interval (20 cm). Even if these hotspots were
smaller fractures (of an opening between 5 and 10 cm) included in the measurement points, the measured tem-
cannot be identified in the profiles during hardening perature could be significantly reduced, and hence be
of the groutting material in this case-study This could undetectable in the temperature profile.
be attributed to the width of hotspots corresponding
to random fractures, which is lower than the spatial
resolution (2 m) and sampling interval (20 cm). 5 DURING THE HEATING PHASE OF A
LONG-DURATION DTRT
A TRT of a heating phase of 7 months was conducted

4 DURING THE RECOVERY PHASE OF A in B2. During this test, temperature is measured at
DTRT the pipe inlet and outlet in B2, as well as by the
fiber optics in the four boreholes. These measurements
Figure 8 (right) presents the naturally occurring allow to investigate the behaviour of the BHE in B2
gamma radiation data (moving average of 2 m) for through time and to follow the thermal plume in the
B3, which were measured to characterise the clay con- heterogeneous rock mass.
tent of the rock formation. High gamma-ray values
indicate shale/siltstone layers while low gamma-ray
values indicate sandstone/siltstone layers. Figure 8
5.1 Mean ground thermal conductivity and mean
(left) presents the temperature difference between a
borehole thermal resistance
recovery profile (after 4h of recovery) and the undis-
turbed temperature profile for B3. Local peaks in this The mean thermal conductivity of the surrounding
profile indicate an uneven heat transfer rate through ground and the mean borehole thermal resistance were
depth. calculated based on the pipe inlet and outlet tem-
It is observed that temperature local minima perature data in B2. The Infinite Line Source model
correspond to gamma-ray local minima, indicating (Carslaw & Jaeger 1959) was used for this calcula-
sandstone/siltstone layers, while temperature local tion and the forward regression technique was applied
maxima to gamma-ray local maxima, indicating (Tinti 2012). In this technique a start time point is cho-
shale/siltstone layers. The higher thermal diffusivity of sen and the end time point increases gradually. During
508
Table 1. Mean ground thermal conductivity and mean
borehole thermal resistance for the four BHEs.
λ∗grout Rb λground
BHE (grouting) W/mK Km/W W/mK
B1 (Geosolid) 2.35 0.079 2.94

B2 (Füllbinder) 0.95 0.095 2.71
B3 (Geosolid) 2.35 0.078 3.09
B4 (Hom. admixture) 2.46 0.080 2.80
* Erol & François 2014.
Table 1 presents ground thermal conductivity and

borehole thermal resistance results, based on TRTs
of a duration of 7 days conducted in the four
Figure 9. Mean ground thermal conductivity through time BHEs. The average ground thermal conductivity is
for the first 3 months of the heating phase in B2. 2.88 ± 0.16 W/mK. The Geosolid admixture and the
homemade admixture with graphite display approx-
imately the same thermal conductivity and the cor-
responding BHEs are characterised by a borehole
thermal resistance in the order of 0.08 Km/W. The
Füllbinder admixture (B2) has a lower thermal con-
ductivity and the corresponding thermal resistance is
19% higher than the other three BHEs. Though the
borehole thermal resistance in B2 decreases with time
for a longer duration of the heating phase, as presented
previously in Figure 10.
5.2 Temperature evolution in the rock mass

Figure 11 displays the temperature increase through
time in the four boreholes. Data for B2 correspond
to the mean temperature of the pipe inlet and out-
let. Data for the other three boreholes correspond to
the depth-average temperature measured by the fiber
Figure 10. Mean borehole thermal resistance through time optics. Temperature starts to increase after approxi-
for the first 3 months of the heating phase in B2. mately 5 days in B3, 10 days in B1 and 16 days in B4.
Temperature increases at a higher rate in B3 than the
other two boreholes, since this borehole has the small-
the first few hours of the test, the temperature evolution est distance to B2. Likewise, the lowest increase rate
is mainly controlled by the borehole filling and is not is observed in B4 which is located at the greatest dis-
representative of the surrounding rock thermal prop- tance to B2. These measurements illustrate the effect
erties (Sanner et al. 2005). A start time point of 10 h of the distance to the heating source on the temperature
was chosen for this analysis, based on optical cross- evolution in the rock mass.
checking of the measured data. Thermal conductivity Figure 12 shows temperature measured in B3 dur-
and borehole thermal resistance were also evaluated ing the heating phase of the DTRT in B2. The first
for data in limited time windows. approximately 18 m are influenced by the air temper-
Figures 9 shows the calculated mean thermal con- ature. Temperature starts to increase at the lower part
ductivity for the first three months of the test. The of the borehole, since this part has a smaller distance
obtained thermal conductivity values range from to B2 than the upper part and is characterised by thick
2.66 W/mK to 3.06 W/mK (difference less than 15%) sandstone layers.
based on the two techniques for the period of 3 months. Figure 13 shows the temperature increase evolution
Figures 10 shows the calculated mean borehole at a depth location of 77.2 m, which corresponds to a
thermal resistance for the first three months of the local maxima in the temperature profile, and at a depth
test. The borehole thermal resistance ranges from location of 82.8 m, which corresponds to a local min-
0.093 Km/W to 0.11 Km/W for the first 20 days. ima of the temperature profile. These locations have
Between 20 days and 30 days the thermal resistance approximately the same distance to B2. The tempera-
decreases of 17% and it converges around a value of ture difference between the two profiles increases with
0.085 Km/W for the rest of the time period. time, reaching an order of magnitude of 0.3◦ C after 90
509
Figure 13. Temperature increase evolution at certain depths
in B3.
Figure 11. Temperature evolution in the four BHEs (B2:

mean of pipe inlet and outlet temperature – B1,B3,B4:
depth-average temperature by fiber optics).
Figure 14. Temperature difference after 78 days of heating

and natural gamma radioactivity data for B3.
undisturbed temperature profile (left), together with

the gamma-ray data for B3. It is observed that the
local maxima at 77.2 m depth corresponds to the top
Figure 12. Temperature profiles in B3 during the heating of a sandstone/siltstone layer and the local minima at
phase of the DTRT in B2. 82.8 m depth to the top of a shale/siltstone layer. B3 is
characterised by the same lithostratigraphy with B2 but
days. This indicates a higher heat transfer rate at the at a lower depth, due to the layer dip angle orientation
location of the local maxima, compared to the one of (45◦ SE). The measured profiles could be the result of
the local minima. the heat diffusion due to the uneven heat transfer rate
Figure 14 presents the temperature difference along the direction of the different layers, which are
between the temperature profile after 78 days and the characterised by a dip angle of 45◦ SE.
510
6 CONCLUSIONS Charlier, R., Radu, J.-P. and Collin, F. (2001). Numerical mod
elling of coupled transient phenomena. Revue Française
This study presents the analysis of temperature profiles de Génie Civil, 5(6), 719–741.
measured by fiber optics in four BHEs in an heteroge- Climate-data.org. Climate data for cities worldwide. Last ac
cessed 15-10-2015, http://en.climate-data.org/
nous bedrock. The undisturbed ground temperature
Collin, F., Li, X.L., Radu, J.-P. and Charlier, R. (2002).
profiles are characterised by a negative temperature Ther mo-hydro-mechanical coupling in clay barriers.
gradient. These profiles can be the result of the heating Engineering Geology, 64, 179–193.
of the ground by structures located close to the bore- Erol, S. and François, B. (2014). Efficiency of various grout-
holes, as verified also by the 3D numerical model. ing materials for borehole heat exchangers. Appl. Therm.
Temperature profiles during hardening of the grout- Eng., 70, 788–799.
ing material allow to locate extended fracture zones, Fujii, H., Okubo, H., and Itoi, R. (2006). Thermal Response
more than one meter in this specific case. Based Tests Using Optical Fiber Thermometers. GRC Transac
on temperature measurements during the recovery tions, 30, 545–551.
Hermans, T., Nguyen, F., Robert, T. and Revil, A. (2014). Ge
phase of a DTRT layers with different mineral con-
ophysical methods for monitoring temperature changes in
tent can be detected since they display a different shallow low enthalpy geothermal systems. Energies, 7
thermal behaviour. The results are in good agreement (8), 5083–5118.
with those of the borehole televiewer logging method Hoffmann, L., Müller, M. S., Krämer, S., Giebel, M.,
Layers thinner than 1.2 m or random fractures can- Schwotzer, G. and Wieduwilt, T. (2007). Applications of
not be identified by this procedure in this case-study. Fibre Optic Temperature Measurement. Proc. Estonian
The resolution of the applied procedure is limited by Acad. Sci. Eng., 13 (4), 363–378.
the measurement parameters, spatial resolution and Monier-Williams, M.E., Davis, R.K., Paillet, F.L., Turpen-
sampling interval. ing, R.M., Sol, S.J.Y and Schneider, G.W. (2009). Review
of Borehole Based Geophysical Site Evaluation Tools
The influence of the duration of a TRT to the mean
and Techniques (Rep. NWMO TR-2009-25). Retrieved
ground thermal conductivity and mean borehole ther- from Nuclear Waste Management Organization web-
mal resistance is also investigated. In this case-study, site: http://www.nwmo.ca/uploads_managed/MediaFiles/
thermal conductivity varies less than 15% for a heat- 1770_n wmotr-2009-25boreholebasedgeophysicaltools_
ing period of 3 months, based on temperature data of r0d.pdf
the pipe inlet and outlet. The mean borehole thermal Radioti, G., Charlier, R., Nguyen, F. and Radu, J.-P. (2013).
resistance decreases of 17% after the first 30 days of Thermal Response Test in Borehole Heat Exchangers
the test. Moreover, the BHE with lower grouting ther- Equipped with Fiber Optics. In: Proceedings, Interna-
mal conductivity displays a higher borehole thermal tional Workshop on Geomechanics and Energy: The
Ground as Energy Source and Storage. EAGE, Lausanne,
resistance. Measuring the temperature in all the bore-
Switzerland, 96–100.
holes during the long-duration TRT allows to follow Radioti, G., Delvoie, S., Radu, J.-P., Nguyen, F. and Charlier,
the thermal plume in the heterogenous rock mass. The R. (2015a). Fractured bedrock investigation by using high-
effect of the distance to the heating source and the resolution borehole images and the Distributed Temper-
effect of the rock heterogeneity are displayed in theese ature Sensing technique. In: ISRM Congress 2015 Pro-
measurements. ceedings – Int’l Symposium on Rock Mechanics, ISRM,
Given the increasing number of closed-loop Montreal, Can ada.
geothermal systems and the wide application of TRTs, Radioti, G., Delvoie, S., Sartor, K., Nguyen, F. and Charlier,
it would be of interest to measure the temperature along R. (2015b). Fiber-optic temperature profiles analysis for
closed-loop geothermal systems: a case study. In: Proceed
the BHE length at the undisturbed state, during hard-
ings, Second EAGE Workshop on Geomechanics and Ener
ening of the grouting material and during the recovery gy:The Ground as Energy Source and Storage. EAGE, Cel
phase. Alternatively to fiber optic cables, temper- le, Germany.
ature through depth can be measured by lowering Sanner, B., Hellström, G., Spitler, J., and Gehlin, S. (2005).
down a temperature sensor into the geothermal pipe. Thermal Response Test – Current Status and World-Wide
Analysing theese profiles could give valuable infor- Application. Proceedings World Geothermal Congress
mation about the rock heterogeneity, the influence of 2005, 24–29.
the surrounding structures to the ground temperature Sartor, K., Quoilin, S. and Dewallef, P. (2014). Simulation and
field, as well as about the rock geothermal reservoir optimization of a CHP biomass plant and district heating
network. Appied Energy, 130, 474–483.
potential. These information could be included in an
Tinti, F. (2012). The probabilistic characterization of under
advanced numerical model, to predict the long-term ground as a tool for the optimization of integrated design
behaviour of closed-loop geothermal systems and to of shallow geothermal systems (Doctoral dissertation).
optimise their efficiency. Univer sity of Bologna, Italy.
REFERENCES
Carslaw, H.S., and Jaeger, J.C. (1959).Conduction of Heat
in Solids, second edition. New York: Oxford University
Press.
511
Temperature distribution in the vicinity of a borehole heat exchanger for

intermittent heat extraction: An analytical solution
S. Erol & B. François

Building, Architecture and Town Planning Department (BATir), Université Libre de Bruxelles, Brussels, Belgium
ABSTRACT: Existing analytical solutions for thermal analysis of closed-loop ground heat exchangers systems
evaluate temperature change in the heat carrier-fluid and the surrounding ground in the production period of
Borehole Heat Exchangers (BHE) only if a continuous heat load is assigned. In the present study, we solve
analytically the heat conduction/advection/dispersion equation in porous media, for intermittent heat extrac-
tion. We convolute rectangular function or pulses in time domain both for single and multi-BHEs field. The
solution includes non-symmetric configurations around the BHEs by considering anisotropic features induced
by groundwater convection or intrinsic anisotropy of thermal conductivity. Thermal dispersivity linked to the
ground water flow is also considered.
The validity of the analytical model is checked through the comparison with results obtained from numerical
finite element code. The comparison results agree well with numerical results both for conduction and advection
dominated heat transfer systems, and analytical solutions provide significantly shorter runtime compared to
numerical simulations. The developed tool allows also to investigate the recovery aspects and the sustainability
of closed-loop ground heat exchangers systems in terms of temperature and the energy deficit of the ground.
1 INTRODUCTION et al. 2003; Diao et al. 2004; Marcotte et al. 2010; Man
et al. 2010; Molina et al. 2011). Until now, intermittent
The Ground Source Heat Pump (GSHP) system tech- heat extraction can be taken into account through sim-
nologies are the most often application of the shal- plified assumptions (Eskilson 1987; Hellström 1991;
low geothermal energy use and primarily reduce the Claesson and Eskilson 1988).
energy consumption for the space heating and cooling The objective of this study is to develop an analytical
supplied from the conventional systems. solution to evaluate temperature change in the ground
In order to evaluate the necessary drilling depth of both for single and multi-BHEs that considers inter-
a borehole heat exchanger and the regulation of the mittent heat extraction, thermal conduction, advection
heat input, the specific heat extraction rate should be and dispersion. This new analytical solution may fur-
optimized regarding the characteristics of the hydro- ther help to investigate the regulatory issues such as
geological conditions for the long term operational the recovery of groundwater temperature after the use
effects in the fields. of GSHP systems, and also can be used in TRT oper-
This is particularly the case for multi-BHEs that ations for predicting the ground thermal properties
may affect significantly the ground temperature on (Erol et al., 2015).
a relatively large area. After an operation period of We start from the Green’s function which is the solu-
BHEs, the ground needs time to recover from the tem- tion of heat conduction/advection/dispersion equation
perature drop to sustain the performance of the system in porous media and apply an analytical convolution
in the long-term run (e.g. 30 years) (Signorelli 2004; of that function with a rectangular function or pulses,
Rybach and Eugster 2010). which have different period length and pulse height.
In order to investigate an operation both with the The evolution of the mean fluid temperature of the
heat extraction and the subsequent recovery periods carrying fluid to maintain a constant heat extraction
of BHEs, including the groundwater flow dispersion rate is evaluated along the time. Temperature evalua-
in a porous medium and the axisymmetric heat trans- tion in the surrounding ground is also deduced. The
fer along the BHE, the 3D numerical simulation tools developed equation is verified with the finite element
undergo a large computational effort and require long method software COMSOL Multiphysics. Further-
execution time. On the other hand, most of the analyti- more, the energy balance of the ground is investigated
cal solutions described in literature consider a constant with the analytical solution during 30 years of produc-
continuous heat extraction/injection in time merely for tion period, and the subsequent energy recovery of the
a single BHE (Eskilson 1987; Zeng et al. 2002; Sutton ground after the system is shutdown.
513
2 ANALYTICAL DEVELOPMENT The solution of the partial differential equation for
heat transfer in porous media (Eq. 1) is obtained from
2.1 Single BHE the Green’s function G of a pulse point source QP at
the given point coordinates (x , y , z ) and time t = 0
In geothermal literature, the existing finite and cylin-
(Metzger 2002):
drical analytical solutions with a constant heat load
may provide satisfactory estimation of ground ther-
mal parameters to design closed-loop ground heat
exchangers systems (Deerman and Kavanaugh 1991;
Kavanaugh and Rafferty 2014; Gehlin 2002). In a real
case, the systems can be operated with various periods
in a given time for different heat extraction/injection
rates, instead of a continuous operation as assumed in
most of other previously presented analytical methods.
Some authors evaluated the temperature change for
TRT operation in the vicinity of a single BHE or BHEs
field with an analytical solution by using multiple load
aggregation algorithms (Yavuzturk 1999; Bernier et al. In order to take into account the axial effect and the
2004; Marcotte and Pasquier 2008; Lamarche 2009; groundwater flow, this solution can be applied for the
Michopoulos and Kyriakis 2009; Michopoulos and response of a constant line-source with finite length H
Kyriakis 2010). However, some of those approaches along the vertical z direction with a pulse heat extrac-
may not be appropriate in all cases to evaluate the tion after applying moving source theory (Carslaw and
accurate temperature change in the ground due to Jaeger 1959) by integrating Eq. 5 along the z-axis
neglecting the axial effect, considering only single (Diao et al. 2004):
BHE or not taking into account groundwater flow. In
particular when Darcy’s velocity in porous media is
considered, the thermal dispersion coefficients must
be taken into account, because thermal dispersion has
a large impact on the distribution of the temperature
plume around BHE, for Darcy’s velocity larger than
×10−8 m/s (Molina et al. 2011).
The governing equation of the heat advec-
tion/dispersion in porous media is given as follows:
where QL is a pulse heat input per meter depth and vT

is thermal transport velocity that can be calculated as
follows (Molina et al. 2011):
in which uw,x is the Darcy’s velocity on the x-direction,
s is a volumetric heat source, and ρm cm is the vol-
umetric heat capacity of the medium, which can be
calculated as the weighted arithmetic mean of the
solids ρs cs and volumetric heat capacity of water ρw cw In order to simplify Eq. 6, the exponential function can
(de Marsily 1986): be integrated by using u-substitution method:
The components of effective longitudinal and trans-

verse thermal conductivities are defined on the direc-
tions of x, y and z as follows (Hopmans et al.
2002):
where λm is the bulk thermal conductivity of porous

medium in the absence of groundwater flow, l and
t are the longitudinal and transverse dispersivities,
respectively.The thermal dispersion is a linear function
of groundwater flow and relates to the anisotropy of the
velocity field (Molina et al. 2011; Sauty et al. 1982).
514
The limits of u-value becomes: To apply an intermittent injection or extraction of
heat in time domain, we convolute analytically Eq.
15 with a single or a series of different rectangular
pulses referring to the duration of operations in time.
For instance, f (x, y, z, t) function is convoluted with
a rectangular heat flow rate function qL (t) defined as
Substituting Eq. 10 and Eq. 11 into Eq. 8, allows to follows:
re-write the equation as:
in which T is the period of heat extraction. qL is the

heat flow rate taken as independent of the depth in our
simulations.
The convolution of qL and f function is written as
follows:
The integration of exponential function f = exp(−u2 )

can be expressed with error function: For the analytical evaluation of the convolution inte-
gral equation, we discretize both qL and f functions
with a differential of t. So, the convolution as a sum
of impulse responses at coordinates (x, y, z) is given as:
By taking the integration of exponential function, where n denote the time span, i t is the time delay of
therefore, Eq. 12 reduces to: each unit impulse, and the delayed and shifted impulse
response becomes qL (i t)f (t-i t) t.
By using the same method, it is possible to convolute
f function with rectangular pulses which have different
pulse height and length in given identical time span of
f function. Thus, recovery period of the ground can
be investigated after a production of a single BHE and
the numerical computational effort will be decreased.
2.2 Multi-BHEs
In case of multi-BHEs, analytical solution Eq. 15 can
be solved in a sum function (Eq. 19) depending on the
grid coordinates of each line heat source as illustrated
in Figure 1.
with simplification, Eq. 14 can be expressed with the

error functions as follows:
515
Figure 1. Illustration of multi-BHEs geometry demonstrat- Figure 2. Load profile of heat extraction.
ing the grid coordinates.
Table 1. Common initial input parameters for the model
in which s represents the number of BHEs. We con- domain of single and multi-BHEs field.
sider the impact of groundwater flow on each BHE at
x direction by taking into account thermal transport Parameters Value
velocity vT .
The sum C(x, y, z, t) can be convoluted as described Initial temperature ◦ C (To ) 0
in the previous Section 2.1 to apply intermittent heat Bulk thermal conductivity of porous medium 2.4a
extraction as follows: W m−1 K−1 (λm )
Effective thermal conductivity in the 6.6b
longitudinal direction W m−1 K−1 (λx )
Effective thermal conductivity in the transverse 2.82b
direction W m−1 K−1 (λy = λz )
Volumetric heat capacity MJ m−3 K−1 (ρm cm ) 2.8a
Groundwater flow / discharge m s−1 (uw,x ) 1 × 10−6c
3 VALIDATION
Longitudinal thermal dispersion coefficient ( l ) 1d
Transverse thermal dispersion coefficient ( t ) 0.1d
The developed analytical solutions (Eq. 18 and Eq.
20), for intermittent heat extraction, are verified with a
Representative values taken from (VDI-Richtlinie 2000).
3D numerical models. For the verification, numerical b
Calculated values according to Eq. 3 and 4.
model setup, initial and boundary conditions of the c
Assigned only for the models in which heat advec-
model, input parameters and comparison of the numer- tion/dispersion is considered.
d
ical and the analytical solution results are presented in Values taken from (Hecht-Méndez et al. 2013) to calculate
the following. effective thermal conductivities.
3.1 Numerical model setup

In order to validate the analytical solution devel- As the boundary conditions, the load profile of heat
oped above, simple cases for single and multi-BHEs extraction with three different extraction periods can
have been considered through numerical models be seen in Figure 2. The simulation time is restricted
using COMSOL Multiphysics software. The common to 160 days. The top of the model surface tempera-
numerical characteristics are described here for both ture is fixed to 0◦ C, as well as identically assigned
single and multi-BHEs field models. The study is car- initial temperature, to observe the relative tempera-
ried out by a 3D homogeneous model domain and BHE ture change in the subsurface. Initial input parameters
is represented with the line heat source(s) (W m−1 ). A are given in Table 1. Thermal dispersion is taken into
model domain of 40 m × 40 m in the horizontal direc- account leading to anisotropic thermal conductivity of
tion and 60 m in the vertical direction is set for the the medium, as described by Eq. 3 and 4.
simulation. The length of line heat source is 50 m for The number of elements changes depending on the
each. The mesh is generated using uniform tetrahedral model of single or multi-lines heat sources (Table 2).
elements. The coordinates of single and multi-lines For the simulations, the basic heat transfer module of
heat sources are shown in Figure 3. COMSOL Multiphysics, which uses Fourier’s law, is
In order to get a better resolution of the temperature used, and the groundwater flow is imposed through
variations around the line source, close to it the mesh a homogeneous velocity field. The Backward Euler
is further refined along the length of line source and (Crank-Nicolson Scheme) time marching method with
also along the line on which the observation point are RMS error tolerance of 10−3 is applied and the max-
placed (at the depth of 25 m). imum time step interval set to 86400 s due to better
516
Table 2. Summary of the model setup for verification. approximately 1500 times smaller than the runtime of
numerical models. Note that, for the analytical solu-
Parameter Value tion, the computation time depends on the number
of observation point. It has the advantage that it can
Type of problem 3D reduce the calculation time as a function of the amount
Numerical method for heat Standard Galerkin-FEM
of required information.
transfer
Simulation time 160 days
Number of elements solved for 834,679/1,975,633
single BHE model/multi-BHEs 3.3 Multi-BHEs
model
Solver type Flexible Generalized Eq. 20 is solved on MATLAB and compared with
Minimal Residual method the numerical results. According to the results, again
the analytical method solution agrees with numer-
ical results both with (Figure 6) and without the
Table 3. Comparison the execution times and time steps for groundwater flow (Figure 7).
single BHE. The small discrepancy between the results of advec-
tion/dispersion case can be accounted for the mesh
Number of
discretization of the numerical simulation. Compari-
time stepa Runtime
Model (Total simulation) [s] a son of Figure 6 and Figure 7 shows that the maximum
temperature decrease in the ground is substantially
Analytical solution (Eq. 18) 2562 9b /13c reduced by the groundwater flow (from –12 K to –8 K
Numerical model 1 162 15986 in the simulated case).
no groundwater flow
Numerical model 2 162 16974
with groundwater flow of 4 SUSTAINABILITY AND RECOVERY
1 × 10−6 m/s ASPECTS
a
Hardware specifications: Intel, 4 core i-5 3.10 GHz, RAM:
The objective of this section is to evaluate the long-
16 GB.
b
Calculation for 5 observation points. term sustainability of the system and the energy
c
Calculation for 7 observation points (Figure 4 and Figure 5). deficit of the ground by comparing the temperature
distribution in the vicinity of a BHE and the heat
fluxes.
robustness. Table 2 provides a summary of the model
setups.
For verification plots, temperature changes are
observed in time on the x direction of the coordinate
system (Figure 3).
3.2 Single BHE

The Eq. 18 is solved on MATLAB and compared with
the numerical results. According to the results, the
analytical method solution agrees well with numer-
ical results both under conduction (Figure 4) and
advection/dispersion dominated heat transfer systems
(Figure 5). Comparing the results of temperature dif-
ference between conduction and advection/dispersion
heat transfer systems, the impact of the groundwater
flow/dispersion appears clearly. The first heat extrac-
tion phase generates larger temperature decrease in the
point located in the x-direction when groundwater flow
and dispersion are considered. However, the recovery
phase is accelerated due to water flow. Consequently,
the subsequent heat extraction induces lower tempera-
ture change in the ground close to the BHE. Also, with
time, temperature is impacted at larger distance in the
direction of the water flow when the groundwater flow
and the dispersion are considered (at 10 m, T = 0.8
K with water flow vs T = 0.2 K without water flow). Figure 3. Illustration of temperature observation points on
The significant advantage of analytical method can the x direction: a) for single BHE b) for multi-BHEs field.
be seen in Table 3. The execution time of Eq. 18 is Groundwater flow is set on the x direction.
517
Figure 4. Comparison of numerical and analytical solution Figure 7. Comparison of numerical and analytical solu-
results at the depth of 25 m for single line heat source without tion results at the depth of 25 m for multi-BHEs field under
groundwater flow. Induced by the load profile of Figure 2. groundwater flow of 1 × 10−6 m/s on the x-axis direction.
capacity is 2.5 MJ m−3 K−1 . The thermal properties

of the ground are set as an average value of soils,
which can be seen in shallow subsurface such as
clay or silica-sand, and the groundwater flow is not
considered.
In Figure 8, the relative temperature change results
are plotted (e.g. the temperature difference between
plume temperature ensuing from the operation and ini-
tial ground temperature) during the production and the
subsequent recovery phase of the ground. The results
show that even at 10 m distance from the BHE, the
Figure 5. Comparison of numerical and analytical solution mean temperature decrease is nearly 0.3 K after 30
results at the depth of 25 m for single line heat source. Under years of the ground temperature recovery.
groundwater flow of 1 × 10−6 m/s on the x-axis direction. Moreover, the bulk energy deficit of the ground
Induced by the load profile of Figure 2. demonstrated in Figure 9 is calculated based on the
volume integration method respect to the tempera-
ture change. During the production period, the Eq. 20
accounts for the balance between the energy extrac-
tion and the lateral heat flux in the ground and during
the recovery period only the lateral heat flux in the
ground is considered. The bulk energy deficit in the
ground can be calculated in axisymmetric conditions
as follows:
Figure 6. Comparison of numerical and analytical solution

results at the depth of 25 m for multi-BHEs field without in which H is the length of the BHE and we assume
groundwater flow. that the energy deficit is identical along the length, T
is the temperature change respect to time interval t,
The study is carried out with the developed ana- dR is the radial distance interval, and R is the radial
lytical solution on MATLAB. The scenario contains distance.
a production period of 30 years, and a subsequent Compared to the evaluation of the recovery phase
recovery phase, which is identical duration as the respect to the temperature gradient (Figure 8) and
production period. A constant continuous heat extrac- regarding to the energy balance in the ground (Fig-
tion of 10.27 W m−1 is applied along 30 years oper- ure 9), it can be seen that the justification based on the
ation period (the total amount of heat extraction local change of the ground temperature does not pro-
9000 kWh per year = 50 W m−1 × 100 m × 1800 h is vide a straight insight compared to the replenishment
distributed hourly for a single BHE with a length of of the bulk ground energy deficit Actually, the fast
100 m). The bulk thermal conductivity of the ground decline of the temperature deficit after the shutdown
is λm = 1.5 W m−1 K−1 , and the bulk volumetric heat of the system is not translated by such a rapid drop of
518
plume is obtained due to advection and dispersion pro-
cesses induced by the groundwater movement. The
new approach provides significantly shorter computa-
tion time compared to numerical simulation to obtain
the temperature results of a long-term production of
GSHP systems and subsequent recovery period.
The consideration of the temperature change in the
vicinity of a BHE does not give the direct insight in the
replenishment of the bulk ground energy deficit. By
taking into account the bulk vertical and the lateral heat
fluxes around the BHEs, the evaluation of the energy
recovery may be more realistic.
Figure 8. Temperature probe of the scenario. Temperature As a perspective, our analytical model can serve as
probes at the depth of 50 m from surface. a tool to predict the ground thermal evaluation around
the BHEs during the heat extraction/injection opera-
tions and in the subsequent recovery phase after the
GSHP system is shutdown. However, the limitations
of the model is that we did not take into account the
top surface and the bottom heat fluxes which may
accelerate the recovery process in long-term, and the
performance of the GSHP system may increase more
than we estimated here.
ACKNOWLEDGEMENT
The financial support from Walloon Region in Bel-

Figure 9. The bulk ground energy deficit of Scenario 4, gium is profoundly acknowledged (Grand: 1117492 –
basalt. GeoTherWal – Programme ERable (E24+)).
energy deficit. The total amount of extracted energy REFERENCES

by the BHE over the production period is around
9.72 × 1011 Joule (=10.27 W m−1 × 100 m × 30 years Bernier, M.A., Labib D.P.R & Pine P. 2004. “A Multiple Load
Aggregation Algorithm for Annual Hourly Simulations of
×3.15 × 107 s/year) and the net energy deficit in the GCHP Systems”, HVAC & R Res. 10 (4): 471–488.
ground after this 30 years of the production period is Carslaw, H.S., &. Jaeger J.C. 1959. Conduction of Heat in
around 2.7 × 1011 Joule, which means that approxi- Solids, Second Edition. US-NY: Oxford University Press.
mately 70% of the energy already recovered during Claesson, J & Eskilson P. 1988. “Conductive Heat Extraction
the production period by the ground lateral heat fluxes to a Deep Borehole: Thermal Analyses and Dimensioning
(see Figure 9). After subsequent 30 years of recovery, Rules.” Energy 13 (6): 509–527.
nearly 98 % of the extracted heat is recovered. Deerman, D.J. & Kavanaugh P.S. 1991. “Simulation of Ver-
In the near-field, the consideration of the tem- tical U-Tube Ground – Coupled Heat Pump Systems
perature change is plausible to know if the local Using the Cylindrical Heat Source Solution.” ASHRAE
Transactions 97 (1): 287–295.
temperature drops to the freezing point of the ground- Diao, N., Li Q., & Fang Z. 2004. “Heat Transfer in Ground
water, but the bulk energy deficit gives more global Heat Exchangers with Groundwater Advection.” Int. J. of
information about the energy recovery of the ground. Therm. Sci. 43 (14): 1203–1211.
Erol S., Hashemi M.A. & François B. (2015). Analytical solu-
tion of discontinuous heat extraction for sustainability and
recovery aspects of borehole heat exchangers. Int. J. of
5 CONCLUSION Thermal Sciences, 88, pp. 47–50.
Eskilson, P. 1987. “PhD Thesis: Thermal Analysis of Heat
Analytical solutions of GSHP systems are prefer- Extraction Boreholes.”, Lund: University of Lund.
able to have a better comprehension about the heat Gehlin, S. 2002. “PhD. Thesis: Thermal Response Test.”,
transfer system of the ground. Starting from the Luleå: Luleå University of Technology.
Green’s function, which is a solution of the con- Hecht-Méndez, J., De Paly M., Beck M. & Bayer P. 2013.
duction/advection/dispersion heat transfer in porous “Optimization of Energy Extraction for Vertical Closed-
media, we deduced an analytical solution that provides Loop Geothermal Systems Considering Groundwater
Flow.” Energy Convers. and Manag. 66: 1–10.
temperature distribution around single and multi- Hellström, G. 1991. “Ph.D. Thesis: Ground Heat Storage
BHEs for intermittent heat extraction or storage. Axial Thermal Analyses of Duct Storage Systems, I. Theory.”
effect and groundwater flow are considered. The pro- Lund: University of Lund.
posed analytical solutions are validated with numeri- Hopmans, J.W., Šimunek J. & Bristow K.L. 2002. “Indirect
cal code. Non-symmetric distribution of temperature Estimation of Soil Thermal Properties and Water Flux
519
Using Heat Pulse Probe Measurements: Geometry and Molina-Giraldo, N., Blum P., Zhu K., Bayer P., & Fang Z.
Dispersion Effects.” Water Resour. Res. 38 (1): 7–14. 2011. “A Moving Finite Line Source Model to Simulate
Kavanaugh, S. P., & Rafferty K. 2014. Geothermal Heat- Borehole Heat Exchangers with GroundwaterAdvection.”
ing and Cooling: Design of Ground-Source Heat Pump Int. J. of Therm. Sci. 50 (14): 2506–2513.
Systems. ASHRAE. Atlanta, US–GA. Rybach, L., & Eugster W.J.. 2010. “Sustainability Aspects of
Lamarche, L. 2009. “A Fast Algorithm for the Hourly Sim- Geothermal Heat Pump Operation, with Experience from
ulations of Ground-Source Heat Pumps Using Arbitrary Switzerland.” Geothermics 39 (4): 365–369.
Response Factors.” Renew. Energy 34 (10): 2252–2258. Sauty, J. P., Gringarten A.C., Fabris H., Thiery D., Men-
Man, Y., Yang H., Diao N., Liu J. & Fang Z. 2010. “A New joz A., & Landel P.A. 1982. “Sensible Energy Storage
Model and Analytical Solutions for Borehole and Pile in Aquifers: 2. Field Experiments and Comparison with
Ground Heat Exchangers.” Int. J. Heat Mass Transf. 53 Theoretical Results.” Water Resour. Res. 18 (2): 253–265.
(13–14): 2593–2601. Signorelli, S. 2004. “PhD Thesis: Geoscientific Investiga-
Marcotte, D., Pasquier P., Sheriff F., & Bernier M. 2010. tions for the Use of Shallow Low Enthalpy Systems.”,
“The Importance of Axial Effects for Borehole Design Zurich: Swiss Federal Institute of Technology.
of Geothermal Heat-Pump Systems.” Renew. Energy 35 Sutton, M. G., Nutter D.W., & Couvillion R.J.. 2003. “A
(4): 763–770. Ground Resistance for Vertical Bore Heat Exchangers
Marcotte, D., and Pasquier P. 2008. “Fast Fluid and Ground With Groundwater Flow.” J. of Energy Resour. Technol.
Temperature Computation for Geothermal Ground- 125 (3): 183.
Loop Heat Exchanger Systems.” Geothermics 37 (6): VDI-Richtlinie. 2000. “Thermal Use of the Underground –
651–665. Fundamentals, Approvals and Environmental Aspects,
Metzger, T. 2002. “PhD Thesis (in French): Dispersion VDI 4640 Blatt 1.” Düsseldorf: Verain Deutscher Inge-
Thermique En Milieux Poreux: Caractérisation Expéri- nieure, VDI-Verlag.
mentale Par Technique Inverse.”, Nancy, Institut national Yavuzturk, C. 1999. “PhD Thesis: Modeling of Vertical
polytechnique de Lorraine (INPL). Ground Loop Heat Exchangers for Ground Source Heat
Michopoulos, A., & Kyriakis N. 2009. “A New Energy Anal- Pump Sytems.” US-OK: Oklahoma State University.
ysis Tool for Ground Source Heat Pump Systems.” Energy Zeng, H. Y., N.R. Diao, & Fang Z.H.. 2002. “A Finite
and Buildings 41 (9): 937–941. Line-Source Model for Boreholes in Geothermal Heat
Michopoulos, A., & Kyriakis N. 2010. “The Influence of a Exchangers.” Heat Transfer Asian Research 31 (7):
Vertical Ground Heat Exchanger Length on the Electricity 558–567.
Consumption of the Heat Pumps.” Renew. Energy 35 (7):
1403–1407.
Molina-Giraldo, N., Bayer P., & Blum P.. 2011. “Evaluat-
ing the Influence of Thermal Dispersion on Tempera-
ture Plumes from Geothermal Systems Using Analytical
Solutions.” Int. J. of Therm. Sci. 50 (7): 1223–1231.
520
Use underground reservoir in Taipei basin as cooling source

for air conditioners
H.J. Liao
Department of Civil and Construction and Engineering, National Taiwan University of Science and Technology,
Taipei, Taiwan
Chihping Kuo
Department and Institute of Civil Engineering and Environmental Informatics,
Minghsin University of Science and Technology, Hsinchu, Taiwan
ABSTRACT: To use the underground reservoir in Taipei basin as the cooling source for air conditioners, a full
scale test was carried out in the campus of NTUST to cool down a 15 RT (Refrigeration Tons) air conditioner. Two
types of circulating water system were adopted: one is open system; the other is closed system. The open system
pumps up groundwater directly from the underground reservoir. It offers a constant temperature cooling source
for air conditioners (AC units). After doing heat exchange with AC units, the heated groundwater is discharged
back to the reservoir through an open well. Test results showed that the open system has a high cooling capacity
and was capable of keeping a 15 RT AC unit running continuously. Since no groundwater was pumped out
from the reservoir, ground subsidence was no concern for the open system. In comparison, the closed system
discharges the exhaust heat from the AC units through a closed loop pipe which is submerged in the groundwater
inside a well. The heat is transmitted to the underground reservoir by means of the circulating water in the pipe
loop. However, the heat transmission rate of the water saturated ground is limited and it was unable to dissipate
all the exhaust heat from the 15 RT AC unit. As a result, the heat quickly built up around the pipe loop. The
AC unit was shut down after running for only a few hours due to overheat. Obviously, the cooling capacity of
the closed system is much lower than that of the same well system used by the open system. The groundwater
pumping rate of the open system could also be adjusted using a PLC unit based on the actual cooling need of AC
unit to further cut down the running cost of water circulation. Numerical simulation also confirms that the heat
exchange rate of the closed system is much lower than the open system even though the flow rate of groundwater
is capable to carry the same amount of heat away.
1 INTRODUCTION accumulated in the basin where Taipei city locates.

As a result, the air temperature is increased and makes
1.1 Geographical and climatic conditions the heat island effect more serious (Hsieh et al. 2007).
Taipei Basin, located in the northernTaiwan and in sub-
tropical area, covers approximately an area of 243 km2
1.2 Geological and groundwater conditions
with the average elevation about 20 m above sea level.
It is surrounded by Datun volcanoes in the north; Generally, sediments deposited from different rivers
by Linkou tableland in the west; and by hills and are various: soft clays deposited from Keelung river;
mountains in the east and the south. Several major gravels deposited from Xindian river; and sandy layers
rivers meander through the Taipei basin, namely Tan- deposited from Dahan river. Basically, the sandy and
shui river, Keelung river, Xindian river and Dahan clayey deposits in Taipei basin are underlaid by gravel
river. Taipei city locates right in the Taipei basin. The deposit. The thickness of gravel stratum increases
weather in summer is hot and humid. Over the past gradually from Xindian in the southeast to Tansui in
decades, the mean daily surface temperature of Taipei the northwest (Figure 1).
city had been rising from 28.0◦ C in 1960 to 29.5◦ C As estimated by Chen (2005), there is about 68.4
in 2005 (Chen et al. 2007). In addition, the num- billion m3 of groundwater stored in the gravel stra-
ber of days with temperature above 35◦ C is increased tum of Taipei basin. Kuo and Liao (2011) studied
also. Air-conditioners have become a necessity to cool the directions and quantities of the groundwater flow
down the indoors temperature in summer. Among the in the gravel stratum using the numerical software
air conditioners used, a large majority are air-cooled MODFLOW. The direction of groundwater flow was
type. In other words, most of the heat exhausted from indicated by the vectors in Figure 2. The average quan-
air-conditioners is discharged to the open air and tity of water budget is about 3.5 × 104 3 /day (cmd) for
521
Figure 1. Schematic diagram of the basin profile.
Figure 3. Closed type and open type cooling water circula-

tion.
The feasibility of using open system in Chingmei

gravel stratum has been verified through a field pump-
ing and recharging test and a numerical simulation
(Kuo and Liao 2012). These results were adopted and
Figure 2. Schematic diagram of the basin profile. implemented to the full scale groundwater circulation
test in a power transform station in the campus of
National Taiwan University of Science and Technol-
the entire basin. The value of water budget in the study ogy (NTUST). Both open type and closed type systems
area can be converted into the groundwater flux, i.e., (Figure 3) were studied in full scale and the test results
groundwater velocity, of 13.0 cm/day. The measured are presented in this paper.
temperature of the groundwater changes from 24◦ C in
winter to 26◦ C in summer.
2 HYDROLOGICAL CONDITIONS IN NTUST
1.3 The alternative method 2.1 Location
Since the groundwater resource is rich in the basin, NTUST locates at the southeast side of Taipei basin.As
an alternative method for the cooling of AC units is shown in figure 2, flowing direction of groundwater is
proposed here. To mitigate the above mentioned heat generally from south to north at the NTUST site.
island problem, the exhaust heat from the AC units
can be discharged to the underground reservoir in
2.2 Underground soils profile and flowing of
the gravel stratum underlying Taipei city by means of
groundwater
circulating groundwater/water system. Two types of
groundwater cooling system, namely the open system As shown in figure 4, the gravel stratum underneath the
and the closed system, can be used. The open sys- NTUST site was located 39∼60 m below ground sur-
tem consists of a group of integrated deep wells which face. Inside the gravel stratum, there was a varve clay
pump up cool groundwater and provide a steady cool- layer (Wuku layer) of 4 m thick (GL −46∼ −50 m)
ing source for ACs. After doing heat exchange with dividing the gravel layer into upper layer (Ching-
the ACs, the heated water was discharged back to mei stratum) and lower layer (Banciao stratum). The
the reservoir. During the circulation process, under- groundwater levels of the upper and lower layers were
ground reservoir only acts as a heat sink. All the located at GL −12.5 m and GL −10 m respectively
groundwater being pumped up is recharged back to the during the period of testing. Pumping test were per-
underground reservoir immediately. So, no pumping formed to estimate the transmissive parameter (T) and
induced ground subsidence is expected. In compari- it was about equal to 0.0003 m2 /sec. As shown in Fig-
son, the closed system brings the heat from AC units ure 2, the direction of groundwater flow were measured
down to the underground reservoir through the circu- and matched the results from numerical simulation. Its
lating water in a closed loop pipe. No groundwater is velocity also measured and is equal to 0.005 m/sec.The
pumped up or discharged in the closed system. temperature of the groundwater is about 23.5◦ C.
522
to get the cool from the groundwater without let-
ting the groundwater flow into the air conditioner.
The PHE consisted of 17 heat exchange plates with
the heat exchange area of 1.92 m2 . The heat load of
this PHE unit was ∼80 kW. It was more than the
nominal heat exchange needed for this 15 RT air
conditioner (cooling capacity = 24 kWh). Inside the
PHE unit, the heated water from AC was separated
from the cool groundwater coming from the well
by corrugated stainless steel plates. As the heated
water flowing through the PHE unit, it exchanges heat
with cool groundwater inside the PHE. After that,
the exhaust heat from AC was brought down to the
gravel stratum by the circulating water. Two types of
circulation system were tested here: open system circu-
lates groundwater between aquifer and PHE via wells;
closed type circulation system circulates cooling water
in a closed loop pipe installed in the wells.
4.2 Open system

Two PVC standing pipes with a diameter of 254 mm
were installed in the well. One had openings at GL
−42∼−47 m (the upper well); the other had open-
Figure 4. Aquifer layers under the NTUST test site.
ings at GL −51∼−58 m (the lower well) (Figure 5).
Upper well works as the pumping well while lower
3 METHODOLOGY well works as the discharging well. A submergible
pump of 5 hp (horse power) was installed at 36.5 m
Both full-scale tests and numerical simulation of below ground surface in the upper well to pump up
closed type and open type cooling water circulation groundwater for the PHE unit. A rise pipe of 13 mm
were performed and compared in this study. diameter was connected to the submergible pump. The
rate of groundwater pumped up from the well was con-
trolled by the programmable logic controller (PLC)
and a rotation speed adjustable pump. The pumping
4 FIELD TEST AT NTUST SITE
rate was adjusted based on the measured temperature
of the cooling water used to cool down the compres-
4.1 Heat generating and exchanging devices
sors of AC. If the measured temperature was too high,
The power transform station in NTUST covers a floor the pumping rate was increased to pump up more cool
area of 124 m2 . Without air conditioning (ie., with only water. The flow rate of circulating water was measured
natural air ventilation), the typical indoor temperature with a GF Signet paddle-wheel flow sensor. The read-
rise in the station room was about 5◦ C over a 24 hour ings from the flow sensor were sent to the PLC to
period in summer. adjust the flow rate from the well. The water level in the
To control the room temperature, a 15 RT (Refriger- well was measured with vibration wire type pressure
ation Tons) water-cooled air conditioner was installed. transducers. The electricity consumption of AC and
The air conditioner had a nominal cooling capacity of submergible pump was measured using the electrical
82000 BTU/hr (=24 kWh). It equipped with two scroll Watt meter.
type compressors. The nominal power consumption Figure 6 shows the measured temperatures from
of the compressor was 11A each at full load (=22A various thermometers during a full day running period
when both compressors are operating). The compres- of the air conditioner. The groundwater pumped up
sors operated in two modes: one compressor running from the gravel stratum maintained a constant temper-
and both compressors running. The nominal cooling ature of 23.5◦ C. The indoors temperature of the power
water needed for this air conditioner was 196 liter/min transform station was maintained at a constant value
and the nominal operation temperature of the cooling of 17.5±0.5◦ C despite the outdoors temperature vary-
water was 37◦ C out from and 30∼32◦ C into the air ing from 20◦ C in the early morning to around 30◦ C at
conditioner. Any increase in the temperature of cool- noon. The air temperature measured at the air outlet of
ing water will decrease the cooling efficiency of the AC varied from 12.5 to 16.5◦ C. Occasionally (say, 10
compressor. times over a 24-hour period), the air temperature from
Because the manufacturer of air conditioner did AC dropped to 7∼8◦ C. It is an indication that AC was
not recommend using untreated groundwater to cool in the mode of both compressors running. However,
down the compressors directly. A plate heat exchanger the time period for both compressors running was very
(PHE) unit was used here acting as a buffer unit short. For most of the time, only one compressor was
523
Figure 6. Temperatures measured from different sensors
during a full day operation.
Figure 5. Aquifer layers under NTUST campus.
running. In other words, the AC was able to operate in

an energy efficient mode under this open groundwater
circulation system and consumed less electricity.
Figure 6 also shows that the cooling water tempera-
ture on the AC side of the plate heat exchanger (PHE)
varies from 32∼33◦ C (to AC) to 33∼34◦ C (from AC).
It indicates that the compressors of AC were operat-
ing at a temperature of about 33◦ C compared with the
nominal operating temperature of around 37◦ C when
used with cooling tower system. Meanwhile, the tem-
perature of circulating groundwater to the PHE unit
on the well side maintained a steady 23.5◦ C (from the
well).
The flow rate of the circulating groundwater was
controlled by the PLC unit. Although the flow rate
jumped up and down between 50 and 200 liter/min
Figure 7. Schematic diagram of closed circulating water
in Figure 6, the accumulated flow volume provided to system for AC.
the PHE unit was steady. Compared with the nominal
flow rate (=196 liter/min) required when a cooling
4.3 Closed system
tower is used, the accumulated flow volume of cir-
culating groundwater was much lower than that of Unlike the open system, no groundwater is pumped up
cooling tower [∼120000 liters vs. 282240 liters per from the gravel stratum in the closed system to cool
day (=1440 min * 196l pm)]. It indicates that 23.5◦ C down the AC. Instead, water circulating in the closed
groundwater is capable to keep a 15RT air conditioner system is in a closed pipe (Figure 7). It exchanges
running for a full day using only about 40% of cooling the heat with the surrounding groundwater through
water compared with that of cooling tower. the outer wall of a double pipe. As mentioned earlier,
524
the test well was divided into upper well and lower
well. A 40 m long double pipe with an inner diameter
of 10 cm was installed in the upper well; while a 60 m
long double pipe with an inner diameter of 12.7 cm was
in the lower well. The double pipes in the upper and
lower wells were connected in series manner. Seamless
steel pipe was used for the outer pipe and a PVC pipe of
5 cm in diameter was used as the inner pipe. PVC pipe
was used because it has a better heat isolation property
to minimize the heat exchange inside the double pipe.
Outside the double pipe, there was a PVC pipe of 20 cm
in diameter. It was perforated at the depth of GL −41
to −45 m and GL −52 to −60 m to allow groundwater Figure 8. Temperature curves and heat exchange curves of
flow in and out of the PVC pipe easily. The annular AC for the closed system tested here (AC is off but the water
space between PVC pipe and well wall was back filled circulation pump is on during the cool down period).
with gravel. Only the submerged length of the double
pipe in the well was counted as the effective length for
heat exchange between heated circulating water and
ground/groundwater. At the time of the experiment,
the groundwater water level was at GL −25 m. So the
submerged length (the effective length) of the double
pipe was about total to 50 m.
The schematic diagram for the closed type system
is shown in Figure 7. The same PHE unit as the open
system was used. A pump with 3 hp and flow rate
of 150 liter/min was installed on the well side of the
PHE unit to circulate the water in the closed system to
and fro the ground. No submergible pump was used
Figure 9. Water temperature in and out from the PHE
in the closed system. The heated water flowing out unit (during the cool down period both the AC and water
from the PHE unit was pumped to the PVC inner pipe circulation pump were off).
of the double pipe first and back from the annular space
between PVC pipe and outer steel pipe. On its way more than the heat exchange capacity of the closed
back to PHE unit, the heated water exchanged the heat system, the water temperature will keep increasing
with ground/groundwater through the outer steel pipe. until the AC automatically shut down for safety reason;
The heat load of the PHE unit used here was equal otherwise, the water temperature will only increase to
to 70∼79 kW. But the heat exchange capacity of 50 m a certain value and then remains there.
long submerged pipe was to be determined from this
experiment.
As shown in Figure 8, the water temperatures on the 5 NUMERICAL SIMULATION
AC side and the well side of the PHE unit increased
quickly when AC was running. Obviously, the heat Based on the measured local groundwater flow, a
generated from the AC and the water pump was beyond heat transport model established from the SHEMAT
the heat exchange capacity of this closed circulation (Simulator for HEat and MAss Transport, Clauser
system. After running for about 4 hours, the tempera- 2003) program was used to simulate the heat trans-
ture of the cooling water on the AC side of the PHE unit mission behavior which resulted from two types of
reached about 53◦ C; while the temperature of circulat- circulating water system for AC. 3-D ms consist-
ing water on the well side reached about 46◦ C. The AC ing of 144 × 142 × 3 cells were used. In the cen-
was shut down automatically to protect the compressor ter area (2000 m × 2000 m), each cell represented an
from overheating. After AC shut down, the water circu- area of 1 m × 1 m; outside the center area, each cell
lation pump on the well side stayed running to keep the represented an area of 10 m × 10 m or 20 m × 20 m
water in the closed pipe flowing. Under this circum- depending on the locations. The upper and lower
stance, it took 16.5 hours for the water temperature layers were treated as an aquitard, and the middle
drop back to 30◦ C. But the temperature was unable to layer was a confined aquifer. However, SHEMAT
drop back to the initial 23.5◦ C (i.e., the initial water can only be used in a layer of constant thickness for
temperature when the test started). This was caused the current version. So, a mean value of 30 m was
by the heat generated from the operating water pump. adopted to approximate the thickness of the aquifer.
However, if the water circulation pump was shut down To simulate the flowing groundwater condition in the
and no more heat contributed to the closed system, the Chingmei stratum with the SHEMAT program, a line
temperature was able to drop back to 23.5◦ C after 24 of virtual pumping wells was placed at the effluent
hours (Figure 9). From the phenomenon observed, it boundary of the study area to provide a groundwa-
can be concluded that if the generated heat from AC is ter flow with velocity = 0.005 m/sec. Meanwhile, a
525
line of virtually discharging wells were placed at
the inflow boundary. The parameters used in this
model were as follows: effective porosity = 0.25 and
conductivity = 102 cm/sec (Freeze & Cherry 1979),
thermal capacity = 1.875 MJ/m3◦ K and conductivity
= 1.308 W/m◦ K (Tindall & Kunkel 1999).
5.1 Open system

A steady recharging well with 200 liter/min and 37◦ C
were placed in the central point (1000 m, 1000 m)
in the simulated model. The temperature change and
distribution is shown in Figure 10. At the first four
hours, the heat diffuses mainly radially around the
well. The influence of groundwater flow is not signif-
icant yet at this stage. After 24 hours, the heat center
still keeps diffusing radially around the well but some
heat has started to dissipate with the groundwater flow
to the downstream. After 168 hours (=1 week), the
heat is mainly dissipating along with the groundwa-
ter flow. After 240 hours (=10 days), the dissipating
heat is carried away mainly by the groundwater flow
to the downstream. Very little heat is transmitted lat-
erally to the direction other than the groundwater flow
direction.
The long-term simulation shows the same distribu-
tion pattern. The heat dissipation maintains steadily
toward the downstream of groundwater flow. In other
words, the exhausted heat discharging from the open
system will be carried away by the groundwater flow
and will not accumulate around the well. However,
due to the limitation of boundary setting, the distribu-
tion of exhaust heat extends to a distance of 2 km or
more. But it is understood that the exhaust heat from
the discharging well will dissipate to the surround-
ing water body as the heated water flows downstream
with groundwater flow. In fact, the numerical results
also show that the heat dissipation in vertical direc-
tion (depth direction) can be as far as 20 m. So, the
heat dissipation distance estimated from the numer-
ical analysis can only be treated as the upper bound
cases for the open system.
5.2 Closed system

Closed system means the well will not recharge water
into aquifer. Due to the limit of the numerical software,
the exhausted heat has to be transferred into equivalent
discharging well with steady flow in actual tempera-
ture. From Figure 8, the heat generation rate can be
estimated to be 0.25 kW/min (=50 kW/200 min). That
equals a steady flow with 7.3 liter/min and a 29.5◦ C Figure 10. 2-D view of simulated short-term to long-term
temperature difference. temperature change and distribution for the open system, the
Therefore, a steady discharging well with 7.3 axis presents distance in meter.
liter/min and 53◦ C were placed in the central point
(1000 m, 1000 m) in the simulated model. Without dis-
charging the water to the reservoir, the diffusion of heat to aquifer is much less. It means that most of the heat
(mainly by heat transmission) is very limited. At the is trapped inside the well and was unable to dissipate
first four hours, the heat diffuses radially and accumu- to the reservoir. In fact, it took up to 24 hours for
lates quickly around the well. Comparing to the open the exhaust heat to fully dissipate to restart the closed
system, the quantity of heat exhausted from this well system again.
526
following conclusions can be drawn from the results
of this test.
1. The 23.5◦ C groundwater pumped up from the well
in the open system can provide a steady cooling
source for a 15 RT AC unit running all day long.
Using the same well, the heat discharged from the
closed system could not keep up with the heat gen-
erated from the AC. The water temperature in the
closed system kept increasing and caused the AC
to shut down due to overheat. The open system can
provide a much better cooling capacity to the AC
than the closed system
2. The submergible water pump in the open system
accounts for a large portion of power consumption.
By adjusting the pumping rate using a PLC unit
based on the cooling need of AC can reduce the
overall power consumption of the system.
3. Due to the slow heat exchange rate between
heated water and surrounding ground/groundwater
at NTUST campus, the cooling capacity of the
closed system is much lower than that of the same
well system used by the open system. An optimal
operation pattern for a closed system and its match-
ing airconditioners can be established in advance
through the temperature curve of circulation water
obtained from the field test or numerical analysis.
4. For the open system, the mineral contents of the
groundwater may cause scale or contamination
problem inside the PHE unit and the circulation
pipe. But it has a better heat dissipation capacity.
In comparison, water in the closed system has no
direct contact with groundwater. No contamination
in the pipeline has been experienced over the period
of this experiment.
REFERENCES
Chen, W.F. 2005.Groundwater in Taiwan. Taipei: Sinobooks.
(in Chinese)
Chen, T.C., Wang, S.Y., and Yen M.C. 2007. Enhancement
of Afternoon Thunderstorm Activity by Urbanization
in a Valley: Taipei”, Journal of Applied Meteorology
and Climatology, American Meteorological Society, 46,
1324–1340.
Clauser, Christoph (Ed.), Numerical Simulation of Reactive
Flow in Hot Aquifers, SHEMAT and Processing SHEMAT
Freeze, R.A. & Cherry, J.A. 1979. Groundwater. Prentice-
Hall. Englewood Cliffs NJ.
Hsieh, C.M., Aramakia, T, & Hanakia, K. 2007. Estimation
of heat rejection based on the air-conditioner use time and
its mitigation from buildings in Taipei City. Building and
Environment, 42, 3125–3137.
Figure 11. 2-D view of simulated short-term temperature Kuo, C.P. & Liao, H.J. 2012. The feasibility of using cir-
change and distribution for the closed type circulation system, culating groundwater as renewable energy sources for
the axis presents distance in meter. air-conditioning in Taipei basin. Renewable Energy 39 (1),
175–182.
Tindall, J.A. & Kunkel, J.R. 1999.Unsaturated Zone Hydrol-
ogy for Scientists and Engineers. Prentice-Hall, London.
6 CONCLUSIONS AND SUGGESTIONS Tsao,Y.S., Lin, C.N., Tan,Y.C. & Mao, A.S. 1985. Simulation
and application of mathematical groundwater model in
A full scale test on two types of cooling water circu- Taipei Basin. Ministry of Economic Affairs. Taipei. (in
lation system (open system and closed system) was Chinese)
carried out to cool down a 15 RT (Refrigeration Wang, C.C. 2007. Heat Exchange Design. Wu-Nan Books
Tons) air conditioner in the campus of NTUST. The Company. (in Chinese)
527
Cyclic loading effects on soil-energy pile interaction
S. Xiao, M.T. Suleiman & R. Elzeiny

Lehigh University, Bethlehem, USA
M.J. Al-Khawaja
Qatar University, Doha, Qatar
ABSTRACT: To reduce greenhouse gas emission from space heating and cooling, energy piles have been used
in buildings as an alternative renewable energy source for approximate by two decades. However, the effects of
cyclic thermal loading, due to the intermittent operation of the heat pump, on the soil-energy pile interaction have
not been fully investigated. Energy piles are subjected to cyclic temperature changes that affect the properties
of soil-structure (or pile) interface. In addition, the effects of temperature variations produce cyclic expansion
and contraction of the pile. To evaluate the effects of radial expansion/contraction cycles, a fully controlled
thermal-modified borehole shear test (Thermal-mBST) device was developed at Lehigh University to measure
the thermo-mechanical behavior of the soil-energy pile interface. A pair of concrete plates, representing the
pile surface, were used. The plates have embedded aluminum small diameter pipes that are connected to a heat
pump to control the temperature of the Thermal-mBST. Two linear potentiometers were fixed between the two
concrete plates to control/measure the horizontal displacement between the two plates. The testing system is
capable of simulating temperature change and cycles, expansion/contraction (displacement) change and cycles,
as well as the combination of temperature and expansion/contraction cycles. In this paper, the Thermal-mBST
device was utilized to conduct tests with temperature changes (T) of 0 and +20◦ C at Soil-Concrete Interface
(SCI) under initial normal pressure of 41.4 kPa utilizing an aluminum tank filled with silty clay. The tests were
performed under heating cycles, expansion cycles, and combined heating and expansion cycles. After applying
the cycles, the shearing stage was performed to measure the stress-displacement curves of the soil-pile interface
(t-z curves).
1 INTRODUCTION (2015). Suleiman & Xiao (2014) and Xiao & Suleiman
(2015) who utilized Thermal-mBST reported the
Shallow geothermal energy, which is one of the effects of the monotonic thermal and displacement
promising renewable energy sources, takes advantage cycle on soil-concrete interface properties, at temper-
of the nearly constant and moderate temperature of atures of 2 to 40◦ C. The results showed that expan-
the ground to heat and cool buildings. The Ground sion/contraction has significant effect on the shear
Source Heat Pump (GSHP) exchanges heat with the resistance in normally-consolidated unsaturated soil.
ground through a ground heat exchanger.As a develop- Murphy & McCartney (2014) performed thermal-
ment from conventional GSHP systems, energy piles borehole shear tests in Boulder clay and silty sand
are used to support the superstructure load and also considering only the effects of the temperature rang-
function as ground heat exchangers for heating and ing from 10 to 45◦ C. Short time temperature on the
cooling of buildings (Brandl 2006). Several published interface did not show major impact on the shape
full-scale tests investigated this aspect of energy piles of the normalized T-z curve. Donna et al. (2015)
and showed thermally induced deformation and forces performed soil-concrete interface tests using a mod-
in the foundation element (Brandl 2006, Laloui et al. ified direct shear device with dry sand and saturated
2006, Bourne-Webb et al. 2009, McCartney & Mur- clay in isothermal condition. The testing temperature
phy 2012, Suryatriyastuti et al. 2012 Akrouch et al. ranged from 22 to 60◦ C. The results showed that the
2014 and You et al. 2016). The intermittent operation sand-concrete interface behavior is not affected by the
of the heat pump presents new challenges to founda- temperature changes at constant normal load nor con-
tion engineers, one of which is cyclic thermal loading stant normal stiffness conditions. The tests with clay
effects on soil-energy pile interaction. showed the shear strength increased with increasing
The effects of temperature changes on soil- temperature due to the effects of temperature on clay
structure interface properties were investigated by deformation.
Suleiman & Xiao (2014), Murphy & McCartney The change of temperature leads to expansion and
(2014), Xiao & Suleiman (2015), and Donna et al. contraction of the pile altering the normal pressure
529
at the soil-pile interface and leading to changes of
the soil-pile interface properties (i.e., shaft resis-
tance). Energy pile expands in both axial and radial
directions when subjected to heating, and contracts
when subjected to cooling. Radial expansion and
contraction of energy piles were studied by Sury-
atriyastuti et al. (2012), Suleiman & Xiao (2014),
Xiao & Suleiman (2015), Olgun et al. (2014), and
Wang et al. (2015). However, the effects of temper-
ature cycles and expansion/contraction cycles on the
soil-pile interface properties have not been fully inves-
tigated. In this paper, the Thermal-mBST device was
utilized to evaluate the effects of cyclic thermal loading
and expansion/contraction cycles on soil and soil-pile
interface mechanical properties.
2 TESTING APPARATUS
To investigate the effects of temperature and expan-

sion/contraction cycles on the soil-energy pile inter-
face properties, a device with advanced capabilities
was used in this research (Figure 1). This device,
which is called Thermal-mBST, was developed based
on Handy & Fox (1967), Suleiman & Xiao (2014),
Murphy & McCartney (2014), and Xiao & Suleiman
(2015). Compared to the device used in Xiao &
Suleiman (2015), this improved device included the
following modifications: (1) the concrete shearing
plates were embedded with small diameter aluminum
pipes which connect to a heat pump to apply heating
and cooling, and insulation was used to reduce the con-
densation effect on the water content of surrounding
soil; (2) thermocouples were installed in the soil and
on the surface of the concrete shear plate to measure
temperatures at the soil-concrete interface (3) two lin-
ear potentiometers were installed between the shear
plates to measure the horizontal displacement, and air
pressure in the pneumatic piston can be adjusted to
control displacements representing the expansion and
contraction of energy piles; (4) the testing device was
automated, and loads and displacements were fully Figure 1. Experimental apparatus of the Thermal-mBST:
(a) configuration of the step-up; and (b) shear head.
controlled by a control system during the test. These
modifications allow for assigning thermal loading on
the concrete shearing plates, measuring the temper- measured by a load cell. A pressure bag was placed on
ature at the soil-concrete interface and temperature top of the soil to apply 68.9 kPa overburden pressure
distribution in soil, and controlling the horizontal dis- to consolidate the soil and simulate larger overburden
placement between the plates to simulate the radial pressure in the soil.
expansion and contraction of energy piles.
The Thermal-mBST was conducted in the labo-
ratory inside a soil tank with 470 mm in diameter 3 SOIL MATERIAL AND PREPARATION
and 610 mm height as shown in Figure 1. The soil
borehole at center of the tank where the shear head Soil obtained from a construction site in Lehigh Valley,
was placed is 95 mm in diameter. The shear head of Pennsylvania USA, was used in the Thermal-mBSTs.
the Thermal-mBST includes a pneumatic piston sand- The soil particle size distribution curve is shown in
wiched between two concrete plates as shown in Figure Figure 2. Using the Unified Soil Classification System
1b. Different normal stresses on the soil-concrete inter- (USCS), the soil was classified as silty clay with fines
face can be produced by changing the air pressure of content of 80%. The liquid and plastic limits of the
the pneumatic piston. The dimensions of the shear soil were 28% and 22%, respectively. The maximum
plate are 51 by 102 mm. Linear variable differential dry unit weight and optimum moisture content deter-
transformers (LVDTs) were used to measure the verti- mined using Standard Proctor tests were 17.4 kN/m3
cal movement of the shear head. The pulling force was and 14%, respectively. To perform the Thermal-mBST
530
Figure 2. Particle size distribution of the soil.
test inside the soil tank, the soil was prepared with a
target moisture content of 18%, dry unit weight of Figure 3. Definition of displacement and temperature
13.7 kN/m3 , and degree of saturation of the soil is cycles.
54%. The measured thermal conductivity of the soil
at the same moisture content is ∼1.1 W/mK, and the The normal pressure can be reduced by the con-
volumetric heat capacity is ∼ 2042 kJ/m3 K. troller to make the plates return to its original position
The soil was placed in the tank in 5 layers, each which is 1 displacement cycle as shown in Figure
layer was compacted by 25 blows using a steel tamper 3. The system also can combine the displacement
to achieve the target dry unit weight. A steel tube with and temperature cycles at the same time by con-
a diameter of 95 cm was embedded in the soil along trolling the heat pumps and normal pressure at the
the center of the tank to form a hole through the depth. soil-concrete interface. To separate the temperature
After compaction, the tube was pulled out to allow for cycles and expansion/contraction cycles effects on
lowering the shear head of the Thermal-mBST inside the soil-concrete interaction, the Thermal-mBSTs can
the created hole. An air bag was placed on top of the be conducted with expansion/contraction cycles and
soil surface and inflated to apply a vertical pressure temperature cycles separately.
of 69 kPa simulating overburden pressure, and the soil After applying the test condition (e.g. displace-
was allowed to consolidate for 48 hours. ment cycle), the shearing stage starts. The interface
is sheared with a constant shearing rate of 0.05 mm/s
which is similar to the conventional borehole shear
4 TEST PROCEDURES test (Lutenegger, 1987). Stress vs. displacement rela-
tion (t-z curve) can be obtained through the results of
The shear head was lowered in the hole at the center of the load cell and LVDTs during shearing.
the soil tank, a target normal horizontal pressure was
applied for 12 hours (consolidation stage). In the test
with temperature cycles only, the plates were heated 5 RESUSLTS AND ANALYSIS
or cooled by circulating fluid with different temper-
atures. During the heating cycles, one heat pump is A reference test was conducted without temperature
used and the heat pump setting temperature is 5 to 8◦ C cycle or displacement cycle (D = 0, T = 0) with
higher than the target temperature of the soil-concrete normal stress of 41.4 kPa. The peak stress of t-z curve
interface to account for the heat loss. Once the target for this reference test is 35.8 kPa. The secant shear
temperature of soil-concrete interface is achieved (i.e., modulus E50 is 111.6 kPa/mm.
0.5 temperature cycle as shown in Figure 3), the heat
pump stops functioning, and the other heat pump starts
5.1 Effect of temperature change and cycles
to cool the temperature of the plates down to the room
temperature (i.e. 1 temperature cycle) with the setting Figure 4 shows the t-z curves after 0.5, and 10.5 heat-
temperature of 2◦ C lower than the room temperature. ing cycles under normal stress of 41.4 kPa. With 0.5
In the displacement cycles only, the control sys- heating cycle (T = 0.5), two tests were performed,
tem adjusts the air pressure of the pneumatic piston to one is continuously heating for 0.5 hour, and the other
change the normal pressure at the soil-concrete inter- one is continuously heating for 10.5 hours. For 0.5
face. The piston pushes the plates outward toward the hour’s heating test, the peak shear strength decreased
surrounding soil when the normal pressure is increas- by 9% compared to the reference test. This may be
ing, and the plates move backward when the normal attributed to the undrained heating, where the time for
pressure is decreasing. The expansion to target dis- the dissipation of thermall-induced excess pore water
placement is defined as 0.5 cycle of displacement. pressures was not sufficient (Murphy and McCartney,
531
Figure 5. Horizontal displacement during the heating cycle
with constant normal stress of 41.4 kPa.
Figure 4. Heating effects on t-z curves of SCI.
2014). For the 10.5 hours’ heating test, the shearing

behavior of the soil-concrete interface showed strain-
softening response. The peak shear strength is almost
the same as the reference test. However, the residual
shear strength is 15% lower than the reference test.
The peak strength increased after 10.5 hours’ heating
may be caused by drained heating and increasing suc-
tion where the water content deceased from 18% to
14.2% near the soil-concrete interface. After 10.5 heat-
ing cycles, the peak shear strength is 10% lower than
the reference test, and residual strength is 15% lower
than the reference test, which may also be attributed
to the process of undrained heating.
Figure 5 shows the displacement change between Figure 6. Temperature distribution of the soil during the
the two shear plates during heating cycles with con- heating cycles.
stant normal stress during each temperature cycle that
took one hour. After 10.5 heating cycle, the horizon- heating cycles increases. At the end of the 10.5 heating
tal distance change between the two plates is 489 µm. cycles, the temperature active zone is approximately
The horizontal displacement during the heating cycle 75 mm.
indicated a thermal-softening behavior of the soil and
thermal-induced settlement. The rate of the thermal-
5.2 Effect of expansion and cycles
softening behavior decreases with number of cycles.
The horizontal displacement increases with the num- Figure 7 shows the effects of expansion cycles on
ber of heating cycles in this test. Water migration also the soil-concrete properties. After applying the expan-
was observed in the test. Water contents of 17.4%, sion cycles, the interface was sheared at the expansion
15.8%, and 14.2% were measured after 0.5 hour’s position in the tests with 0.5 and 10.5 expansion
heating, 10.5 hours’ heating, and 10.5 heating cycles, cycles. When compared to the reference test (D = 0,
separately. Water migration followed by suction vari- T = 0), the peak strength has 21.8% increase in the
ation in the soil may affect the soil-concrete interface test with 0.5 expansion cycle (D = 0.5, T = 0) and
properties (Hamid & Miller 2009). The displacement 12.3% increase in the test of 10.5 expansion cycles
change of the plates is around 847 µm after 10.5 hours’ (D = 10.5, T = 0). In the 10 expansion cycle test
continuous heating. Those temperature-induced soil (D = 10, T = 0), the interface was sheared when
settlement changed the relative density of the soil the plates returned back to the initial position, the
which may affect soil-concrete interaction (DeJong reduction of the peak strength of the interface is 73.6%
et al. 2009). compared to the reference test. This may be attributed
Figure 6 shows the temperature distribution at the to the permanent soil deformation occurred during
soil-concrete interface and in the surrounding soil for the expansion cycles. Figure 8 shows the relationship
the 10.5 heating cycles test. As shown in the figure, between normal stresses and displacement in the test
the temperature distribution is uniform at the begin- with 10.5 expansion cycles. In this test, the plates
ning (T = 0) in the soil. The temperature active zone, expand and return back and forth between the ini-
where the soil experiences temperature changes, is tial and target expansion positions. The corresponding
approximately 50 mm for the 0.5 and 1 heating cycles. normal stresses at target expansion position decreased
The temperature active zone increases as the number of with number of cycles which is shown in Figure 8.
532
Figure 7. Expansion effects on the t-z curves of SCI. Figure 9. Combination of expansion and heating effects on
t-z curves with INS of 41.4 kPa.
6 CONCLUSION
Shallow geothermal energy, which is one of

the promising renewable energy sources, utilizes the
nearly constant and moderate temperature of the
ground to heat and cool buildings. The ground source
heat pump (GSHP) exchanges heat with the ground
through a ground heat exchanger. As a development
from conventional GSHP systems, energy piles can
be used to support the superstructure load and also
function as ground heat exchangers for heating and
cooling of buildings. The intermittent operation of the
heat pump presents new challenges to foundation engi-
neers, one of which is cyclic thermal loading effects on
Figure 8. Normal stress vs. expansion displacement during soil-energy pile interface properties. Thermal-mBSTs
the 10.5 expansion cycles. were conducted to evaluate the expansion/contraction
and temperature cycles on the soil-energy pile inter-
The initial normal stress was 41.4 kPa which was action by directly measuring t-z curves. The results
increased to 49.6 kPa at the expansion position at 0.5 showed that the shear strength decreased when the
expansion cycle. After 1 expansion cycle, the normal soil-pile interface was subjected to short-term heating.
stress was 17.1 kPa which had 38.0% reduction com- After 0.5 heating cycle in half hour, the peak strength
pared to the initial normal stress. The normal stresses decreased by 9%. The stress-displacement response
were 11.4 and 40.4 kPa after 10 and 10.5 expansion of the soil-pile interface showed a peak followed by
cycles. a reduction of shear resistance when subjected to
long-term heating (drained heating) or cycles. In the
5.3 Combined effect of heating and expansion 10.5 hours’ heating test, the peak strength is almost
the same as the reference test, however, the residual
The effects of temperature and expansion cycles can be
strength is 15% lower than the peak strength. The
combined together using the Thermal-mBST device.
permeant deformations can be caused by both dis-
Figure 9 shows the t-z curves after 0.5 expansion
placement cycles and temperature cycles which lower
cycle (D = 0.5, T = 0), 0.5 heating cycle (D = 0,
the normal stress on the soil-pile interface and result
T = 0.5), and combination of 0.5 expansion and
in a significant decrease of the shear strength. The
0.5 heating cycle D = 0.5 and T = 0.5). The peak
soil subjected to a constant normal stress and thermal
shear strength of the test with D = 0.5, T = 0.5
loads will deform with number of heating cycles. The
is 40.6 kPa compared to 43.6 kPa of the test with
normally-consolidated soil has significant permanent
D = 0.5, T = 0. During the 0.5 expansion cycle,
deformation after the first expansion cycle.
the normal stress increases as the concrete plates
expand. Due to the thermal-softening behavior of the
soil in vicinity of the interface during heating, the nor- REFERENCES
mal stress is 46.7 kPa when the plates have 120 µm
expansion with heating cycle (D = 0.5, T = 0.5) Akpinar, M. V., & Benson, C. H. Effect of temperature on
compared with the normal stress of 49.5 kPa in the shear strength of two geomembrane–geotextile interfaces.
test with D = 0.5, T = 0. Geotextiles and Geomembranes, 23(5): 443–453.
533
Akrouch, G., Sanchez, M., & Briaud, J. L. 2014. Thermo- Lutenegger, A. J. 1987. Suggested Method for Performing the
mechanical behavior of energy piles in high plasticity Borehole Shear Test. Geotechnical Testing Journal, 10(1):
clays. Acta Geotechnica, 9(3): 399–412. 19–25.
Bourne-Webb, P. J.,Amatya, B., Soga, K.,Amis, T., Davidson, McCartney, J.S. & Murphy, K.D. 2012. Strain distribu-
C. & Payne, P. 2009. Energy Pile Test at Lambeth Col- tions in full-scale energy foundations. DFI Journal, 6(2):
lege, London: Geotechnical and Thermodynamic Aspects 26–38.
of Pile Response to Heat Cycles. Géotechnique, 59(3): Murphy, K., & McCartney, J. 2014. Thermal Borehole
237–248. Shear Device. Geotechnical Testing Journal, 37(6):
Brandl, H. 2006. Energy Foundations and other Geothermal 1–16.
Ground Structures. Géotechnique, 56 (2): 81–122. Olgun, G. Ozudogru, T.Y., & Arson, C. F. 2014. Thermo-
DeJong, J. & Westgate, Z. 2009. Role of Initial State, Mate- mechanical radial expansion of heat exchanger piles and
rial Properties, and Confinement Condition on Local possible effects on contact pressures at pile–soil interface.
and Global Soil-Structure Interface Behavior. J. Geotech. Géotechnique Letters, 4(3): 170–178.
Geoenviron. Eng., 135(11):1646–1660. Suleiman, M. T., & Xiao, S. 2014. Soil-Pile Interaction of
Di Donna, A. Ferrari, A., & Laloui, L. 2015. Experimen- Geothermal Deep Foundations. Proceedings of the 27th
tal investigations of the soil–concrete interface: physical Central Pennsylvania Geotechnical Conference, Hershey,
mechanisms, cyclic mobilization, and behaviour at dif- PA on April 23–25, 2014.
ferent temperatures. Canadian Geotechnical Journal, 53: Suryatriyastuti, M.E., Mroueh, H., & Burlon, S. 2012. Under-
1–14. standing the temperature-induced mechanical behaviour
Hamid, T.B, and Miller, G.A. 2009. Shear strength of unsat- of energy pile foundations. Renewable and Sustainable
urated soil interfaces. Canadian Geotechnical Journal, Energy Reviews, 16(5): 3344–3354.
46(5): 595–606. Xiao, S. & Suleiman, M. T. 2015. Investigation of Thermo-
Handy, R. L. & Fox, N. S. 1967. A Soil Borehole Direct Shear mechanical Behavior of Soil-Energy Pile Interface Using
Test Device. Highway Research News, Transportation Modified Borehole Shear Tests. IFCEE 2015, San Anto-
Research Record, 27: 42–51. nio, Texas.
Karademir, T., & Frost, J. 2011. Elevated Temperature Xiao, S., Suleiman, T. M. & McCartney, J. S. 2014. Shear
Effects on Geotextile-Geomembrane Interface Strength. Behavior of Silty Soil and Soil-Structure Interface under
Geo-Frontiers 2011: 1023–1033. Temperature Effects. GeoCongress 2014. Atlanta, GA.
Laloui, L., Nuth, M., & Vulliet, L. 2006. Experimental Feb. 23–26, 2014.
and Numerical Investigations of the Behavior of a Heat You, S., Cheng, X., Guo, H., & Yao, Z. 2016. Experimental
Exchanger Pile. International Journal for Numerical and study on structural response of CFG energy piles. 2016.
Analytical Methods in Geomechanics, 30(8):763–781. Applied Thermal Engineering, 96(5): 640–651.
534
Benefits and optimisation of district hybrid ground source

heat pump systems
O. Mikhaylova
R. Choudhary & K. Soga

The University of Cambridge, Cambridge, UK
I.W. Johnston
ABSTRACT: Buildings consume large amounts of energy, largely to satisfy their heating and cooling needs.
Since most of this energy is derived from fossil fuels, buildings are responsible for a significant share of the
world’s CO2 emissions. Shallow geothermal energy is a promising sustainable source of energy which can
potentially satisfy the thermal requirements of buildings not only economically but also with reduced carbon
emissions. For high density urban areas, buildings can be connected to district Hybrid Ground Source Heat
Pump (HGSHP) systems for heating and cooling purposes. This study discusses the benefits of district over
individual HGSHP systems. A methodology for the optimisation of district HGSHP systems is proposed which
considers building thermal demand regimes. Such optimisation can reduce the total lifetime costs of heating
and cooling, capital investments and payback periods of HGSHP systems. The importance of considering the
demand regimes in the optimisation is demonstrated through a case study. The case study shows that an optimised
district HGSHP system can have significant financial benefits over individual HGSHP systems and therefore
make district systems more attractive to investors.
1 INTRODUCTION by a traditional system, for example, a gas boiler in

heating or a cooling tower in cooling. This arrange-
Currently, buildings consume large amounts of fossil ment allows the expensive Ground Heat Exchangers
fuels for heating and cooling purposes to make them (GHEs) to provide most of the thermal energy (usu-
significant CO2 emitters. To reduce emissions, low- ally 70–95 % of the annual required energy) while the
carbon geothermal energy from shallow depths can be comparatively cheaper traditional systems to provide
supplied to buildings to satisfy their thermal needs. the balance of the energy.
The capital costs of shallow geothermal energy instal- In highly populated urban areas, buildings are
lations, or Ground Source Heat Pump (GSHP) sys- located close to each other, so they can be combined
tems, can be high in comparison to traditional heating into districts for thermal energy supply purposes. In
and cooling systems. Some governments, for example such cases, the energy would be provided by cen-
the UK, have introduced incentives to compensate high tralised district systems. Just as occurs with individual
installation costs of these systems. More research into HGSHP systems, district HGSHP systems could be
optimisation of GSHP system is needed to increase more economical when sized to a certain shave factor
their financial attractiveness to potential investors. α. In the design and optimisation of district HGSHP
Previous research into the economics of geothermal systems, the specific variations of thermal power
systems suggests that hybrids of GSHP and traditional demand, particularly in time, of different types of
heating and cooling systems (HGSHP systems) can buildings needs to be taken into account.
often be more financially beneficial heating and/or This study evaluates likely financial benefits of dis-
cooling options than GSHP only systems (for example, trict over individual HGSHP systems. In particular,
Hackel et al. (2009), Alavy et al. (2013), Nguyen et al. the study proposes a methodology to consider thermal
(2014) and Hénault et al. (2015)). In such systems, the demand regimes of individual buildings in the design
GSHPs are sized for a certain portion (or shave fac- of district HGSHP systems, so that such systems can
tor) of a peak thermal demand α (Alavy et al., 2013). be optimised. An example case demonstrates the pro-
As such, GSHPs provide baseline thermal power up posed methodology and the potential advantages of
to their installed capacities and the rest is topped-up district HGSHP system arrangements.
535
2 METHODOLOGY of tbase = 16◦ C. The baseline temperature determines
when heating is needed: if an ambient air tempera-
2.1 Building occupancy regimes ture is less than the baseline temperature, a build-
ing needs heating. The peak heating power demand
Building occupancy is a major factor influencing
HDmax occurs at the minimum ambient air temperature
buildings’ thermal demand regimes or the periods
tamb_min . In the analysis, both Type A and B build-
when buildings need thermal power. When buildings
ings were assumed to have the same heating power
are unoccupied, their heating or cooling systems are
demands HDdesign,A = HDdesign,B = 10 kW at the aver-
usually switched off or on set-back mode. When build-
age ambient air temperature of the heating design
ings are occupied, their heating or cooling systems are
month in London, tdesign = 4.4◦ C (Table 1). The heat-
usually switched on and may supply thermal energy
ing demand of the building k at any particular hourly
to the buildings depending on the current ambient air
timestep i, when the ambient air temperature is tamb,i ,
temperature and other factors (eg. internal gains). Dif-
is calculated as
ferent buildings tend to be occupied at different rates at
any given time, albeit there are similarities depending
on nature of activities in buildings. For example, resi-
dential buildings are likely to be less occupied during
daytime and predominantly occupied during nights,
whereas commercial buildings are likely to be occu- If occupied, a building is assumed to require 100%
pied during office hours and unoccupied for the rest of heating power estimated for a particular ambient air
of the time. temperature. If unoccupied, the building is assumed
In this study, two building types were distinguished: to have a zero power demand. Since, overall, Type A
a residential building (Type A) and an office building buildings have a greater occupancy than Type B build-
(Type B). The occupancy of a building at a partic- ings, Type A buildings require more heating energy
ular point of time is modelled by its probability of annually than Type B buildings, all other factors being
presence P(p). The probability of presence is the prob- equal (Table 1).
ability of the building being occupied at a particular The annual building heating energy is calculated as
point of time. The assumed P(p)’s of the two build- the sum of the energy required by the building at each
ing types at two distinct time periods (9 am–6 pm and hourly timestep i of a year j. Considering P(p)k,i of the
6 pm–9 am) are shown in Table 1. This is a simplified building k at the hourly timestep i, the annual heating
representation of possible and more complex building energy of the building, Ek,j , is
occupancy regimes as the purpose here is to illustrate
the principles of the proposed HGSHP optimisation
methodology.
The lifetime heating energy required for the build-
2.2 Building thermal demands considering ing k can be estimated as
occupancy regimes
The example case was performed for the climatic con-
ditions of London, UK. In this climate, even if cooling
is provided to a building, the annual building heat- where Tlife is the life time of the HGSHP system.
ing demand tends to be higher than its annual cooling Similarly, if two building A and B are considered,
demand. Thus, HGSHP systems are generally sized for their annual heating energy at year j, EA + B,j , and
heating with GHEs sized for a reduced yearly aver- lifetime heating energy, EA + B,tot , are
age ground load (Section 2.3). The HGSHP design
methodology is very similar for both heating only
and heating dominant cases, so the impact of thermal
demand regimes on sizing of HGSHP systems would
be the same in these cases. To demonstrate general
principles of the proposed optimisation methodology,
the buildings are assumed to require heating only. 2.3 Sizing GHEs
Building heating demand amounts and regimes In this study, the ASHRAE design approach is used to
were assumed to depend on ambient air tempera- size GHEs (Philippe et al., 2010). According to this
tures and occupancy of buildings. Other factors, such method, the design length of GHEs is
as fractions of heating use and internal temperature
set points, can also affect building thermal demands
and should be taken into account in detailed analy-
ses. However, to keep the demonstration simple, these
additional parameters are not considered. The descriptions of the parameters in Equation 6 are
The building heating demands are assumed to be given in Table 2.
proportional to ambient air temperatures with an The GHE ground thermal load design parameters
assumed baseline ambient air temperature for heating (qy , qm and qh ) were calculated based on the estimated
536
Table 1. Occupancy and thermal demand of the two building types for the example problem.
Probability of presence P(p)

Design heating demand Annual heating energy
9 am–6 pm 6 pm–9 am HDdesign , kW Ej , kWh
Type A (Residential) 0.5 0.9 10 37,890

Type B (Office) 0.9 0.1 10 15,340
Table 2. GHE design parameters.
Design Value Description

parameter
qh , kW Calculated based on Peak hourly ground load

qm , kW assumed building thermal Monthly ground load
qy , kW demands, P(p)s, α and COPGSHP Yearly average ground load
Rb , m · K/W 0.111 Effective thermal resistance of the borehole
R1 0y, m · K/W 0.212 Effective thermal resistance of the ground to
10 years ground load
R1 m, m · K/W 0.197 Effective thermal resistance of the ground to
one month ground load
R6 h, m · K/W 0.113 Effective thermal resistance of the ground
to 6 hours ground load
Tm , ◦ C 2.3 Mean GHE fluid temperature
Tg , ◦ C 13.0 Undisturbed ground temperature
Tp , ◦ C 0 Temperature penalty
building thermal demands, their P(p)’s, the shave fac- Table 3. Financial comparison assumptions.
tor of a particular HGSHP configuration α and the
coefficient of performance of GSHPs COPGSHP . The Parameter Value
rest of the parameters are defined by the ground
thermal properties and GHE configurations. For the Installation cost of 1 kW of GSHPs, icGSHP , £ 240
Installation cost of 1 kW of gas boiler, icboil , £ 25
example case, these parameters are set following the
Installation cost of 1 m of GHEs, icGHE , £ 37.5
typical values used for vertical borehole GHEs in 1 kWh from electricity, ce , £ 0.17
London and assumed to be the same for all system con- 1 kWh from gas, cg , £ 0.05
figurations of the example case (Table 2). The GHEs COP of GSHPs, COPGSHP 3.5
are assumed to be installed sufficiently apart from each COP of gas boiler, COPboil 0.95
other to not thermally interact (hence, Tp = 0). For Life time of HGSHP system Tlife , years 20
a more detailed explanation of the GHE calculation Discount rate, DR 0.05
method, Philippe et al. (2010) should be consulted. Government incentive rate per 1 kWh of 0.0884
geothermal heat, rRHI , £
2.4 Financial indicators

Financial comparisons of different heating system The capital cost of the particular configuration of
arrangements and configurations were performed a HGSHP system, where the GSHP is sized to a
based on (a) total normalised costs of heating over particular α, is calculated as
lifetimes of heating systems and (b) payback periods.
The total normalised costs of heating were calculated
considering the capital costs of HGSHP systems and
the lifetime costs of the heating energy delivered to where CGSHP is the capital cost of the ground source
the building in net present values. The costs of heating heat pump; CGHE is the capital cost of the GHEs and
systems inside buildings (for example, in-room heat- Cboil is the capital cost of the boiler. These costs are
ing units, internal distribution pipework) and HGSHP calculated as
maintenance costs were assumed to be the same for all
HGSHP configurations and, hence, are not included
into the comparisons. The unit costs and other param-
eters for the financial comparisons are summarised in
Table 3.
537
Note that the capital costs of the HGSHP system sized heating system, Etot . Hence, all comparisons are per-
to the same α will be different for buildings A and B formed on the basis of the total normalised costs of
as the lengths (and therefore the costs) of GHEs for heating in £/MWh which is
these buildings are different (see Equation 9).
In the calculations of annual heating energy
costs, the UK government’s renewable heat incentive
(Ofgem, 2016) is taken into account. The incentive
compensates high capital investments into GSHP sys- The payback period of a particular HGSHP system
tems and is paid on a prorata basis for each kWh configuration, PBP, is calculated by dividing initial
of geothermal energy delivered. The incentives were capital investment by the total annual returns in energy
calculated as for non-domestic (commercial) instal- savings as
lations at the rRHI = 0.0884 £/kWh rate paid over the
first 20 years of GSHP installation. The annual cost of
heating energy at any particular year j, including the
renewable heat incentive, is calculated as
In PBP calculations, the initial capital investments
were taken as the difference in the capital costs of
the HGSHP system, Ccap (Equation 7), and a gas-only
where Ch_GSHP,j is the cost of heating energy provided heating system, Cboil_α = 0 (hence, for a gas heating sys-
by the GSHPs at the year j; Ch_boil,j is the cost of heating tem, the payback period is zero). The annual returns
energy provided by the boiler at the year j and RHIj is are quantified as the difference between annual heat-
the renewable heat incentives received at the year j. ing energy costs in cases of the HGSHP including
The annual heating energy provided by GSHPs dur- renewable heat incentives, Ch,j (Equation 11), and the
ing year j, EGSHP,j, is a share of the required annual gas-only heating systems, Ch_α = 0,j , converted into net
building energy, Ej , and determined by the design present values. The payback periods were calculated
shave factor, α. When needed, the boiler tops the heat- in years needed to return initial investments.
ing energy supplied by GSHPs up to the required
amount. The annual heating energy provided by the
boiler during year j, Eboil,j , is 3 RESULTS
In the case study, individual and district arrangements

of HGSHP systems for mixes of two buildings are
Considering the coefficients of performance of the compared. The following three mixes of Type A and B
equipment, the annual costs of the heating energy in buildings are evaluated:
Equation 11 are estimated as
• “A+A” mix: a Type A + a Type A buildings
• “B+B” mix: a Type B + a Type B buildings
• “A+B” mix: a Type A + a Type B buildings
In the individual arrangements, the thermal
demands per building are used to size the GHEs. In
the district arrangements, the thermal demands of the
two buildings are combined before sizing the GHE.
Thus, for these two arrangements of a same mix, the
The annual costs of heating energy are converted qy and qm thermal loads might be different at the same
into net present values to estimate the total lifetime qh .
cost of heating as For both individual and district heating system
arrangements, the GSHPs are sized for a range of
α values from 0 to 100 % with a step of 3 % to
cover possible configurations of the HGSHP systems.
Financial indicators of the resultant configurations of
where the assumed discount rate DR. Note that the dis- HGSHP systems are calculated to determine their cost
count rate prediction is a complex process and should effectiveness.
be performed for a particular project (see examples The results of the financial comparisons for the
in Oxera ( 2011)). Here, for demonstration purposes, three building mixes are presented in Figure 1. In
DR of 5% is assumed as a reasonable value for a the figure, total normalised costs of heating TNC in
low-carbon technology projects (Oxera, 2011). The £/MWh (Equation 17) are plotted against capital costs
Equation 16 follows Alavy et al. (2013). of different heating system configurations, Ccap in
Since heating systems with different total lifetime £ (Equation 7). In each plot,Ccap along the horizon-
required heating energy will be evaluated, the total life- tal axis varies from the cost of a gas-only (the most
time cost of heating Ctot is normalised by the amount left point, α = 0) to the cost of GSHP-only (the most
of heating energy provided during the lifetime of the right point, α = 100 %) heating systems. The points
538
in between represent the hybrid heating system con-
figurations with α increasing from left to right. For
each building mix, the minimum value of TNC rep-
resents the most financially optimal solution. Key
observations from Figure 1 are summarised in Table 4.
From Figure 1, if GSHP-only heating is compared
with gas-only alternatives, a GSHP-only heating is
more economical than a gas-only heating in the “A+A”
and the “A+B” mix. However, in the “B+B” mix, a
GSHP-only heating is a more expensive option com-
pared to a gas-only heating. In addition, TNC for the
gas-only options are the same for both “A+A” and
“B+B” mixes while there is a significant difference in
TNC between these two mixes for GSHP-only options.
Such differences are explained by the principles of
GHE sizing. The Type A and B buildings have the
same design heating demands (Table 1) which largely
determine the required lengths of “expensive” GHEs,
and therefore influence the cost. At the same time,
annually, less “free” geothermal energy is supplied to
Type B buildings compared to Type A buildings due to
the differences in their annual heating energy demands
(Table 1). Hence, the high capital costs of the GSHP
systems are not compensated by the delivered “free”
geothermal energy in Type B buildings as much as they
are in Type A buildings. In other words, “expensive”
GHEs are utilized more efficiently in Type A buildings
compared to Type B buildings.
In all three building mixes considered, the TNC is at
a minimum when the hybrids of a GSHP and a boiler
are used for the heating in comparison to the GSHP-
only and gas-only systems. In Figure 1, the minimum
points at the TNC curves represent the most financially
optimum HGSHP configurations which correspond
to the GSHPs sized to certain shares of peak power
demand HDmax or shave factors α.
In the optimum cases of all three building mixes,
the district HGSHP systems would ensure lower TNC
compared to individual HGSHP systems installed for
the same buildings (Figure 1). However, the difference
between the optimum district and individual TNC’s are
much higher in the “A+B” mix compared to the other
two mixes. This is explained by the significant differ-
ences in occupancy regimes of Type A and B buildings
(Table 1), so that, when combined, these buildings effi-
ciently share GSHP installed capacities, and maximize
the utility of GHEs. Note that in all optimum hybrid
cases, significant shares of heating energy are pro-
vided from sustainable geothermal sources (Table 4)
ensuring low CO2 emissions from the buildings.
If two buildings with different occupancy rates are
connected to a district HGSHP system, the building
with a lower overall occupancy rate would benefit
more financially from the district arrangement than
the building with a higher overall occupancy rate.
Indeed, the minimum TNC for residential buildings
Figure 1. Total normalised costs of heating TNC for three
in the “A+A” mix is the same as the minimum TNC
building mixes: a) “A+A”, b) “B+B” and c) “A+B”.
for buildings in the “A+B” mix, 14.6 £/MWh. At the
same time, the minimum TNC for office buildings in normalised cost of heating if they are in the “A+B”
the “B+B” mix is much higher, 23.1 £/MWh. The min- mixes compared to when they are in the “B+B” mixes.
imum TNC in the “A+B” mix is 14.6 £/MWh. Hence, However, Type A buildings would also benefit from
Type B buildings would access a much lower total being in the “A+B” mixes, since such districts would
539
Table 4. Financial comparison.
At optimum configuration
Share of geothermal
Building mix system Type of HGSHP α, % TNC, £/MWh energy, % Ccap , £
“A+A” Individual 56 14.6 91.3 20,300

District 53 15.0 91.1 19,600
“B+B” Individual 33 25.5 75.5 11,300
District 31 23.1 79.2 10,700
“A+B” Individual 47 18.7 86.6 16,600
District sized conventionally 47 15.8 96.1 16,900
District 33 14.6 88.8 13,200
allow them to have lower capital costs compared to the

“A+A” mixes (£13,200 in comparison to £19,600, see
Table 4) while still accessing the same low TNC.
Results demonstrate that building thermal demand
regimes must be combined when sizing GHEs in order
to better optimise district HGSHP systems. Indeed,
in the “A+B” mix, the individual HGSHP system
would be sized for a shave factor of α = 47 % to
achieve the minimum TNC = 18.7 £/MWh (the min-
imum of the “Individual” curve in Figure 1c). If the
district HGSHP system was designed by designing the
GHEs as per individual demands, the GSHP would be
sized for the same α = 47 % which would result in
TNC = 15.8 £/MWh. However, if the thermal demand
of the buildings is combined, the GSHPs would be Figure 2. Payback periods for optimum HGSHP config-
sized for α = 33 % which would reduce the TNC to urations for three building mixes: “A+A”, “B+B” and
14.6 £/MWh (the minimum of the District curve in Fig- “A+B”.
ure 1c). In addition, 22 % less of the capital cost would
be invested to achieve the minimum TNC, £13,200 in costs and total normalised costs of heating (Table 4).
comparison to £16,900 (Table 4). For the individual as well as for district HGSHP sys-
To evaluate risks of investments into HGSHP heat- tems sized conventionally, without taking into account
ing systems, payback periods were calculated. The building thermal demand regimes, the payback period
payback periods were estimated for the optimum con- of the “A+B” mix would be higher, 7.6 and 6.8 years
figurations of HGSHP systems installed individually respectively. This, again, demonstrates that district
and for districts of two buildings as well as for district HGSHP systems are more economical than individual
arrangements when HGSHP systems were sized indi- HGSHP systems especially when the district systems
vidually, without combining building thermal demand are optimised with building thermal demand regimes.
regimes. The results of the estimation are shown in The example case presented investigates some gen-
Figure 2. eral aspects of the optimisation of district HGSHP
From the figure, the initial investments into HGSHP systems by considering building thermal demand
systems of Type A (residential) buildings have sig- regimes. More research should be performed to quan-
nificantly lower payback periods compared to the tify thermal demand regimes of typical building mixes
initial investments into HGSHP systems of Type B based on factors, additional to building occupancy.
(office) buildings. For example, for the cases of district These detailed heating demand regimes could be used
arrangements, the payback period for the “A+A” mix in comprehensive studies of their effects on optimi-
is 5.8 years whereas the payback period for the “B+B” sation of district HGSHP systems. Based on these
mix is much higher, 9.3 years. This is explained by studies, optimum building mixes can be suggested, so
higher annual returns in the “A+A” mixes due to the that the minimum total normalised costs of heating,
higher annual heating demands for Type A buildings capital investments and payback periods are achieved.
in comparison to Type B buildings for the same design
heating demands (Table 1).
Among all calculated payback periods presented in 4 CONCLUSIONS
Figure 2, the minimum of 5.4 years is expected in the
district arrangement of the “A+B” mix. In this mix, The paper discusses potential financial benefits of
both Type A and Type B buildings would benefit from district hybrid ground source heat pump (HGSHP)
low payback periods as well as from the low capital systems in comparison to individual HGSHP systems.
540
A methodology of the financial optimisation of district Fellowship funded by The Australian Government
HGSHP systems was presented where building ther- Department of Education.
mal demand regimes were taken into account to find
the optimum share of GSHP systems in the hybrids.
An illustrative example demonstrates the economic REFERENCES
advantages of district systems for two building mixes
and the importance of combining the building ther- Alavy M, Nguyen H V, Leong WH, et al. (2013) A method-
ology and computerized approach for optimizing hybrid
mal demand regimes when district GHEs are designed.
ground source heat pump system design. Renewable
It shows that if buildings with significantly different Energy, 57, 404–412.
thermal demand regimes were connected to one (dis- Hackel S, Nellis G and Klein S (2009) Optimization of
trict) HGSHP system, they would benefit financially Cooling-Dominated Hybrid Ground-Coupled Heat Pump
from such a system arrangement. In particular, their Systems. ASHRAE Transactions, 115(1), 565–580.
total normalised costs of heating and payback peri- Hénault B, Pasquier P and Kummert M (2015) Financial opti-
ods would be significantly lower in comparison to mization and design of hybrid ground-coupled heat pump
individual HGSHP systems. systems. Applied Thermal Engineering, 93, 72–82.
In addition, the example shows that for the optimum Nguyen H V, Law YLE, Alavy M, et al. (2014) An analysis
of the factors affecting hybrid ground-source heat pump
sizing of district HGSHP systems, the building ther-
installation potential in North America. Applied Energy,
mal demand regimes have to be combined when sizing 125, 28–38.
the GHEs. If GHEs in such hybrids were sized con- Ofgem (2016) Non-Domestic RHI Main Guidance. Avail-
ventionally, without taking into account the demand able from: https://www.ofgem.gov.uk/environmental-
regimes, the resultant total normalised costs of heating programmes/non-domestic-renewable-heat-incentive-rhi
and capital costs would be higher than at the optimum (accessed 1 February 2016).
sizing. Oxera (2011) Discount rates for low-carbon and renewable
generation technologies. Available from: https://www.the
ccc.org.uk/archive/aws/Renewables Review/Oxera low
carbon discount rates 180411.pdf (accessed 1 February
ACKNOWLEDGEMENT 2016).
Philippe M, Bernier M and Marchio D (2010) Sizing Cal-
The first author would like to acknowledge the finan- culation Spreadsheet Vertical Geothermal Borefields.
cial support provided by an Endeavour Research ASHRAE Journal, 52(7), 20–28.
541
Thermo-mechanical finite element analyses of energy walls
D. Sterpi & L. Mauri

Department of Civil and Environmental Engineering, Politecnico di Milano, Milano, Italy
ABSTRACT: Energy walls are thermo–active diaphragm walls that embed heat exchanger pipes with the
purpose to exploit the thermal energy of the subsoil. Their geotechnical and structural performance is related
to both thermal and mechanical loads and still needs to be thoroughly investigated. To this purpose, three–
dimensional thermo–mechanical finite element analyses allow to highlight the effects of the heat transfer process
on the soil temperature field, on the wall and on the soil–structure interaction, in terms of internal actions and
earth pressures. The main findings show that the effects of the thermal loads can be considered admissible for
the geostructure, in terms of its global stability and structural safety, though they are generally not negligible.
Unexpected overstress conditions can occur, such as tensile stresses and out–of–plane effects, whose magnitude
depends on the thermal boundary conditions and on the constraint degree of the structure.
1 INTRODUCTION to combined thermal and mechanical loads in their

ordinary working conditions. The earliest investiga-
The energy consumption in Europe is rapidly increas- tion was addressed to the energy piles (Knellwolf et al
ing especially in the sector of buildings, which has 2011, Amatya et al 2012, Suryatriyastuti et al 2012),
surpassed the sectors of industry and transport. Out more recently the study broadened to energy tunnels
of the building total energy demand, heating currently and retaining walls (Xia et al 2012, Barla et al 2016).
accounts for 40% and, moreover, the cooling demand is Monitoring data from instrumented full scale
expected to rise significantly in the next years (Inter- energy piles provided major insights in their thermo-
national Energy Agency, 2014, Pérez-Lombard et al mechanical behaviour and established important
2008). databases for the validation of numerical analyses and
The promotion of geothermal energy for building the calibration of the relevant parameters (Laloui et al
thermal conditioning is therefore crucial to meet the 2006, Bourne–Webb et al 2009). Taking advantage
European targets about renewable energy exploitation of ideal, controlled and repeatable conditions, also
and greenhouse gas emissions reduction (European laboratory tests on small scale models helped in iden-
Parliament, 2010). In fact, even at low depths or at tifying the pile response (Kramer et al 2015, Stewart &
the near-surface the geothermal energy, directly used McCartney 2014).
as thermal energy, is pervasively available, a charac- With respect to the energy piles, the field experience
teristic that makes it optimal for local harvesting and with energy walls is newer and still there is some lack
diffuse distribution at both single space and district of monitoring data about their thermo–mechanical
scales (Lund et al. 2011). response. Following a preliminary study on the perfor-
Energy walls are thermo–active reinforced concrete mance of an energy wall, from the perspective of both
diaphragm walls designed as soil retaining structures the geothermal energy exploitation and the short and
that, similarly to other so called “energy geostructures” long term influence on the soil temperatures (Sterpi
(Brandl 2006, Adam & Markiewicz 2009, Amis et al et al 2014), the research is now addressed to inves-
2010, Laloui & Di Donna 2013, Nicholson et al 2013), tigate the effects of the heat transfer process on the
embed heat exchanger pipes with the purpose to use diaphragm wall and on the soil–structure interaction
the subsoil as a natural reservoir, where heat can be (Mauri 2015). Three–dimensional finite element anal-
dispersed in the summer season and extracted in the yses allow to get insights into the thermal induced
winter season. stresses that, ultimately, affect the earth pressure dis-
Although basically limited to the new constructions, tributions and the wall internal actions, such as axial
this system offers the advantage of using existing com- loads and bending moments.
ponents of the structure without requiring additional
works, and associated costs, as in the case of the more
traditional borehole heat exchanger systems. 2 ENERGY PILES AND ENERGY WALLS
With the spread of this technology, the issue raised
about the behaviour of these structures that, designed All the data collected from the experimental investi-
to serve a prior structural function, are subjected gations on energy piles, together with the numerical
545
and analytical predictions, confirm that the pile-soil different temperatures, the two sections are subjected
interface resistance and the conditions at the pile head to different thermal expansions or contractions. Con-
and toe, such as the presence of the over-structure or sequently, different transversal sections of the wall
of a stiff substratum, exert constraints on the thermal will behave differently and will interact, so that three–
expansion of the pile when heated, and on the ther- dimensional effects in the stress–strain distribution are
mal contraction when cooled. These constraints lead to be expected.
to thermal induced stresses that act in addition to the In the following, these aspects are discussed with
stresses induced by the mechanical loads, and influ- reference to an energy wall, assembled with segments
ence the final stress distribution (e.g. Laloui et al 2006, hosting two heat exchanger pipes each. A preliminary
Amatya et al 2012). Moreover, the cyclic nature of the FEM thermal analysis of the soil–structure system
thermal load has raised the question about the possible permits to investigate the cyclic thermal working con-
occurrence of a shaft resistance loss, due to both a pro- ditions. The most demanding condition for the wall,
gressive degradation of the interface and a reduction of corresponding to the highest temperature variations,
the normal effective stress induced by soil volume con- is then identified and considered as thermal load in a
tractions (Suryatriyastuti et al 2014). In energy piles, thermo–mechanical analysis.
significant changes were eventually observed in the
mobilised shaft resistance and in the pile axial load and,
although they are not expected to lead to detrimental 3 DESCRIPTION OF THE PROBLEM
consequences, they should be taken into consideration
at the design stage. 3.1 Geometry
Thermal induced stresses develop also in energy
A 10 m high excavation, hosting a three level base-
walls, but their effects are less predictable than in
ment of a four floor residential building, is supported
energy piles. Firstly, the wall has a greater complexity
by two facing energy walls, reaching the depth of 15
in terms of geometry: the axisymmetric approxima-
m. A vertical symmetry plane allows to model only
tion is not applicable and various restraints could act
half of the excavation (fig. 1.a). The wall and the base-
on the wall from structural components as anchors,
ment floor slab have both a thickness of 0.5 m. In these
roof and base slabs, etc. Secondly, the wall is fully
analyses the possible presence of a roof slab and of an
embedded in the soil in its lowest part only, and the
over–structure is disregarded.
thermal boundary condition at the side exposed to the
The energy wall is formed by single connected seg-
excavation could be undefined. In addition, the section
ments of reinforced concrete, each of them is 2.4 m
area of the energy wall is large enough to host a variety
large and embeds two heat exchanger pipes (fig. 1.b),
of different suitable layouts of the heat exchanger, with
fixed to the reinforcement cage at the sides of the seg-
consequent different temperature gradients induced by
ment, so to keep clear the central part where concrete is
the heat transfer process.
cast. Each pipe is made of high-density polyethylene,
With respect to the energy performance, the heat
has a 30 mm diameter and circulates a heat carrier
transfer models developed for axisymmetric struc-
fluid, mixture of water and glycol with antifreeze
tures, such as borehole heat exchangers or energy piles,
function.
cannot be straightforwardly extended to energy walls,
For the sake of computational simplicity, the loops
but specific models have to be developed (Sun et al
were assumed U-shaped and laid in the longitudinal
2013). The applicability of Thermal Response Tests
mid-section of the wall (fig. 1.c), though the results
to energy geostructures, even to piles, is still under
of previous analyses suggested that different layouts
investigation (Park et al 2013, Cecinato & Loveridge
would increase the energy efficiency (Sterpi et al.
2015).
2014).
Due to difficulties inherent to the problem mod-
In the plan section of the wall, a series of parallel
elling and to the current exiguity of data from
symmetry planes can be identified, at a distance of
monitored cases, mainly focused on the thermal per-
1.2 m from each other (fig. 1.b).Therefore, the analysis
formance and on the temperature gradients within the
of the problem can be reduced to the three-dimensional
structure and the ground (Brandl, 2006, Xia et al 2012),
analysis of a 1.2 m wide slice corresponding to half of
the thermo-mechanical behaviour of energy walls has
the single segment.
not yet been fully investigated.
The energy wall, as soil retaining structure, is basi-
3.2 Materials
cally subjected to lateral pressures contrasted by its
flexural response, a structural behaviour entirely dif- The subsoil consists of a saturated, well graded, silty
ferent from the one of piles; yet, the thermal loads will sand, of increasing stiffness with depth, in a hydro-
induce mainly an axial expansion/contraction effect static regime. Although the silty component yields
similar to the one of piles. However, this effect is a not negligible cohesion, the influence of temper-
generally not uniformly distributed along the wall lon- ature on the soil behaviour was limited to thermal
gitudinal axis. In fact, the distance between the two expansion and the hydro-mechanical coupling effects
transversal sections hosting the descending and the were neglected. The soil mechanical behaviour was
ascending portions of the heat exchanger is not neg- assumed isotropic, elastic perfectly plastic, with Mohr-
ligible and, since the two portions carry the fluid at Coulomb failure criterion and non-associated flow
546
Figure 1. (a) Section of the energy wall and the base slab in the (y,z) plane; (b) plan section in the (x,y) plane of a single
segment of the energy wall with two embedded U-shaped pipes; (c) 3D sketch with (x,y,z) reference system. The two opposite
symmetry planes are 1.2 m apart in the x direction. (units: m).
Table 1. Thermo–mechanical properties. Prescribed thermal loading and boundary condi-

tions govern the heat transfer process and the final
E ν c φ ψ α temperature gradients. In this problem, constant tem-
MPa – kPa ◦ ◦
1/◦ C peratures of 15◦ C at the soil base and 18◦ C at the
internal sides of the excavation were assumed. The
Saturated soil 80–120 0.3 20 32 15 10−5
lateral boundaries and the front and back faces of the
Soil–wall interface 80–100 0.3 1 22.6 5 10−5
domain, as symmetry planes, were given an adiabatic
Reinforced concrete 30000 0.2 – – – 10−5
condition. A yearly cyclic thermal condition was pre-
scribed at the ground surface, corresponding to the
seasonally varying air temperatures (fig. 2).
Table 2. Physical and thermal properties.
The thermal loading condition provided by the heat
Thermal Specific
exchanger is in the form of a mass flux having given
Density conductivity heat temperature at the inlet, i.e. the fluid velocity and inlet
kg/m3 W/(m◦ C) J/(kg◦ C) temperature have to be assigned.
A fluid velocity of 0.05 m/s was assumed. It should
Saturated soil 1930 2.2 1642 be also pointed out that, for any given loop layout, one
Reinforced concrete 2500 2.6 880 velocity exists that maximises the heat flux and that
Heat carrier fluid 1000 0.57 4186 high velocities in the circuit could lead to local and
distributed energy losses that would reduce the energy
performance (Sterpi et al. 2014).
rule. The reinforced concrete elements are modelled as The inlet temperature was prescribed as time–
homogeneous, isotropic, perfectly elastic, and suscep- dependent condition of yearly periodicity (fig. 2),
tible to thermal expansion. The presence of a soil-wall assuming a dual operating mode (heating/cooling).
interface, with poorer mechanical properties, was also The extreme values of 2◦ C in heating mode and 30◦ C
modelled (Table 1). in cooling mode and the extended periods of three sum-
The thermal properties of saturated soil and rein- mer months and six winter months were chosen in
forced concrete (Table 2) were assumed from the order to analyse the effects of upper and lower limits
weighted arithmetic means of the properties of the sin- of actual thermal loads.
gle components, the weights being volume fractions
for the thermal conductivity and mass fractions for the
specific heat capacity (Rees et al., 2000).
4 NUMERICAL MODEL
3.3 Thermal loading and boundary conditions 4.1 Heat transfer process and thermal analysis
The heat transfer occurs by convection within the heat The initial condition of the soil temperature field,
exchanger and by conduction within the concrete wall T0 (x,t), is the one obtained as steady state solution
and the soil, since there is no groundwater flow and of a heat transfer analysis in absence of the geother-
radiation can be neglected. mal system. The steady state solution is characterized
547
Figure 2. Seasonal variation of the ground surface tem-
perature (solid line) and fluid inlet temperature in a heat-
ing/cooling operating mode (grey areas).
by the same yearly periodicity of the ground surface

temperatures.
Once the initial temperature condition is set, the
thermal analysis can be carried out adding the thermal
load applied by the heat exchanger in dual operating
mode. With the fluid temperatures and operating peri-
ods shown in figure 2, the steady state soil temperature
field T(x,t) is reached after a transient phase lasting 2
years.
Note that a single operating mode (for instance
winter heating mode), in a hydrostatic regime, would
induce the transient phase to last longer, due to the lack
of thermal energy recharge (in the summer period) and
the consequent thermal drift, i.e. a continuous decrease Figure 3. Contour lines of the highest temperature increase,
in the soil temperature. Alternate periods of thermal at time taug : (a) in the warmest section (max value +11.5◦ C)
energy injections and extractions, of possibly compa- and (b) in the coolest section (max value +9.2◦ C); (min
rable amounts, limit the thermal drift effect and the value 0).
steady state solution is reached in a shorter time.
The most demanding thermal working conditions Figure 4 shows the temperature gradients within
for the energy wall over the year are those corre- the energy wall, along the vertical and longitudinal
sponding to the highest temperature increases and directions and in the thickness. The marked influence
decreases. of the thermal boundary condition at the excavation
Focusing on the temperature increases, the highest side is evident from figure 4.b.
values are reached at time taug corresponding to the end
of August. Then, the temperature variation T(x,taug )
4.2 Thermo-mechanical analysis
is calculated at each point x, as difference between
the initial values T0 (x,taug ) and the values T(x,taug ) The initial stress-strain state σ 0 (x) − ε0 (x), was
resulting from operating the geothermal system. obtained simulating the construction process in a
The temperature variation T(x,taug ) is shown in conventional mechanical analysis. Then, the tem-
figures 3 and 4 with reference to respectively the perature variation T(x,taug ), corresponding to the
soil mass and the energy wall. In particular, figure highest temperature increase in the summer period,
3 shows the variation in the temperature field along was applied as thermal load in a thermo-mechanical
the wall longitudinal axis, due to the presence of analysis on the system characterized by initial condi-
warmer (x = 1.2 m, fig. 3.a) and cooler (x = 0. m, fig. tions σ 0 (x) − ε0 (x) − T0 (x,taug ). The associated ther-
3.b) sections, thus proving the necessity to perform mal expansions in the soil and the wall induce
a three-dimensional analysis. The geothermal system thermal strains and stresses that lead to a stress-strain
influences the soil temperatures up to a distance of state σ 1 (x) − ε1 (x), that can be compared with the
about 4 m from the wall, but the major variations state σ 0 (x) − ε0 (x) to analyse the influence of the
(greater than 2◦ C) are found within 2.5–3 m from geothermal system.
it. The highest temperature increase, at the soil-wall With respect to the initial condition, the wall
interface of the wall deepest part, is about 9.5◦ C. exhibits a negligible variation of vertical displacements
548
Figure 5. Axial force in the energy wall under
thermo-mechanical loads, at the warmest (x = 1.2 m)
and coolest (x = 0.) sections, with respect to the initial
condition.
Figure 4. Contour lines of the highest temperature increase

at time taug in the energy wall, from the viewpoint of (a)
the wall side facing the soil and (b) the wall side facing the
excavation (min value 0, max value +11.5◦ C).
and a limited increase of horizontal displacements

(+8% at the wall top). The earth lateral pressures
increase in the active zone (soil mass side), at the
depth of 11–14 m from the ground surface, of about
10%, while in the passive zone (excavation side) the
variation is negligible.
The thermal induced stresses lead to significant
variations in the wall internal actions. Figure 5 shows
the axial force distribution within the wall, for the ini-
tial condition (solid line) and the conditions reached Figure 6. Bending moment in the energy wall under
under thermal loads (dashed lines). As expected, the thermo-mechanical loads, at the warmest (x = 1.2 m) and
warmest section, close to the descending portions of coolest (x = 0.) sections, with respect to the initial condition.
the heat exchangers (x = 1.2 m), undergoes an increase
in compressive axial load due to restraints that limit its greater than those calculated in the fully embedded
free thermal expansion. These restraints are exerted by part of the wall (10–15 m).
the soil, at the wall faces and base, by the thermal con- The influence of thermal loads reflects also, though
dition imposed at the excavation side, and by adjacent not to the same extent, on the wall bending moment dis-
sections of the wall itself that are not equally heated. tribution, shown in figure 6 with the same notation as
At the same time, the coolest section (x = 0. m) is sub- in figure 5. The thermal induced stresses lead to minor
jected to a tensile action, exerted by the adjacent heated variations in the upper part of the wall, with positive or
sections, that eventually results in a positive (tensile) slightly negative values in respectively the warmest or
axial force. the coolest sections, where the initial condition would
Note that the thermal condition at the excavation conversely set a vanishing bending moment.
side (0–10 m from the ground surface) exerts a great The non-uniformity in the wall structural response
influence, leading to variations in the axial force along the longitudinal axis (x direction), due to the not
549
uniform distribution of the thermal load T(x,taug ), is of pile response to heat cycles, Géotechnique, 59(3),
therefore confirmed by the various results. The generic 237–248.
transversal section of the wall will undergo internal Brandl, H. 2006. Energy foundations and other thermo-active
actions within the limits calculated in sections x = 0 ground structures. Géotechnique 56(2), 81–122.
Cecinato, F. & Loveridge, F. 2015. Influences on the thermal
and x = 1.2 m. Unexpected stress states, such as tensile efficiency of energy piles. Energy 82, 1021–1033.
stresses, torsion and out-of-plane effects, result from European Parliament and Council (2010). Directive 2010/31/
this non-uniformity. EU of the European Parliament and of the Council of
19 May 2010 on the energy performance of buildings.
Official Journal of the European Union, L. 153/13.
International Energy Agency (2014). Key World Energy
5 CONCLUSIONS Statistics 2014. OECD/IEA, Paris (www.iea.org).
Knellwolf, C., Peron, H. & Laloui, L. 2011. Geotechni-
The thermal analysis of the soil–structure system per- cal analysis of heat exchanger piles. J. Geotech. Geoenv.
mitted, first, to investigate the thermal working condi- Engng. ASCE 137(10), 890–902.
tions of the energy wall, subjected to a dual operating Kramer, C.A., Ghasemi–Fare, O. & Basu, P. 2015. Laboratory
mode and to seasonal variations of the atmospheric thermal performance tests on a model heat exchanger pile
temperature. The most demanding working conditions in sand. Geotech. Geol. Engineering 33, 253–271.
Laloui, L. & Di Donna, A. 2013. Energy Geostructures, ISTE
were then identified and considered as thermal loads
and John Wiley & Sons.
in a thermo–mechanical analysis. Laloui, L., Nuth, M. & Vulliet, L. 2006. Experimental
The structural response of energy walls is similar to and numerical investigations of the behaviour of a heat
the one observed in energy piles, but while for piles exchanger pile. Int. J. Numer. Anal. Meth. Geomech.
the conventional assumption of axisymmetry is still 30(8), 763–781.
appropriate, in the analysis and design of walls the con- Lund, J.W., Freeston, D.H. & Boyd, T.L. 2011. Direct appli-
ventional plane strain assumption has to be replaced cation of geothermal energy 2010 worldwide review.
by a three-dimensional scheme. Geothermics 40, 159–180.
The numerical findings show that the effects of Mauri, L. 2015.A study on energy walls behaviour by thermo-
mechanical numerical analyses (in italian). MS Thesis,
the thermal loads can be considered admissible for
Politecnico di Milano, Italy.
the geostructure, in terms of its global stability and Nicholson, D. P., Chen, Q., Pillai, A. & Chendorain, M.
structural safety. In fact, they are prominent in the 2013. Developments in thermal pile and thermal tunnel
axial direction and therefore mildly affect the struc- linings for city scale GSHP systems. Proc. 38th Workshop
tural response of the wall that is mainly based on a Geothermal Reservoir Engineering, Stanford University,
flexural behaviour. California, SGP-TR-198.
However, these effects are generally not negligi- Park, H., Lee, S-R., Yoon, S. & Choi, J-C. 2013. Evalua-
ble. The results indicate the development of internal tion of thermal response and performance of PHC energy
actions that are unusual in ordinary diaphragm walls, pile: field experiments and numerical simulation. Applied
Energy 103, 12–24.
such as tensile stresses. The magnitude of these effects
Pérez-Lombard, L., Ortiz, J. & Pout, C. 2008. A review
depends on the thermal boundary conditions and on the on buildings energy consumption information. Energy &
constraint degree of the structure, i.e. on the presence Buildings 40, 394–398.
of connected structures. Therefore, the optimal struc- Rees, S.W., Adjali, M.H., Zhou, Z., Davies, M. &
tural design of energy walls should take into account Thomas H.R. 2000, Ground heat transfer effects on the
possible situations of unexpected overstress conditions thermal performance of earth-contact structures. Renew-
as consequences of the additional thermal loads. able Sustainable Energy Rev. 4(3), 213–265
Sterpi, D., Angelotti, A., Corti, D. & Ramus, M. 2014.
Numerical analysis of heat transfer in thermo-active
diaphragm walls. Proc. 8th NUMGE Conf. (eds. Hicks–
REFERENCES Brinkgreve–Rohe), London: Taylor&Francis Group,
Vol.2, 1043–1048.
Adam, D. & Markiewicz, R. 2009. Energy from earth-coupled Stewart, M.A. & McCartney, J.S. 2014. Centrifuge modeling
structures, foundations, tunnels and sewers. Géotechnique of soil-structure interaction in energy foundations. ASCE
59(3), 229–236. J. Geotech. Geoenviron. Eng. 140(4), 04013044.
Amatya, B.L., Soga, K., Bourne-Webb, P.J., Amis, T. & Sun, M., Xia, C. & Zhang, G. 2013. Heat transfer model
Laloui, L. 2012. Thermo-mechanical behaviour of energy and design method for geothermal heat exchange tubes in
piles. Géotechnique 62(6), 503–519. diaphragm walls. Energy & Buildings 61, 250–259
Amis, T., Robinson, C.A.W. & Wong, S. 2010. Integrat- Suryatriyastuti, M.E., Mroueh, H. & Burlon, S. 2012. Under-
ing geothermal loops into the diaphragm walls of the standing the temperature-induced mechanical behavior of
Knights–bridge Palace Hotel project. Proc. 11th Int. Conf. energy pile foundations, Renewable Sustainable Energy
Geotechnical Challenges in Urban Regeneration, London. Rev. 16, 3344–3354
Barla, M., Di Donna, A. & Perino, A. 2016. Application of Suryatriyastuti, M.E., Mroueh, H. & Burlon, S. 2014. A load
energy tunnels to an urban environment. Geothermics 61, transfer approach for studying the cyclic behaviour of
104–113. thermo-active piles, Comp. & Geot. 55, 378–391
Bourne-Webb, P.J., Amatya B., Soga, K., Amis, T., Davidson, Xia, C., Sun, M., Zhang, G., Xiao, S. & Zou, Y. 2012. Exper-
C. & Payne P. 2009. Energy pile test at Lambeth Col- imental study on geothermal heat exchangers buried in
lege, London: geotechnical and thermodynamic aspects diaphragm walls. Energy & Buildings 52, 50–55
550
Lessons learned from mechanical monitoring of a thermoactive pile
J. Habert & M. El’Mejahed

Cerema, Lille, France
J.B. Bernard
Ecome, Paris, France
ABSTRACT: This paper deals with a one year-long monitoring a thermo-active pile under a two-storey resi-
dential building in Gonesse (France). To fulfil the heating demands of the building, flight auger piles have been
installed and equipped with PEHD pipes, allowing the circulation of a cold fluid, which could be heated by the
surrounding soil. To apprehend the pile’s behaviour, displacements sensors have been fixed to the reinforcement
cage and embedded in concrete, to enable an automatic monitoring of strains and temperatures. On the one
hand, it provides elements on strains, temperatures and stress along the pile and caused by real heating demands
and circulation of a cold fluid. On the other hand, it also gives access to phenomena relevant for all piles (both
thermoactive and standard ones) but currently neglected for the design of structures : significant evolutions of
strains and stress in piles caused by the daily variations of outdoor temperature and sunshine can be underlined.
The analysis of the measurements hence allows to improve understanding the behaviour of thermoactive but also
standard piles.
1 INTRODUCTION to i) additional displacements, ii) additional structural

forces and iii) modified factors of geotechnical safety
1.1 A promising method (pile compressive or tensile capacity).
In order to reduce the consumption of fossil energy
sources and the energy bill for heating and cooling 1.3 Current practice for thermoactive piles design
demands of buildings, energy geostructures constitute
an interesting alternative. Integrating heat exchanger Eurocodes and especially Eurocode 7 (EN 1997-1:
elements into geotechnical structures, and especially 2004) don’t address especially thermoactive geostruc-
in piles needed to ensure the baring capacity of the tures. Consequently there is currently neither design
structure, leads to minimized drilling costs, especially standards nor even design method for thermo-active
in comparison with additional vertical geothermal piles. To ensure safety of such structures, contractors
probes. have been constructing the buildings on thermoactive
This technology has been used for almost 30 years piles with empirical consideration or with conservative
in different countries in Europe : Keble College in design by increasing the safety factor for geotechnical
Oxford-UK on 2001 (Kefford & al., 2010), Dock Mid- resistance (Boënnec, 2009, Knellwolf et al., 2011).
field Zurich Terminal Airport-Switzerland on 2003 However mechanical design of thermoactive piles
(Pahud & al., 2007), thermoactive embedded walls of appears clearly as a two steps problem : on the first
metropolitan stations, (Brandl, 2006, Bardoneschi & step (but most important step of the process thermoac-
Bernard, 2014). tive piles behaviour has to be reproduced. It involved
to understand the behaviour of such piles, bases on a
wide range of tests, including in situ tests, physical
1.2 Thermo-Mechanical effects modelling and long term monitoring : the present arti-
However mechanical effects induce by temperature cle deals with a one year-long monitoring thermoactive
variations shall be addressed. The injected fluid in piles under a building.
the pile, whose temperature comprised between 1◦ C
to 35◦ C, I) leads indeed to either contraction (when
a cold fluid is injected) or dilation (when a warm 2 MONITORED BUILDING DESCRIPTION
fluid is injected) of the pile itself and the surrounding
soil, but also ii) potentially modifies the soil strength Piles of a two storey residential building located
and stiffness parameters due to thermo-mechanic in Gonesse (France) have been monitored. Figure 1
effects. These phenomena can in a second step lead shows an overview of the building.
551
Figure 1. Overview of the monitored building, on the right Figure 2. Vibrating wire gauge fixed on reinforcing cage.
side of the picture (source Google Earth).
2.1 Building description and energy demands

41 flats have been build, with a global area of 2995 m2 .
The overall building presents a peak heating demand
of 49 kW, combined with a global heating demand of
10.5 MWh/year.
2.2 Soil conditions and foundation system

The building is located on alluvial modern deposits
with poor mechanical properties on the first 7
to 9 meters, covering sandy layers (Sables de
Beauchamps). To ensure its bearing capacity, pile
foundations have been chosen.
The average SLS load on each pile is 250 kN. Con-
tinuous flight auger piles have been installed, with a
12,00 m length and a diameter comprised between 500 Figure 3. Location of monitored piles and strain gauges in
each pile.
and 600 mm.
2.3 Thermoactive piles energetic design

– temperature sensors, consisting of continuous optic
Crossing energy demands of the supported buildings fiber and thermocouples.
and the length of the piles, it has been chosen to only
fulfil the heating demands of the building. Due to construction process, it was finally not pos-
In the present case, after estimating thermal proper- sible to have access to temperature datas. That’s why
ties of the surrounding soils, 63 geothermal piles (out only results of the vibrating wire gauges will be pre-
of 80 available piles) have been equipped, each with sented in this article. However, these gauges also allow
two U-shaped geothermal pipes. a measurement of the temperature. Figure 3 shows The
location of the seven gauges initially installed : it has
been tried to keep (before embedding in concrete) a
2.4 Monitoring of piles
constant spacing between the strain sensors and the
Two piles have been monitored : a first thermoac- U-loop PEHD pipes.
tive 600 mm in diameter pile (P71) , located under
the building. A second standard (non thermoactive)
500 mm in diameter pile (P65) located under the edge
3 OBTAINED MEASUREMENTS
of the building.
To ensure the long term monitoring of piles, two
The building has been under construction from the
kind of sensors have been installed:
beginning of 2012 until middle of 2013. The first mea-
– strain sensors, consisting in vibrating-wire gauges, surements have been made in October 2014, a priori
to monitor the vertical strain. These sensors have just before the second heating cycle.
been fixed to the reinforced cages (as shown on Fig- For different reasons it was note possible to have
ure 2), before their embedding in the fresh concrete. access to measurement of strain of gauges 3 and
C-110 vibrating-wire gauges provided by Telemac 7. Measurements have been performed automatically
have been used, every 5 minutes since 22nd October 2014.
552
Figure 4. Typical temperatures measurements. Figure 6. Characteristic evolving of vertical strains
(geothermal pile, z = 7 m).
Figure 5. Temperatures measurements in thermoactive pile. Figure 7. Overall strain measurements.
3.1 Temperature measurements expansion factor of the vibrating wire. As thermal

expansion factors of the vibrating wire gauges and the
Figures 4 and 5 show the characteristic evolving of
pile materials (concretes and steel are very similar, the
temperature both in thermoactive and standard piles.
vertical strain εel
v can easily and directly be obtained,
For the thermoactive pile, the progressive decreas-
using Equation 3.
ing of temperature can be stated during winter. For
the standard pile, the first temperature sensor show
important temperature variations, due to his loca-
tion under the building’s edge. This sensor is indeed
affected by the outside temperature. On the contrary, Figure 6 shows the characteristic evolving of these
the deeper sensors do not show significant temperature different strains for the gauge V2 in thermoactive pile.
variations. When a cold fluid is injected in a mechanically free
pile, the vertical strain would be equal to the thermal
3.2 Strain measurements strain εth
v and traduce contraction. Because of the sur-
Vertical strain in the pile can be studied through tak- rounding soil and supported structure, the (observed)
ing into account temperature effects and decomposing vertical shrinkage εv of the pile which is a limited part
the total vertical strain εv in an elastic part εel of the thermal strain. This phenomena however leads
v and at
to additional tensile stresses in the pile, equal to Eεel
thermal part εth
v , as quoted in Equation 1 (Laloui & al, v
(where E is the Young modulus of the pile material).
1999).
During winter positive elastic vertical strain means
tensile stress.
Figure 7 shows the characteristic evolving of tem-
perature and strains for different sensors both in the
Through measuring the frequency F of vibrating-
thermoactive pile (P71) and the standard pile (P65)
wire gauges at time t, the vertical strain can be obtained
(each point is the average measurement during 4 days).
using Equation 2, taking into account temperature T .
Figure 8 shows the same results on a shorter
period to ease the analysis (each point is the average
measurement during 1 day).
For the standard pile, since the temperature remains
quite constant, there is no strain caused by temperature
where K is a gauge factor (for C110 vibrating wire changes, and the elastic vertical strain is equal to the
gauges, K = 1.875 10−9 Hz−2 ) and αT the thermal vertical strain.
553
Figure 8. Strain measurements during the first winter. Figure 10. Correlation between outside temperature and
strains, pile P71, z = 7 m.
Figure 9. Influence of outside temperature, pile P71, Figure 11. Influence of time interval between consecutive
z = 7 m. measurements.
4 FURTHER ANALYSIS
4.1 Influence of outside temperature

Figure 9 shows the daily evolution of elastic verti-
cal strain during the 26th and 27th October 2014.
The correlation between the pile temperature and the
strains hence appears clearly. When the outside tem-
perature increase, the load on the geothermal pile P71
decreases. At the same time, the load on the Standard
pile decreases. Moreover the effect of the outside tem-
perature is more important for the shallow sensors than
for the deepest ones. It’s supposed that these results Figure 12. Strain variations in geothermal pile, z = 7 m.
are linked to the thermal expansion and contraction of
the building itself accompanying outside temperature Hence a 5 minutes interval (as chosen here) might
variations. appear quite small and at the same lead to a signifi-
To better analyse the phenomena, the Figure 10 cant load of datas. However, increasing this interval up
shows the correlation and the phase angle between to 2 hour will prevent to have access extreme values.
outside temperature and strains. Consequently the present study leads us to maintain at
Especially in summer, daily evolutions of tempera- least on measurement each hour.
ture can lead to additional strain up to 30 µdef. If the
short term modulus of deformation of concrete is con-
sidered, with an average value of 30000 MPa, a daily 4.3 Influence of geothermal fluid
variation of vertical stress in pile up to 90 kPa can also The evolving of the vertical strain is then compared to
be obtained. temperature changes in geothermal pile P71. In a first
step, measurements obtained for the thermoactive pile
at z = 7 m are plotted in Figure 12 (each point is the
4.2 Choosing the right time interval
average measurement during a 4 days period).
Figure 11 shows the influence of the chosen time inter- Strains hence appear to evolve quite linearly with
val between consecutive measurements during 27th temperature variations. The same analysis can finally
October 2014, for pile P71 at z = 7 m. be performed for each strain gauges, and also for
554
Moreover, the measurement provides useful infor-
mation for standard piles located under buildings. The
importance of the variations of strains due to outside
temperature has indeed been stated. This phenomena
highlights the importance of automated measurements
to finely understand the behaviour of piles under
buildings.
The vertical strain monitoring is still functioning
and will allow in the future year to better understand
the behaviour of thermoactive pile systems.
Figure 13. Strain variations in the two monitored piles. ACKNOWLEDGEMENTS
The work described in this paper forms part of a

Table 1. Ratio between temperature variations in pile 71 research project “GECKO” which is supported by
and strain. a grant from the French National Research Agency
(ANR). It is an industrial project with interna-
P71 P67 tional collaboration, involving companies and research
Pile laboratories in civil and energy engineering sec-
z (m) 0.5 7 11.8 6 11.8 tor: ECOME, BRGM, Cerema, IFSTTAR, LEMTA,
◦
LGCgE-Université Lille 1.
εel
v /T (µdef/ C) −9.1 −10.7 −7.3 4.5 2.8
REFERENCES
gauges installed in the standard pile. Strains variations Bardoneschi, B. & Bernard, J.B., 2014, Captage géother-
are presented on Figure 13. This allows to determine mique en parois moulées de stations de métro, Xpair.com
the ratio εel Bourne-Webb P.J., Amatya B., Soga K., Amis T., Davidson
v /T. The results for the first winter are
presented in Table 1. C., Payne P., 2009: Energy pile test at Lambeth College,
London: Geotechnical and thermodynamic aspects of pile
The previous analysis hence shows that during response to heat cycles, Geotechnique, vol. 59, no. 3,
winter, due to the circulation of a cold fluid, the 237–248.
load on geothermal piles will slightly decrease, quite Brandl H., Energy foundations and other thermo-active
linearly with temperature variations. Consequently, ground structures, Geotechnique, 56 n◦ 2, 81–122.
the mechanical vertical load on standard piles will Kefford N, 2010, Case study, long term monitoring of energy
increase. piles at Keble College GSHPA Research Seminar., Oxford.
ARUP.
Pahud, D. & Hubbuch, M., 2007 : Measured thermal perfor-
4.4 Conclusions mances of the energy pile system of the dock midfield at
Zurich Airport, European Geothermal Congress.
The present paper deals with a long term monitoring Knellwolf, C., Peron H., andLaloui L., 2011 : Geotechnical
of a thermoactive pile but also of a standard piles. Analysis of Heat Exchanger Piles. Journal of Geotech-
Based on different analysis of the strains in two piles nical and Geoenvironmental Engineering, 137(10),
under a residential building, different phenomena have 890–902.
Boënnec, O., 2009 : Piling on the Energy, GeoDrilling
been output. During winter, the temperature variation
International, 25–28.
leads to contraction of geothermal piles, and hence a Laloui, L., Moreni, M., Steinmann, G., Vulliet, L.,
decrease of the compressive vertical stress and of the Fromentin, A., Pahud, D. 1999 : Test en conditions réelles
load at pile head. This phenomena is accompanied by du comportement statique d’un pieu soumis à des sollic-
an increase of the load the measurements on narrow itations thermo-mécaniques. Report of the Swiss Federal
standard piles. Office of Energy.
555
Effect of forced thermal recharging on the thermal behaviour

of a field scale geothermal energy pile
Mohammed Faizal & Abdelmalek Bouazza

Department of Civil Engineering, Monash University, Melbourne, Australia
ABSTRACT: Many applications require ground source heat pumps coupled with energy piles to operate for a
given number of hours per day. During the times when the heat pumps are switched off, the ground temperatures
normally recover naturally. The ground can be forcefully recharged during the rest periods using solar heaters or
cooling towers combined with energy piles. The effect of solar thermal recharging on energy extracted, ground
and pile temperatures, and pile thermal strains and stresses are presented in this paper. Forced ground thermal
recovery is experimentally investigated and included for 8 hours operation with 16 hours heating and 16 hours
operation with 8 hours heating. The longer recharge duration gives higher benefits in terms of energy extracted
and ground temperature recoveries with larger thermal loads on the pile. The pile temperatures, thermal strains
and stresses undergo cyclic response for both operating modes.
1 INTRODUCTION the thermal strain and stress distribution in field scale

energy piles has mostly been studied for continuous
Geothermal energy piles are foundation piles used as operating modes (Laloui et al., 2006, Bourne-Webb
underground heat exchangers to maintain human ther- et al., 2009, Murphy et al., 2014, Murphy and McCart-
mal comfort in buildings, via the use of a Ground ney, 2014, McCartney and Murphy, 2012, Akrouch
Source Heat Pump (GSHP). The GSHPs may be oper- et al., 2014, Wang et al., 2015, Mimouni, 2014).
ational for different intermittent hours per day or may The thermal behaviour of energy piles for intermit-
run continuously for 24 hours daily. tent operation with forced ground thermal recharge
The ground temperatures during the rest periods remains yet to be studied.
of intermittent modes can be forcefully recharged The objective of this paper is thus to experimen-
through the use of optimized hybrid systems that utilise tally investigate the influence of forced ground thermal
solar collectors or cooling towers combined with recharging for two operation modes of a field scale
GSHPs (Wood et al., 2010, Man et al., 2010). Forced energy pile. The forced thermal recharging is done
thermal recharging improves geothermal energy usage during the non-operating times of the cooling unit,
as well as effectively alleviates the heat build-up and simulating a solar hybrid GSHP system for a sched-
provides a balance of ground temperatures around uled intermittent mode of operation. The recharging
the heat exchanger (Jalaluddin and Miyara, 2012). hot water temperature is maximized to study the effects
Unlike borehole heat exchangers, the dimensions and of extreme operating conditions.
spacing of energy piles are restricted by the superstruc-
ture design. Hybrid systems are therefore a feasible
option to improve energy usage from energy piles
and preventing ground thermal imbalances in a given 2 PILE INSTRUMENTATION AND
building footprint. It thus becomes important to study EXPERIMENTAL PROCEDURE
the ground temperatures and pile thermal behaviour to
identify any potential limitations as a result of forced Two operating modes were experimentally investi-
thermal ground recharge. gated on a field scale energy pile located at Monash
Intermittent operating modes of borehole heat University, Australia (Wang et al., 2015, Singh et al.,
exchangers with forced ground recharge have shown 2015), one mode with 8 hours of daily cooling fol-
that improvement in ground temperatures and per- lowed by 16 hours of heating (8C16H mode) and the
formance of the GSHP is more effective compared other mode was 16 hours of daily cooling followed
to natural ground recovery (Dai et al., 2015, Wang by 8 hours of heating (16C8H mode). Six days (128
et al., 2010, Wang et al., 2012, Man et al., 2010). The hours) of results are assessed for each mode. Only six
few studies on energy piles with ground solar thermal days of data are presented due to site constraints in
recharge have also shown to oppose the ground tem- the 8C16H mode. The results of the 16C8H mode are
perature reduction and improve performance of the adapted from recent studies reported by (Faizal et al.,
GSHPs (Wood et al., 2010, Wood, 2011). At present, 2015). In these experiments, the hot water is circulated
557
in the same loops as the cold water immediately after
switching off the chiller.
A United Refrigeration chiller with a tank volume
of approximately 35 litres was used to circulate cold
water. A GeoCube TRT unit with a tank volume of
approximately 4 litres was used to circulate hot water
in the pile, at flowrates of 15 and 13.5 litres/per min,
respectively. A GPI TM075 water flowmeter mounted
on the inlet to pile pipe recorded the water flowrates.
The chiller and the heater were connected to the pile
inlet and outlets using insulated HDPE pipes, with an
approximate length of 15 m. The water temperature in
the chiller and heater tanks were manually controlled
using bypass valves, before the water was released
into the pile (Fig. 1). The purpose of regulating the
temperatures at the beginning of each thermal cycle
was to study the pile thermal behaviour under extreme
operating conditions.
The hot water in the pipe loops from the heating Figure 1. Chiller and heater connection to the pile.
cycles got mixed and increased the cold water tem-
peratures in the cooling cycle for both modes. This of the pile (Fig. 2). Type K thermocouples recorded
change in temperatures is expected practically in alter- atmospheric temperatures, inlet and outlet water tem-
nating heating/cooling applications of hybrid systems peratures at the pile head, and ground temperatures
(Dai et al., 2015). Also, the elevated temperatures of at eight depths in two boreholes located at radial
the ground from the heating cycle contributed to the distances of 0.5 m and 2 m from the pile edge. The
increase of the cold water temperature in the chiller data from the thermocouples were logged using Pico
tank during the cooling cycle. Similarly, the hot water Technology’s USB-TC08 data logger, whereas data
temperature was affected by the cold water in the loops from the strain gauges were logged using DataTaker’s
when released into the pile immediately after the cool- DT80G and CEM20 data loggers. For the purpose
ing cycle. Hence flow in the pile was stopped multiple of this study, the temperatures and thermal strains
times by shutting the inlet and outlet valves of the cool- observed from the axial strain gauges on the upper and
ing/heating unit and using the bypass valves to control middle pile sections are considered, at depths of 5.4 m
the water temperature before releasing it into the pile. and 8.2 m, and 11.6 m and 13.3 m, respectively. The
Since it was desired to have a continuous flow in the soil profile at the test site consisted of sands and fine
pile, the stoppage times were minimized as much as to coarse to very dense clayey sands, also summarized
possible. Due to extreme high temperatures from the in Fig. 2 (Yu et al., 2015, Barry-Macaulay et al., 2013).
heating cycle in the 8C16H mode (as high as 70◦ C), The top end of the pile was not restrained and the pile
the compressor of the chilling unit would overload and head was exposed to the atmosphere. The undisturbed
was stopped to cool down. Hence, the effective cooling ground temperatures for the year 2013 recorded at this
and heating periods are approximately 15.5–16 hours site showed that surface temperature fluctuations were
for the 16C8H mode and between 4–7 hours in the felt up to 6 m depth (Yu et al., 2014).
8C16H mode. The thermal strains, εT , induced in the concrete were
Figure 2 shows a schematic of the instrumented calculated as:
bored pile used for this study. This pile (0.6 m diam-
eter with a depth of 16.1 m), which was installed in
December 2010, was specifically designed to study
the changes in the shaft capacity of pile after heating where fi is the resonant frequencies of the strain gauges
and cooling operations (Wang et al., 2015), and thus at time i, fo is the reference resonant frequency, GB is
is slightly different from a conventional pile. There the calibration factor (G is gauge factor and B is batch
were two OsterbergCell (O-Cell) load testing sys- factor), Ti is the temperature of the strain gauges at
tems installed at approximately 10 m and 14 m below time i, To is the reference temperature, and αs is the
ground level, dividing the pile into three sections: the thermal expansion of the steel wire.
10 m upper section, the 4 m middle section, and the The free or unrestrained thermal strains, εT ,free , of
lower 1 m section. Three standard HDPE pipe loops, each gauge were calculated as (Murphy et al., 2014,
with outer diameter of 25 mm and inner diameter of Murphy and McCartney, 2014):
20 mm were attached to the reinforcement pile cage.
The pipes were installed 50 mm from the edge of the
pile and to the top of the lower O-Cell (LOC), to a
depth of 14.2 m. The spacing between the loops was where αc is the coefficient of linear thermal expan-
about 175 mm. sion of concrete and T is the change in temperature
Embedment and sister bar vibrating wire strain of concrete. The thermal expansion coefficient of con-
gauges were installed at different depths along the axis crete depends on the constituents of the mix, and can
558
Figure 3. Atmospheric temperatures for the two modes
recorded during experimentation.
They ranged between 21– 35◦ C for 8C16H mode and

16–37◦ C for 16C8H mode.
Figures 4a and 4b show the fluid inlet, outlet, and
change in fluid temperatures of the 8C16H and 16C8H
modes, respectively, presented after the target inlet
temperatures were achieved using the bypass valves,
as discussed in section 2.
The lowest initial inlet cold water temperature
achieved in the 8C16H mode for cooling operation
Figure 2. (a) Schematic of the instrumented energy pile was approximately 20–28◦ C (Fig. 4b), outlet temper-
(LOC: lower O-cell, UOC: upper O-cell, Emb: embedment atures ranged between 16◦ C to 34◦ C, and difference
strain gauges, Sis: sister bar strain gauges) (Adapted from between inlet and outlet between 4◦ C to 6◦ C for all
(Wang et al., 2015). cooling cycles.
The inlet hot water temperature for the heating cycle
be as high as 14.5 µε/◦ C, while that of steel rein- of 8C16H mode was initially 40◦ C (Fig. 4b) and went
forcements ranges from 11.9 to 13 µε/◦ C (McCartney as high as 70◦ C in the heating cycle. The outlet tem-
et al., 2010, Stewart and McCartney, 2014). Thermally peratures ranged from 35◦ C to 62◦ C and the difference
induced stresses, σT , were calculated from the differ- between the outlet and inlet hot fluid temperatures was
ence between the measured thermal strains and the almost a constant of 9◦ C.
free thermal strains from the gauges as (Murphy and The lowest initial inlet cold water temperature
McCartney, 2014, Murphy et al., 2014): achieved in the 16C8H mode for cooling operation
was approximately 16◦ C (Fig. 4a), outlet temperatures
ranged between 11◦ C to 21◦ C, and difference between
inlet and outlet between 3◦ C to 5◦ C for all cooling
where E is the Young’s modulus of the pile concrete cycles.
(taken as 30 GPa). In the present study, cooling is taken The inlet hot water temperature for the heating cycle
as compressive, giving negative thermal strains. of 16C8H mode was initially between 30◦ C–33◦ C and
The difference in the water temperatures at inlet went as high as 55◦ C in the heating cycle. The outlet
and outlet of the pile is used to calculate the energy temperatures ranged from 25◦ C to 45◦ C and the differ-
extracted or rejected, Q̇, as: ence between the outlet and inlet hot fluid temperatures
was almost a constant of 9◦ C.
The difference between the outlet and inlet hot fluid
temperatures was almost constant for both modes, pos-
where ρ is density, V̇ is volume flowrate, and Cp is sibly due to the low volume capacity of the heating unit
the specific heat capacity of water, Tf ,out − Tf ,in is the and powerful electrical heaters which rapidly heated
temperature difference between inlet and outlet water the returning fluid from the pile before re-circulating it
at pile head, and L is the length of the pile (14.2 m since to the inlet. Hence a constant amount of thermal energy
the HDPE pipes only extended up to this depth in the was injected into the ground daily during the forced
pile). thermal recharging period (Fig. 7). It should be noted
that the heating cycle temperatures are maximized to
investigate extreme operating conditions.
3 RESULTS AND DISCUSSIONS Figures 5 and 6 show the ground temperatures for
both modes at radial distances R = 0.5 m and R = 2 m
The atmospheric temperatures recorded at the test site from the pile edge. The ground temperatures at 2 m and
for the two operating modes during the experimenta- 4 m depths were affected by atmospheric temperatures.
tion period (128 hours, 6 days) are shown in Fig. 3. There was no change in ground temperatures at 16 m
559
Figure 4. Fluid temperatures a) 8C16H mode and b) 16C8H mode (Note: the vertical axis scales are different).
depth for both modes since the pipe loops in the pile is positive and energy injected is negative. The energy
were only installed up to 14.2 m depth. injected in both modes is almost constant at 570–
The daily ground temperature recovery are more 590 W/m due to almost constant difference between
evident in the 8C16H mode at R = 0.5 m due to inlet and outlet temperatures (Fig. 4). The 8C16H
longer recharging time and higher inlet temperatures mode gives almost twice higher heat extraction rates
which considerably increased the ground tempera- than the 16C8H mode. Compared to the 16C8H mode,
ture compared to the 16C8H mode. Optimized hybrid the 8C16H mode significantly improves the thermal
systems in actual operating conditions with design gradient between the ground and the fluid as a result
control strategies will have better balance of ground of higher ground temperatures; therefore more energy
temperatures (Man et al., 2010). is extracted in the cooling cycle of the 8C16H mode.
The ground temperature increase at R = 0.5 m in the Figures 8a and 8b shows pile temperatures for
8C16H mode at end of experiments was approximately 8C16H and 16C8H modes, respectively. There is an
10◦ C and 3◦ C in the 16C8H mode, with reference almost cyclic pile temperature change from approxi-
to initial temperature for a depth of 12 m (Fig. 5c) mately 26◦ C to 54◦ C in the 8C16H mode from end
The ground temperature changes were almost negli- of recharge to end of cooling, and from approximately
gible at R = 2 m for both modes. This leads to the 13◦ C to 34◦ C in the 16C8H mode from end of recharge
conclusion that forced thermal recharging (even under to end of cooling, for depth of 5.4 m (Fig. 8c). There are
extreme operating conditions) may not have any poten- differences in 8C16H mode peak temps due to chiller
tial thermal interactions with other nearby energy piles stoppage which resulted from very high temperatures
in long-term actual building operations. Further stud- circulating in the chiller tank. The 8C16H mode gener-
ies with longer operation times will give more insight ally induces higher temperatures in the pile compared
into ground temperature response at different radial to 16C8H mode (Fig. 8c); hence larger thermal strains
distances. and stresses are expected for this mode. Pile tempera-
Figure 7 compares the average energies extracted tures of both modes however undergo cyclic operation
and injected for the two modes, where energy extracted and return to almost similar values at end of cooling
560
Figure 5. Transient ground temperatures at R = 0.5 m. a) 8C16H mode, b) 16C8H mode and c) 12 m depth for both modes.
and recovery cycles for consecutive days. This gives and 2◦ C for 16C8H mode). The difference between
rise to cyclic response of thermally induced strains and recovery and cooled peak temperature is also higher
stresses in the pile. Due to the discontinuity in the piles in 8C16H mode due to high heating fluid temperatures
created by the presence of the O-cells, the upper 10 m (28◦ C for 8C16H and 22◦ C for 16C8H mode for depth
and the middle 4 m sections of the pile were analysed of 5.4 m).
separately. Figures 10a and 10b show the pile axial thermal
Figures 9a and 9b shows pile temperatures against strains for 8C16H and 16C8H modes, respectively,
depth for 8C16H and 16C8H modes, respectively, calculated using equation 1 and zeroed at beginning of
between 80–120 hours. There are slight differences experiments. The lowest observed thermal strains, and
between peak temperatures at end of cooling and hence the most restricted, is observed at 5.4 m depth
recharging for 8C16H mode due to inconsistency in for both modes. The peak thermal strains at end of
maintaining the inlet fluid temperatures. The 8C16H cooling/recovery for both modes return to similar val-
mode gave larger temperature difference between ues for daily thermal cycles. It appears that the soil
depths compared to 16C8H mode due to higher fluid surrounding the pile did not affect or limit the thermal
temperatures in the heating cycle (5◦ C for 8C16H strain behaviour as a result of soil-structure interaction,
561
Figure 6. Transient ground temperatures at R = 2 m. a) 8C16H mode and b) 16C8H modes.
Figure 7. Daily average energy extracted/injected for both modes.
for the present case. A further study of forced ther- strains compared to the 16C8H mode (due to higher
mal recharging on effects of soil properties and pile recharging temperatures and longer recharging time).
settlement would give more insight into the coupled Even though the 8C16H mode induces higher ther-
behaviour of pile and soil. mal loads in the pile, the thermal axial strains from
There is an almost cyclic change in thermal strains the present results however show a cyclic response for
from approximately 110 µε to 300 µε in the 8C16H daily thermal cycles for both modes. Hence, it is pos-
mode from end of cooling to end of recharge, and sible that no substantial deformations in the pile and
from approximately −20 µε to 110 µε◦ C in the 16C8H surrounding soil as a result of increased thermal load-
mode from end of cooling to end of recharge, for a ing are expected due to forced thermal recharging in
depth of 5.4 m (Figure 10c). the present study.
The 8C16H mode generally induces higher thermal Figures 11a and 11b show the axial thermal
loads in the pile as a result of higher thermally induced strains against depth, for 8C16H and 16C8H modes,
562
Figure 8. Transient pile temperatures for a) 8C16H mode, b) 16C8H mode and c) at 5.4 m depth for both modes (Note: the
vertical axis scales are different).
respectively. The lowest thermal strains were observed and longer cooling period, the thermal strains in the
at 5.4 m for both modes indicating that the largest con- 16C8H mode were observed to be negative at end of
straint was at this location in the upper pile section. daily cooling cycles.
There are slight differences between thermal strains Figures 12a and 12b show the axial thermal stresses
at end of cooling and recovery since the pile did not in the pile for 8C16H and 16C8H modes, respectively.
reach thermal equilibrium with the ground as a result The axial thermal stresses are estimated using equa-
of short term operation. tion 3. As expected, significant differences in axial
The difference between recovery and cooled peak thermal stresses are observed between the two modes
thermal strains is higher in the 8C16H mode due to due to significant differences in pile temperatures and
high heating fluid temperatures (190 µε for 8C16H and thermal strains. The maximum cyclic thermal stresses
130 µε absolute for 16C8H mode for depth of 5.4 m). in the 8C16H mode at end of cooling and recovery
Higher inlet temperatures in the heating cycle also lead was approximately 0 MPa and 6.5 MPa, respectively,
to higher strain constraints at 5.4 m depth compared to whereas the maximum cyclic thermal stresses after
8.2 m depth. The thermal strains are less restrained equilibrium in the 16C8H mode at end of cooling and
at 8.2 m depth due to partial restraint provided by the recovery was approximately −1.1 MPa and 3.5 MPa,
upper load cell. Due to lower cooling fluid temperature respectively, at 5.4 m depth (Fig. 12c). The thermal
563
Figure 9. Pile temperatures against depth for a) 8C16H mode and b) 16C8H mode (Note: the horizontal axis scales are
different).
stresses at end of cooling for 16C8H mode was con- depth are different from previously studied field scale
sistently negative as a result of lower fluid temperature energy piles (Laloui et al., 2006, Bourne-Webb et al.,
and longer operation time. 2009, Murphy et al., 2014, Murphy and McCartney,
The response of thermal stresses for both modes is 2014, McCartney and Murphy, 2012, Akrouch et al.,
cyclic with minimal differences between peak values at 2014) due to presence of O-cells which divided the
end of cooling and recovery cycles between consecu- pile into sections, and due to less number of gauges.
tive days. The results indicate that the induced thermal The thermal stresses for both modes are highest at
stresses in energy piles should be stable for forced 5.4 m depth at end of both cooling and heating cycles
thermal recharging. Hence no significant effects on due to the larger restraint on thermal strains com-
the pile performance and surrounding soil deformation pared to other depths. The lowest thermal stresses for
are expected as a result of forced thermal recharging. both modes are for 8.2 m depth possibly due to partial
It should be noted that the recharging temperatures restraint provided by the upper O-cell which created a
in the present studies have been maximized to study discontinuity in the pile.
the effects of extreme operating conditions. Thermally The difference between recovery and cooled peak
induced stresses in energy piles in actual operations thermal stresses is higher in the 8C16H mode due to
of hybrid systems will be much lower due to control high heating fluid temperatures (6200 kPa for 8C16H
strategies put in place to activate ground recharging. and 5000 kPa absolute for 16C8H mode for depth of
Figures 13a and 13b show the pile axial thermal 5.4 m). Higher inlet temperatures in the heating cycle
stresses against depth for 8C16H and 16C8H modes, also lead to higher strain constraints at 5.4 m depth
respectively. The trends in thermal stresses against compared to 8.2 m depth.
564
Figure 10. Transient thermal strains for a) 8C16H mode, b) 16C8H mode and c) 5.4 m depth for both modes (Note: the
vertical axis scales are different).
Figure 11. Axial thermal strains against depth for a) 8C16H mode and b) 16C8H mode (Note: the horizontal axis scales are
different).
565
Figure 12. Transient axial thermal stresses for a) 8C16H mode, b) 8C16H mode and c) 5.4 m depth for both modes (Note:
the vertical axis scales are different).
4 CONCLUDING REMARKS operating conditions.The following major conclusions

are drawn from the results of this study:
The objective of this paper was to study the thermal
behaviour of a field scale energy pile under forced • The ground temperatures at 2 m radial distance
ground recharge (simulating solar thermal recharge) from the pile surface was not affected by high pile
for 8 hours cooling and 16 hours heating (8C16H), and emperatures; hence it is possible that no potential
16 hours cooling and 8 hours heating (16C8H), respec- thermal interactions between energy piles with 2 m
tively. The hot water temperature was maximized for or more spacing in actual building operations are
a scheduled recharging during non-operating times expected due to elevated temperatures from forced
of the cooling unit to study the effects of extreme thermal recharging.
566
Figure 13. Axial thermal stresses against depth for a) 8C16H mode and b) 16C8H mode (Note: the horizontal axis scales are
different).
• Higher recharging temperatures with longer recharg- (project number LP120200613) and contributions
ing time leads to higher ground temperatures near from Vibropile Pty. Ltd., Golder Associates Pty. Ltd.
the pile. and GeoExchange Australia Pty. Ltd. Their support is
• The energy extracted was almost twice in the 8C16H gratefully acknowledged.
mode compared to 16C8H mode.
• The pile temperatures, thermal strains and stresses
are higher in the 8C16H mode, but undergo cyclic
response for both operating modes; hence it is possi- REFERENCES
ble that no significant structural effects are expected
from forced thermal recharging for the pile and soil Akrouch, G., Sanchez, M. & Briaud, J.L. 2014. Thermo-
mechanical behavior of energy piles in high plasticity
in the current study.
clays. Acta Geotechnica, 9, 399–412.
Intermittent operating modes with controlled forced Barry-Macaulay, D., Bouazza, A., Singh, R. M., Wang, B.
thermal recharging can be beneficial in terms of & Ranjith, P. G. 2013. Thermal conductivity of soils and
energy extraction and ground temperature balance for rocks from the Melbourne (Australia) region. Engineering
Geology, 164, 131–138.
geothermal energy piles since no significant structural
Bourne-Webb, P. J., Amatya, B., Soga, K., Amis, T., David-
effects are expected. Further studies on conventional son, C. & Payne, P. 2009. Energy pile test at Lam-
piles with optimized hybrid systems will give more beth College, London: geotechnical and thermodynamic
insight on coupled thermal behaviour of soil and pile. aspects of pile response to heat cycles. Geotechnique, 59,
237–248.
Dai, L., Li, S., Duanmu, L., Li, X., Shang, Y. & Dong,
ACKNOWLEDGEMENTS M. 2015. Experimental performance analysis of a solar
assisted ground source heat pump system under different
This research was supported under the Australian heating operation modes. Applied Thermal Engineering,
Research Council’s Linkage Projects funding scheme 75, 325–333.
567
Faizal, M., Bouazza, A. & Singh, R. M. 2015. An experi- Stewart, M. A. & McCartney, J. S. 2014. Centrifuge modeling
mental investigation of the influence of intermittent and of soil-structure interaction in energy foundations. Jour-
continuous operating modes on the thermal behaviour of nal of Geotechnical and Geoenvironmental Engineering,
a full scale geothermal energy pile (Under review). 140, 04013044.
Jalaluddin & Miyara, A. 2012. Thermal performance investi- Wang, B., Bouazza, A., Singh, R., Haberfield, C., Barry-
gation of several types of vertical ground heat exchangers Macaulay, D. & Baycan, S. 2015. Posttemperature effects
with different operation mode.AppliedThermal Engineer- on shaft capacity of a full-scale geothermal energy pile.
ing, 33–34, 167–174. Journal of Geotechnical and Geoenvironmental Engineer-
Laloui, L., Nuth, M. & Vulliet, L. 2006. Experimental ing, 0, 04014125.
and numerical investigations of the behaviour of a heat Wang, E., Fung, A. S., Qi, C. & Leong, W. H. 2012. Per-
exchanger pile. International Journal for Numerical and formance prediction of a hybrid solar ground-source heat
Analytical Methods in Geomechanics, 30, 763–781. pump system. Energy and Buildings, 47, 600–611.
Man, Y., Yang, H. & Wang, J. 2010. Study on hybrid Wang, X., Zheng, M., Zhang, W., Zhang, S. & Yang, T. 2010.
ground-coupled heat pump system for air-conditioning in Experimental study of a solar-assisted ground-coupled
hot-weather areas like Hong Kong. Applied Energy, 87, heat pump system with solar seasonal thermal storage in
2826–2833. severe cold areas. Energy and Buildings, 42, 2104–2110.
McCartney, J. S. & Murphy, K. D. 2012. Strain distributions in Wood, C. 2011. Energy piles for residential installations and
full-scale energy foundations. DFI Journal: The Journal other low rise buildings. Ground source live sustainable
of the Deep Foundations Institute, 6, 26–38. heating & cooling. Peterborough.
McCartney, J. S., Rosenberg, J. E. & Sultanova, A. 2010. Wood, C. J., Liu, H. & Riffat, S. B. 2010. Comparison of
Engineering performance of thermo-active foundations. a modelled and field tested piled ground heat exchanger
GeoTrends. system for a residential building and the simulated effect
Mimouni, T. 2014. Thermomechanical characterization of of assisted ground heat recharge. International Journal of
energy geostructures with emphasis on energy piles. PhD Low-Carbon Technologies, 5, 137–143.
thesis, École Polytechnique Fédérale de Lausanne. Yu, K. L., Singh, R. M., Bouazza, A. & Bui, H. H. 2014.
Murphy, K. D. & McCartney, J. 2014. Seasonal response of Evaluation of soil thermal properties through numeri-
energy foundations during building operation. Geotech- cal simulations of a heating test on a geothermal energy
nical and Geological Engineering, 1–14. pile. Proceedings of the 7th International Congress on
Murphy, K. D., McCartney, J. S. & Henry, K. S. 2014. Eval- Environmental Geotechnics.
uation of thermo-mechanical and thermal behavior of Yu, K. L., Singh, R. M., Bouazza, A. & Bui, H. H. 2015.
full-scale energy foundations. Acta Geotechnica, 1–17. Determining soil thermal conductivity through numeri-
Singh, R., Bouazza, A. & Wang, B. 2015. Near-field ground cal simulation of a heating test on a heat exchanger pile.
thermal response to heating of a geothermal energy pile: Geotechnical and Geological Engineering, 33, 239–252.
observations from a field test. Soils and Foundations.
568
Numerical investigation of the mechanical behaviour

of single energy piles and energy pile groups
C. Tsetoulidis, A. Naskos & K. Georgiadis

Aristotle University of Thessaloniki, Thessaloniki, Greece
ABSTRACT: The effect of thermal loading on the mechanical behaviour of single energy piles and energy piles
in pile groups is investigated. Finite element analyses are presented, in which both thermal and mechanical loads
are applied to the piles. The numerical modelling procedure is first validated against a documented in the literature
full-scale experiment. Subsequently, two numerical investigations are presented. In the first investigation, the
effect of thermal loading on the axial response of single piles is examined. The second numerical investigation
involves axially loaded 3 × 3 pile groups with a rigid cap, in which either the centre pile or all piles of the group
are thermo-active. It is shown that the effect of thermal loading on the axial pile forces is significant and depends
on the number of piles that are thermally active in the pile group and on the relative values of the coefficient of
thermal expansion of the pile and the ground.
1 INTRODUCTION for improvement. Recently, numerical investigations

have been published that investigate aspects of the
One of the most unexploited sources of renewable behaviour of energy piles (e.g. Ozudogru et al., 2014,
energy in our days, that escapes the boundaries of Salciarini et al., 2014, Bourne-Webb et al., 2015, Di
classical geothermal applications, is the use of the Donna et al., 2016). This paper presents such a numer-
almost constant temperature of the ground in the super- ical study using finite element analysis conducted in
ficial layers of the earth’s crust (Brandl, 2006). Up order to investigate the behaviour of single energy piles
until recently, the few attempts that had been made and energy pile groups under axial loading.
to harvest this form of energy were mostly focused
on “energy wells” or “energy boreholes” acting as
heat exchangers. However, based on this idea, over 2 FINITE ELEMENT ANALYSES
the last three decades, this technique has also been
implemented on foundation elements with significant The numerical analyses presented in this paper were
environmental and economic results. The most com- performed with the finite element program ABAQUS.
mon type of such a foundation element in practice Two finite element meshes were used in the analy-
is the energy pile (e.g. “Dock Midfield” station of ses; one for the single pile analyses and one for the
the airport in Zurich, Norddeutsche Landesbank in 3 × 3 pile group analyses. The pile dimensions and soil
Hamburg, The School Center in Berlin, etc.). stratigraphy of the Lambeth College, London energy
Nonetheless, despite the applications that have pile test (Amis et al., 2008 and Bourne-Webb et al.,
taken place so far, there is still a lot of uncertainty 2009) were used. Therefore, in all cases considered,
on the issue of the mechanical behaviour of such ele- the pile(s) had a length of L = 23 m and a diameter
ments, due to the considerable thermal loading that is of D = 0.6 m, while the stratigraphy consisted of a
applied on top of the mechanical loading of the piles. top 1.5 m thick Made Ground layer, overlaying a 4 m
Investigations on the thermo-mechanical behaviour of thick layer of Terrace Deposits, which is underlain by
energy piles that involve thermal pile tests or mon- a deep London Clay layer. The same centre-to-centre
itoring, have been conducted in Bad Schallerbach, pile spacing of 1.8 m (3 pile diameters) was considered
Austria (Brandl, 1998 & 2006), Frankfurt am Main, in all pile group analyses.
Germany (Ennigkeit & Katzenbach, 2001), Lausanne, The soil behaviour was modelled as linear elastic-
Switzerland (Laloui et al., 2006) and London, UK perfectly plastic with a Mohr-Coulomb failure crite-
(Amis et al., 2008 and Bourne-Webb et al., 2009). rion and a Drucker-Prager plastic potential surface.
Fundamental theoretical models have been pro- The mechanical material properties of the ground
posed to describe the mechanical behaviour of energy layers are presented in Table 1. Typical thermal prop-
piles (e.g. Bourne-Webb et al., 2011 and Amatya et al., erties were selected for the materials and are shown in
2012), which incorporate some critical basic aspects. Table 2.
However, they are based on limited data and several The piles were modelled as linear elastic with
assumptions and therefore, there appears to be room Young’s modulus Ep = 2.9 · 107 kPa, Poisson’s ratio
569
Table 1. Mechanical properties of soil layers.
E v φ c or su γ
◦
Layer MPa – kPa kN/m3
Made Ground 20 0.2 20 0 19

Terrace 20 0.2 30 0 19
Deposits London Clay 800 · su 0.49 0 70+7z 20
E: Young’s modulus, v: Poisson’s ratio, φ: angle of shearing

resistance, c: effective cohesion, su : undrained shear strength,
γ: bulk unit weight.
Table 2. Thermal properties of soil layers.
λ C α
Layer W/m/K J/kg/K m/m/K Figure 2. Finite element mesh for the pile group analyses.
Made Ground 2 1200 5.6 · 10−6 and 12 m in the mesh used for the pile group analyses.
Terrace 2 1200 5.6 · 10−6 First order hexahedral elements were used for the rest
Deposits London Clay 1.5 1000 10 · 10−6 of the mesh. In total, the finite element mesh for the
single pile analyses (Fig. 1) consisted of 31382 ele-
λ: thermal conductivity, C: heat capacity, α: coefficient of
thermal expansion. ments, while that for the pile group analyses consisted
of 92274 elements.
The vertical boundaries of both meshes were fixed
in the normal direction, while the bottom boundary
was fixed in all directions. A constant temperature
field of 18◦ C was prescribed for all elements of the
meshes at the beginning of each analysis and the same
temperature was prescribed as a boundary condition
at all boundaries throughout the analyses.
Thermal loading was modelled in separate transient
heat transfer analyses in which temperature variations
were prescribed to the pile elements. The calculated
temperature fields for each time increment of the heat
transfer analysis, were then imposed on the same time
steps of the mechanical FE analyses in order to obtain
the mechanical response to the temperature changes.
The numerical procedure adopted in this study was
validated by simulating the Lambeth College thermal
Figure 1. Finite element mesh for the single pile analyses. pile test in London presented by Bourne-Webb et al.
(2009). In this test, a single pile was subjected to a
vp = 0.1, unit weight γ = 25 kN/m3 , thermal conduc- cooling and a heating stage under constant working
tivity λp = 1.7 W/m/K, heat capacity Cp = 960 J/kg/K load. The test procedure was simulated numerically
and coefficient of thermal expansion αp = 8.5 · using the finite element mesh of Figure 1 and the
10−6 m/m/K. numerical procedure described above and the cal-
The pile soil interface was modelled as frictional culated pile head settlements were compared to the
with a friction angle equal to the angle of shearing measurements reported in Bourne-Webb et al. (2009).
resistance of the adjacent soil, for the top layers, and A relatively good agreement between the numerical
equal to 28◦ for the London Clay layer. A maximum results and the measurements was observed; the dif-
(cut-off) shear stress equal to a · su was specified for ferences between the computed and experimental pile
the part of the interface in the London Clay layer, head settlements were 1%, 2% and 15%, at the end
where a is the adhesion factor. The adhesion factor of the mechanical loading (1200 kN working load),
was calculated according to Kulhawy (1991). cooling (T = −18◦ C) and heating (T = +28◦ C),
The finite element mesh used for the single pile respectively.
analyses is shown in Figure 1, while the mesh used
for the pile group analyses is shown in Figure 2. Both
3 SINGLE ENERGY PILE
meshes have a diameter of 48 m (80 × D) and a height
of 35 m (1.4 × L). Second order hexahedral elements
3.1 Overview of analyses
were used for the pile(s) and for a cylindrical region
around the pile or pile group. This region had a diame- The mechanical behaviour of an axially loaded sin-
ter of 3 m in the mesh used for the single pile analyses gle energy pile is first investigated. Thermal loading
570
cycles of cooling and heating are applied to the pile
at different constant axial working loads. For reasons
of comparison, the cooling and heating cycles that
are applied are the same as those in the simulated
experiment at Lambeth College. The effect of thermal
loading on the axial bearing capacity is first investi-
gated, by simulating pile loading to failure at different
applied temperatures. The interface shear stress and
axial load distributions at different mechanical and
thermal loads are then examined, by simulating cool-
ing and heating cycles under constant axial working
loads. Finally, the pile settlements are discussed.
3.2 Effect of thermal loading on the pile axial

bearing capacity
Three loading cases are examined: (i) axial pile loading
to failure with no thermal load, (ii) application of a
1200 kN axial load at the pile head followed by cooling Figure 3. Axial load-settlement curves.
of the pile to 0◦ C at constant pile head axial load and
finally loading of the pile to failure under constant
temperature, and (iii) application of a 1200 kN axial
load at the pile head, heating of the pile to 30◦ C at
constant pile head axial load and finally loading of the
pile to failure under constant temperature.
The calculated ultimate loads were 4785 kN,
4641 kN and 4847 kN for cases (i), (ii) and (iii), respec-
tively.The load-displacement curves for the three cases
are shown in Figure 3. It can be observed that, as
expected, the bearing capacity of the pile does not
change significantly due to the thermal load (approxi-
mately 3% in cooling and 1.3% in heating). The small
decrease due to cooling and increase due to heating
can be attributed to the material contraction or expan-
sion due to cooling or heating, respectively. During
the cooling stage, the pile and soil contract, which
causes the normal stresses on the pile interface to
decrease and consequently the interface shear strength
to also decrease. Similarly, heating of the pile and soil
results in expansion and increase of the normal inter-
face stresses and interface shear strength. Obviously, Figure 4. Axial force distributions with depth.
this decrease or increase of the interface shear strength
leads to the respective decrease or increase of the axial
bearing capacity due to cooling or heating that is seen The results of the analyses indicate that heating and
in Figure 3. cooling significantly affect the magnitude and dis-
Furthermore, it is also worth mentioning that a tribution of the axial stresses along an energy pile.
greater vertical displacement is required to mobilise Figure 4 presents the axial force variation with depth
the full axial pile resistance during cooling than dur- for the twelve cases analysed. Differences of up to
ing heating. This occurs due to the pile contraction due 670 kN in the axial force can be observed between
to cooling that is added to the pile settlement due to the the mechanically loaded and the thermally loaded pile
mechanical load. The reverse mechanism takes place cases.
in the heating stage. It can be seen in Figure 4 that heating causes the
axial loads to increase along the whole length of the
pile. This increase is clearly caused by the pile axial
3.3 Effect of thermal loading on the pile axial
thermal expansion. This expansion is restrained at the
force distribution
base and also by the interface shear stresses, which
Four different pile-head axial load cases are consid- lead to the observed increase in the axial forces.
ered (the pile self-weight is applied in all cases): zero, On the other hand, as illustrated in the same figure,
1200 kN, 2400 kN and 3600 kN. For all four loads, the the effect of cooling on the axial pile forces depends
same three cases of thermal loading (under constant on the applied mechanical load at the pile head. At
pile head mechanical load) as above are examined: no zero or low axial pile head loads, the axial forces, as
thermal loading, cooling to 0◦ C and heating to 30◦ C. anticipated, decrease due to the pile contraction and
571
Table 3. Calculated pile settlements (mm).
End of End of
Axial pile Mechanical cooling heating
head load only loading cycle cycle
No load +0.2 +2.7 −0.95

1200 kN +2.2 +5.3 +1.6
2400 kN +5.2 +8.5 +4.7
3600 kN +9 +12.2 +8.6
At failure +12.3 +15.9 +11.8
3.4 Pile settlements

The calculated pile settlements at the end of the initial
pile head axial loading, the end of cooling to 0◦ C and
the end of heating to 30◦ C are presented in Table 3
for zero axial load (the pile is loaded with only its
self-weight), 1200 kN, 2400 kN and 3600 kN.
Figure 5. Interface shear stress distributions with depth for It can be seen in Table 3 that, as expected, cooling
no thermal loading and after cooling. of the pile causes further settlement of the pile, which
varies between 2.5 mm and 3.6 mm for the different
pile head load levels. The opposite effect is observed
the resulting interface shear stress changes. However,
during heating, where an upward movement of the pile
an increase in the axial forces over the whole length
between 3.6 mm and 4.1 mm is observed. While the
of the pile is observed for the case of a large pile head
magnitude of the calculated displacements is small, the
load of 3600 kN. At the intermediate pile head load
percentage increase or decrease of the settlement due
of 2400 kN a reversal in the behaviour is observed at
to cooling or heating, respectively, is significant. Set-
mid-length of the pile. The axial forces increase in the
tlement changes due to thermal loading can be greater
upper half of the pile and decrease in the lower part.
than the initial settlement due to the axial pile head
This unexpected response of the pile at cooling
load. For an axial load of 1200 kN an increase of 141%
can be attributed to the pile–soil interface behaviour,
is calculated during cooling and a decrease of 168%
and specifically, as explained in more detail below,
is calculated during heating. The respective percent-
to the reduction in the interface shear strength due to
age differences for 2400 kN applied pile head load are
cooling.
63% and 73%.
As the applied pile head load increases the interface
shear stresses also increase. When the applied load
is small the interface stresses are much smaller than
4 ENERGY PILES IN A PILE GROUP
the interface shear strength. However, as the applied
load becomes larger, the interface shear strength is
4.1 Overview of analyses
mobilized. Initially, this occurs only in the upper part
of the pile, but as the pile head load increases larger In the second part of this study, the behaviour of pile
parts of the pile interface are mobilized. As seen in groups in which one or all of the piles are thermo-
Figure 5, the interface shear strength under mechanical active was investigated. A 9-pile group was considered
only loading is mobilised in the top 12 m and 7.5 m for in a square 3 × 3 grid with a 3 pile diameter centre-
pile head loads of 3600 kN and 2400 kN, respectively. to-centre spacing (= 0.6 m × 3 = 1.8 m spacing). All
In contrast, no shear strength mobilization is evident the piles of the group were connected with a rigid pile
for zero applied load and for 1200 kN applied load. cap. The same phases of cooling and heating, as in
Cooling, generally tends to cause an increase in the the single pile investigation, were applied. However,
interface shear stresses in the upper part of the pile and two separate cases were analysed. In the first case,
a decrease in the lower pile. This is due to the cooling- the whole group was composed of thermo-active piles
induced contraction and is observed clearly for the and in the second, only the centre pile was thermo-
self-weight only case in Figure 5. At large pile head active. In the following paragraphs, the axial force
loads, however, where the maximum shear stress has distributions along the pile length for the centre, a
already been reached in part of the pile due to mechani- corner and a side pile of the 3 × 3 pile group, are pre-
cal loading, this shear stress increase in the upper part sented and discussed. The pile group settlements are
of the pile is not possible. Moreover, because of the also considered.
decrease in the shear strength due to cooling (which
was discussed above), the shear stresses in this part of
4.2 Thermo-active pile group
the pile are in fact forced to decrease. This has an effect
similar to heating and leads to the observed increase The case in which all the piles of the 3 × 3 pile group
in the axial forces (Figure 4). are thermo-active is first presented. The pile group
572
Figure 6. Thermo-active pile group: Axial force distribu-
tions with depth for the centre pile, a corner pile and a side Figure 7. Temperature fields at the end of cooling: (a) single
pile of the pile group. energy pile, (b) thermo-active pile group.
This can be attributed to the end restraints imposed on

self-weight was first activated and then a 7200 kN the piles at the pile bases and the pile cap and also
vertical load was applied at the centre of the pile cap. on the restraint imposed on the ground surface by the
Following the load application, all the piles were sub- pile cap. These effects counteract the tendency of the
jected to a cooling cycle (cooling from 18◦ C to 0◦ C) piles to contract compared to the soil, and leads to an
and then a heating cycle (heating to 30◦ C). increase of the pile axial forces along the whole length
The axial force distributions for the centre pile, a of all the piles. This increase is smaller than that during
side pile and a corner pile, obtained from the numerical cooling in the part of the piles that is situated in the
analyses, are presented in Figure 6. A first interesting London Clay layer.
finding, which can be seen in this figure, is that in It can also be observed in Figure 6 that while cooling
the case of cooling a significant increase of the axial has no effect on the loads carried by each pile of the
forces is observed throughout the whole length of all pile group (the axial forces at the pile heads remain
the piles. This is in contrast with the response of the unchanged), heating causes a redistribution of the axial
single energy pile discussed above, in which the pile- loads. The pile head load increases at the centre pile
soil interface shear stress changes generated by the and the four side piles and decreases at the four corner
pile contraction due to cooling caused a decrease of piles.
the pile axial forces.
The reason for the observed increase in the pile axial
forces during cooling lies mostly in the thermal dila-
4.3 Thermo-active centre pile in pile group
tion coefficient α of the London Clay, which is greater
than that of the pile. Because of this, the greater part The similar procedure as above was followed. The pile
of the soil (the London Clay layer) contracts more than group was loaded with 7200 kN and then a cooling and
the pile for the same temperature increment. a thermal cycle were applied only to the centre pile.
As illustrated in Figure 7a, in the case of a single It is evident from Figure 8 that the behavior of the
pile, only a small region around the pile is affected pile group is significantly different in the case of one
by the temperature reduction in the pile and the aver- thermo-active pile compared to the case in which all
age temperature change in this region is much smaller piles are thermo-active (discussed above). In this case,
than that of the pile. In contrast, in the case of a pile as expected, heating and cooling have the most signif-
group in which all piles are thermo-active (Figure 7b), icant effect on the axial forces of the centre pile. Much
a large volume of soil that contains the whole pile smaller axial force changes are observed in the other
group undergoes approximately the same temperature piles of the group. These force changes are of oppo-
reduction as that of the piles. Consequently, although site sign to the force changes of the centre pile, which
the piles obviously contract during cooling, they con- indicates that the behaviour of the rest of the piles of
tract less than the ground, leading to interface shear the pile group is similar to that of reaction piles.
stress changes that increase the pile axial forces. During cooling, the centre pile axial compressive
Interestingly, the axial forces also increase in all forces decrease remarkably, to the point that they
piles during heating, despite the fact that the London become tensile along almost the entire length of the
Clay layer can potentially dilate more than the pile. pile. The reason for this decrease is that as the centre
573
Table 4. Calculated pile group settlements (mm).
End of End of
Mechanical cooling heating
only loading cycle cycle
All piles +2.6 +7.6 +3.9

thermo-active
Only centre pile +2.6 +3.6 +2.9
thermo-active
As expected, the impact of cooling or heating is

much more significant when all piles are thermo-
active. A further settlement of 4 mm (154% increase)
is computed during cooling of all piles of the group,
compared to 1 mm (38% increase) when only the
centre pile is cooled.
It is also interesting to observe that the extra set-
Figure 8. Thermo-active centre pile in pile group: Axial tlement induced by cooling the pile(s) is not reversed
force distributions with depth for the centre pile, a corner during heating. In fact at the end of heating, the final
pile and a side pile of the pile group. settlement is always greater than the initial settle-
ment of 2.6 mm. This indicates that significant plastic
deformation occurs during cooling.
pile is cooled, it tends to contract, however, movement
is restricted at the pile cap, while contraction is also
restricted by the resisting side friction. As a result, ten- 5 CONCLUSIONS
sile axial stresses develop along its length. In contrast
with the fully thermo-active pile group, in this case, An investigation of the mechanical behaviour of
the temperature field of the model resembles more energy piles and energy pile groups with the use of
that of a single energy pile, with a much smaller area finite elements analysis was presented.
of soil where the temperature changes decidedly. As a The behaviour of single energy piles subjected to
result, the overall pile contraction is greater than the cooling and heating cycles under different constant
contraction of the ground. axial pile head loads was first examined. It was found
In the heating phase, the centre pile axial com- that heating and cooling have a small effect on the
pressive forces increase significantly. The reverse pile axial bearing capacity. The small changes in the
behaviour to that of the cooling stage is observed. As bearing capacity can be attributed to the increase or
the pile tends to expand, the interface side friction, and decrease of the normal stresses acting on the pile-soil
the base and the cap restraints, prevent its deformation interface during heating or cooling, respectively.
and consequently generate further compressive axial It was also shown that thermal loading cycles have a
stresses along the pile length. significant effect on the axial forces and their distribu-
As mentioned above, opposite sign axial force tion along the pile length. Heating of the pile increased
changes to those of the centre pile are observed in the axial forces along its entire length. The response
the rest of the piles of the group. Their axial forces to cooling was found to depend on the level of shear
increase when the centre pile is cooled and decrease strength mobilisation at the pile-soil interface and con-
when the centre pile is heated. This is essentially a sequently on the applied axial load at the pile head. For
redistribution of pile head loads within the pile group. low pile loads cooling caused a decrease of the axial
The rest of the piles in the group are largely unaffected forces. However, for higher pile axial loads, an increase
by the temperature changes around the centre pile and of the compressive forces was observed in part of the
simply react to the axial load applied to the pile cap by pile and for very high loads along the entire length of
the centre pile during each thermal load phase. the pile.
In the second part of the study, the behaviour of a
3 × 3 energy pile group was examined. Analyses were
performed in which either all piles of the pile group
4.4 Pile group settlements or only the centre pile of the pile group were thermo-
The calculated pile settlements at the end of the initial active.
pile group axial loading, the end of cooling to 0◦ C and A significant increase in the axial compressive
the end of heating to 30◦ C are presented in Table 4 for forces of the piles both in the cooling and in the heat-
the case in which all piles of the group are thermo- ing phases was observed when all piles were modelled
active and the case in which only the centre pile is as thermo-active. This increase in the cooling phase
thermo-active. can be attributed to the fact that the London Clay was
574
assumed to have a higher coefficient of thermal expan- Bourne-Webb, P.J., Bodas Freitas, T.M., Freitas Assuncao,
sion than the pile concrete and to the much higher R.M. 2015. Soil–pile thermal interactions in energy foun-
temperature changes in the ground, compared to those dations. Géotechnique 66(2): 167–171.
computed for a single pile. The axial compressive Brandl, H. 1998. Energy piles and diaphragm walls for heat
transfer from and into the ground. Proceedings of the 3rd
forces also increased during heating, because of the International Geotechnical Seminar, Deep Foundations
imposed end constraints. on Bored and Auger Piles (BAP III), University of Ghent,
When only the centre pile was modelled as thermo- Belgium, vol. 1, pp. 37–60.
active, very large decreases or increases of the axial Brandl, H. 2006. Energy foundations and other thermo-active
forces in the centre pile were found for cooling and ground structures. Géotechnique 56(2): 81–122.
heating, respectively. The decrease of the axial forces Di Donna, A., Rotta Loria, A.F. & Laloui, L. 2016. Numer-
during cooling were so large that the greater part of ical study of the response of a group of energy piles
the pile went into tension. The rest of the piles of the under different combinations of thermo-mechanical loads.
group were much less affected. Computers and Geotechnics 72: 126–142.
Ennigkeit, A., Katzenbach, R. 2001. The double use of piles
as foundation and heat exchanging elements. Proceedings
of XVth International Conference on Soil Mechanics and
Geotechnical Engineering (ICSMGE), Istanbul Turkey,
REFERENCES 27–31 August 2001, Vol. 2, pp. 893–896.
Kulhawy, F.H. _1991. Drilled shaft foundations. Foundation
Amatya, B.L., Soga, K., Bourne-Webb, P.J., Amis, T. &
engineering handbook, Chapman and Hill, New York.
Laloui, L. 2012. Thermo-mechanical behaviour of energy
Laloui, L., Nuth, M. & Vulliet, L. 2006. Experimental
piles. Géotechnique 62(6): 503–519.
and numerical investigations of the behaviour of a heat
Amis, T., Bourne-Webb, P., Davidson, C., Amatya, B. & Soga,
exchanger pile. Int. J. Numer. Analyt. Methods Geomech.
K. 2008. The effects of heating and cooling energy piles
30(8): 763–781.
under working load at Lambeth College, UK. Proceedings
Ozudogru, T.Y., Olgun, C.G. & Arson, C.F. 2015. Analy-
of the 33rd Annual and 11th International Conference of
sis of Friction Induced Thermo-Mechanical Stresses on
the Deep Foundations Institute, New York, USA.
a Heat Exchanger Pile in Isothermal Soil. Geotechnical
Bourne-Webb, P.J., Amatya, B.L., Soga, K., Amis, T.,
and Geological Engineering 33(2): 357–371.
Davidson, C. & Payne, P. 2009. Energy pile test at Lam-
Salciarini, D., Ronchi, F., Cattoni, E. & Tamagnini C.
beth College, London: Geotechnical and thermodynamic
2015. Thermomechanical Effects Induced by Energy Piles
aspects of pile response to heat cycles. Géotechnique
Operation in a Small Piled Raft. Int. J. Geomech. 15(2).
59(3): 237–248.
Bourne-Webb, P.J., Amatya, B.L. & Soga, K. 2011. A frame-
work for understanding energy pile behavior. Proceed-
ings of the Institution of Civil Engineers – Geotechnical
Engineering 166(2): 170–177.
575
Thermo-mechanical behavior of small-scale energy pile in dry sand
V.T. Nguyen
Universite Paris-Est, Laboratoire Navier (UMR CNRS 8205), France
Hanoi University of Mining and Geology, Vietnam
A.M. Tang & J.M. Pereira

Universite Paris-Est, Laboratoire Navier (UMR CNRS 8205), France
ABSTRACT: A small-scale energy pile has been developed in laboratory to study its thermo-mechanical
behaviour under thermal cycles. A model pile (20 mm external diameter) of ultimate axial load capacity of
500 N was installed in dry sand. During the experiment, the pile was initially loaded with a series of axial load,
which range from 0 N to 300 N, with an increment of 50 N. For each step of axial loading, a thermal cycle was
applied immediately after the pile head settlement due to axial loading had stabilised. The pile behaviour is
discussed in terms of cumulative thermal settlement, effect of creep during thermo-mechanical loading stage
and also the change of mobilized friction along the pile wall through the analysis of the evaluation of axial force
profiles along the thermo-mechanical loading stages.
1 INTRODUCTION The present paper focuses on experiments on a

small-scale pile in laboratory to study its thermo-
Energy piles, or heat exchanger piles, have a dual mechanical behaviour under thermal cycles.
function: providing support for overhead structures as
a conventional pile foundation and exchanging heat
with the ground for purpose of heating and/or cool-
ing the building. Energy piles have been used in some 2 EXPERIMENTAL METHOD
European countries during the last two decades. This
technique has gained encouraging efficiencies in the 2.1 Experiment setup
use of renewable energy in modern cities and con-
tributed to the reduction of CO2 emission Brandl 1998; A pile model of 20 mm external diameter and 600 mm
Brandl 2006, Laloui et al. 2006, Adam & Markiewicz length was installed in a dry sand sample of 548 mm
2009, Pahud & Hubbuch 2007). However, experiences inner diameter and 900 mm height (Figures 1 & 2). The
and installation are not homogeneous across countries pile model is an aluminum tube with an internal diame-
due to the lack of design standards, including rules ter of 18 mm and sealed at the bottom. The pile surface
defining geotechnical design of energy piles. was coated with sand to simulate the roughness of the
Many studies have been performed to investigate pile. The sand used in this study (Fontainebleau sand)
on the thermo-mechanical behaviour of energy piles has the following physical properties: particle den-
(Laloui et al. 1999, Bourne-Webb et al. 2009, Lam sity ρs = 2.67 Mg/m3 ; maximal void ratio emax = 0.94;
et al. 2009, Laloui 2011, McCartney & Rosenberg minimal void ratio emin = 0.54; and median grain size
2011, Amatya et al. 2012, Kalantidou et al. 2012, D50 = 0.23 mm.
Yavari et al. 2013, Murphy et al. 2014, Ng et al. Installation process began with the compaction of
2014, Suryatriyastuti et al. 2013, Di Donna & Laloui the first two layers of 100 mm thickness, then two lay-
2014, Yavari et al. 2014, Wang et al. 2014, Saggu & ers of 50 mm beneath the pile toe. Pile was then placed
Chakraborty 2014, Akrouch et al. 2014, Goode in its position inside the soil container and was fixed
et al. 2014, Stewart & Mccartney 2014, Mimouni & by a steel bar that was blocked on the above surface of
Laloui 2015, Wang et al. 2016). The results evi- soil container. Finally, the other sand layers of 100 mm
denced the effects of temperature change on the were compacted around the pile. The soil was com-
pile-soil interaction and the mobilized resistance of pacted manually by using a wooden tamper, at a unit
piles. weight of 15.1 kN/m3 .
577
Figure 2. View of the experimental setup.
according to the process described above. Two exper-

iments T1 and T2 were performed to investigate the
load-displacement curve of the pile under mechani-
cal loading. The test procedure follows The French
standard NF P 94-150-1 (1999) (Figure 3). In the
preparation step, the pile was firstly loaded to 0.1Qmax
for 15 minutes and then unloaded to remove the dis-
turbed settlement component due to soil compaction
Figure 1. Pile model and sensors distribution profile. related to the pile installation process (Qmax is the ulti-
mate load which induced 10%D settlement of pile
head, D being the external diameter of pile). In the
experiment, the pile was loaded in steps and subjected
During the compaction, three temperature sensors to two loading cycles to obtain a load-displacement
and two pressure gauges were installed as showed in relationship.
Figure 1. The two pressure sensors (P12 & P20) were The loading paths of the thermo-mechanical exper-
located at 50 mm below the pile toe. P12 measures iments (T3 and T4) are shown in Figure 4. For test T3,
the total horizontal pressure and P20 measures the after the installation, the pile was first heated from 20◦
total vertical pressure of soil. The three soil temper- to 21◦ C for 60 min and then cooled to 19◦ C for 60 min.
ature sensors (S5, S6 & S7) are placed at the same The second thermal cycle was repeated using the same
level (300 mm below the soil surface) and at three procedure as for the first cycle, after a recovery period
distances from the pile axis, 20, 40 and 80 mm, respec- allowing returning to the initial thermal condition. The
tively. To measure the pile axial strain, five strain total duration of the two cycles equals to 6 hours. Once
gauges were distributed along the pile length. Three the temperature reached the initial value (20◦ C), the
displacement transducers (LVDT) were used to mea- pile head load was increased to 100 N. When the set-
sure the pile head displacement, and a load cell is tlement stabilized (after 60 min), two thermal cycles
placed at the pile top to record the pile head load. similar to the previous ones were applied. By doing
By using a water tank placed above the pile head, the so, the pile will be loaded to 300 N. For test T4, pile
pile head load is controlled by the water level in the was first heated from 20◦ to 21◦ C for 120 min and then
tank. A metallic U-tube, connected to a temperature- cooled to 19◦ C for 120 min without head load, after
controlled bath, is placed inside the pile tube for a recovery period allowing to return to initial tem-
heating and cooling the pile. The soil container was perature. The total duration of one cycle equals to 8
thermally isolated to avoid heat exchange with the hours. Once the temperature reached the initial value
ambient air. A similar pile model could be found in (20◦ C), the pile was loaded to 100 N, when the settle-
the studies of Kalantidou et al. (2012) and Yavari et al. ment reached stabilization (after 120 min), one thermal
(2014a). cycle similar to the previous ones was applied. Then,
at the end of thermal cycle, the pile was unloaded for
overnight. The next day, pile was loaded again to 150 N
2.2 Test program
for 120 min, then a thermal cycle was applied to the
In this study, four experiments have been performed. pile. Similar loading processes will be applied until
After each experiment, the pile model was reinstalled the loading step of 300 N.
578
Figure 3. Mechanical loading experiments (Test T1 & T2).
3.1 Mechanical behavior

In this section, the effects of mechanical loading phase
on the pile behaviou will be discussed.At first, effect of
time on pile settlement was evaluated over a period of
60 minutes according to the French Recommendation
NF P 94-150-1 (1999). In each constant loading step,
the pile settled immediately and then the settlement
continues over time but with a lower rate. In the exper-
iment, the results were recorded continuously with an
interval of 1 minute. Figure 5 shows the time evolu-
tion of pile head settlement in logarithmic scale over
a 60 minutes period for test T1. It can be seen that,
after only a few minutes of loading, the pile settled at
a lower rate with a linear trend. Therefore, creep rate
of pile in each loading step could be evaluated by the
slope of the curve settlement-time in log scale. Fol-
lowing the French Recommendation NF P 94-150-1
(1999), under each constant loading, creep of pile can
be measured by the following index:
where αn = creep rate; n = step number; S60 =

settlement of pile head at time t = 60 minutes;
S10 = settlement of pile head at time t = 10 minutes.
Figure 6 shows creep rate results of the four tests.
The settlement obtained in the four tests is illus-
trated by the load-settlement curve in Figure 7. The
pile head settlement value of each loading step in T3
is determined after one hour of mechanical load and it
includes the settlement caused by thermal loading of
pile in previous loading steps. For test T4, pile head set-
tlement was measured after 1 hour of each mechanical Figure 4. Thermo-mechanical loading of T3(a) and T4(b).
579
Figure 8. Axial force profile for test T3.
load step and plus the irreversible settlement of the pre-

Figure 5. Settlement under constant mechanical load versus vious loading steps. From the load-settlement curve,
logarithm of time (test T1). the ultimate bearing capacity of pile (corresponding to
a normalised settlement of 10% of pile diameter) can
be approximately estimated at 500 N. In addition, the
results of the four tests show a good repeatability of
the applied procedure.
Although experiments T3 and T4 comprised cou-
pled thermo-mechanical load, the results in Figure 7
showed a similar load-settlement curve for all four
tests. This result implies that the contribution of ther-
mal cycles in each step of loading to cumulative
displacement of pile is negligible when compared to
that of mechanical loading.
The axial deformations that were recorded from
the strains gauges were used to measure axial force
of along the pile. Figure 8 shows a result of axial
force distribution along the pile length obtained at the
end of each loading step. Axial load of pile at levels
of 100, 200, 300, 400 and 500 mm from the surface
Figure 6. Creep behaviour of pile. were derived from the gauges G5, G4, G3, G2 and
G1 respectively. Axial load value on the surface is
used directly from the load cell. This result indicates
that most of pile head load is transmitted to the soil
via mobilized friction on the pile shaft, while a small
amount of head load is transmitted into the soil through
the pile toe. This latter value could be estimated by
extrapolation from the last two measured forces at G1
and G2. This value is about 20% of the head load. It
can be said that the mobilized friction on the pile shaft
is significant, and that the mobilized resistance of the
pile depends mainly on skin friction.
3.2 Thermo-mechanical behavior of pile under

constant mechanical loading
Figure 9 shows temperature as measured at various
locations in the soil and pile. After 1 hour of heating
or cooling, temperature change of pile seems to be sta-
ble. Soil temperature change is non-significant. This
is due to the small thermal loading applied to the pile,
about ±1◦ C. However, after each heating and cool-
Figure 7. Load-displacement curve. ing cycle, cumulative settlement of pile continues to
580
Figure 9. Temperature and settlement versus time at 100 N
loading step (test T3).
Figure 11. Pile head settlement in T4. (a) axial loading

equal to 0 N; (b) 100 N; (c) 150 N; (d) 200 N; (e) 250 N and
(f) 300 N.
increase. Settlement of pile in cooling phase is signif-

icant, but in the heating phase the pile head heave is
almost negligible. At the end of each thermal loading
step, the settlement of pile tends to increase when the
temperature of pile reached stable value.
Figure 10 and 11 represent incremental settlement
of pile during heating-cooling cycles. Accordingly, the
initial point corresponding to the start of each ther-
mal cycle is set to zero displacement. The final point
corresponds to the end of thermal cycles.
It was observed that the pile behavior during the
thermal cycle phase depends on the mechanical load
applied to the pile. When the mechanical loading was
null (0 N), the displacement of the pile head was the
same as the pile’s thermal expansion curve, which
expresses the deformation of a pile restrained at its toe
but free in other direction under a temperature change
(Figure 10a, 11a). During the following loading steps,
the pile settlement during cooling phase was not com-
pensated by heave during heating so that permanent
and cumulating settlement was observed. In test T3
Figure 10. Pile head settlement in T3. (a) axial loading (Figure 10), the slope of the settlement curve under
equal to 0 N; (b) 100 N; (c) 150 N; (d) 200 N; (e) 250 N and the first cooling phase is the same as in the second
(f) 300 N. cooling phase. It is also similar to the slope of the pile
581
Figure 12. Axial force along the pile length in test T3
thermal expansion curve. It can be said that, the set-

tlement of pile of two cooling phase increases linearly
with temperature reduction. In case of Figure 10d, the
irreversible settlement at the end of second cooling
phase continues to increase after the temperature has
stabilized. Pile uplift in heating phase was observed
only in case of Figure 10b, c, but very small. In test
T4, time of heating/cooling phase is twice as large as
the one of heating/cooling phase in T3, and the dis-
placement result is quite similar to the T3. However,
in the case of Figure 11d, e, f the settlement increases
rapidly at the beginning of cooling phase, and then at
the end of cooling phase it continues to increase after
the temperature has stabilized. This could be explained
by the effect of creep behaviour, it means the settle- Figure 13. Decoupling thermal and mechanical effects on
ment of pile would include thermal contraction and pile head displacement.
creep settlement.
The effect of temperature cycles on the pile head
displacement is quite small, about 0.2% when the pile Another aspect was also examined in this study: the
worked under a specified load. effect of thermal cycles and the effect of creep on the
Axial force measured during test T3 is shown in Fig- pile displacement during thermo-mechanical loading
ure 12. As it can be seen in the first step in which pile stages. The thermal settlement in each test is obtained
was heated and cooled without axial loading, the effect by removing the mechanical settlement of the pile
of cyclic loading on the axial force along the pile length under the corresponding load. The result in Figure 5
does not a clear trend. Axial force increases during shows that the effect of time in log scale on the pile
cooling phase and during the heating this value does settlement is linear when pile worked under a constant
not seem to change from the previous state. However, head load. For this reason the mechanical settlement
during the following steps the effect of thermal cycles during each loading step in the thermo-mechanical
was quite clear, with an increase during heating, and period can be derived from the creep rate. Finally, from
a decrease during cooling. This result can be seen as the results in Figure 10 and 11, the cumulative settle-
reasonable for this floating pile, which transmits most ment of the pile could be measured to decouple the
of head load to the soil through skin friction. Actually, thermal and mechanical effects, as shown in Figure
this phenomenon is similar to the experimental results 13. Although the creep value in T3 and T4 increases as
on the full-scale energy pile model of Bourne-Webb the pile head load increases, this relationship is not lin-
et al. (2009), Laloui (2011). ear. The creep rate increased insignificantly when pile
The influence of thermal cycle on the evolution of worked under small head load, while under higher head
axial force can be seen more clearly at the middle of load the creep rate increased significantly. It should
the pile length, while at the pile base this influence be noted that, the settlement of pile caused by creep
seems insignificant. A similar result was obtained in behaviour increases with time, the duration of a cycle
experiment T4. loading of test T4 is longer than that of test T3, so that
582
after one cycle loading a greater settlement obtained Kalantidou, A. et al., 2012. Preliminary study on the mechan-
in test T4. The irreversible settlement related to ther- ical behaviour of heat exchanger pile in physical model.
mal cycles was larger than the effect of creep during Géotechnique, 62(11), pp.1047–1051.
thermo-mechanical loading stage. However, the evolu- Laloui et al., 1999. In-situ thermo-mechanical load test on
a heat exchanger pile. In 4th International confrence on
tion of cumulated thermal settlement with the loading Deep foundation practice. Singapore, pp. 273–279.
steps does not show a clear trend. Laloui, L., 2011. In Situ Testing of a Heat Exchanger
Although the thermal settlement of pile is very small Pile. Geo-Frontiers 2011 – Advances in Geotechnical
compared to the safety limit (10%D) more work is Engineering. ASCE, pp.410–419.
needed to confirm the effect of many cycles of loading Laloui, L., Nuth, M. & Vulliet, L., 2006. Experimental
on the behavior of energy pile in the long term. and numerical investigations of the behaviour of a heat
exchanger pile. International Journal for Numerical and
Analytical Methods in Geomechanics, 30(8), pp.763–781.
4 CONCLUSIONS Lam, S.Y. et al., 2009. Centrifuge and numerical modeling
of axial load effects on piles in consolidating ground.
Canadian Geotechnical Journal, 46(1), pp.10–24.
This paper presents an experimental study of the McCartney, J.S. & Rosenberg, J.E., 2011. Impact of Heat
behavior of small-scale energy pile in dry sand. Four Exchange on Side Shear in Thermo-Active Foundations.
tests focused on the influence of thermal cycles on Geo-Frontiers 2011 © ASCE 2011, pp.488–498.
the thermo-mechanical behavior of pile. The following Mimouni, T. & Laloui, L., 2015. Behaviour of a group of
conclusions can be drawn: energy piles. , 17(May), pp.1–17.
Murphy, K.D., McCartney, J.S. & Henry, K.S., 2014. Eval-
– The maximum thermal cumulating settlement of uation of thermo-mechanical and thermal behavior of
pile head is very small: about 0.25%D. full-scale energy foundations. Acta Geotechnica, 10(2),
– The axial force along the pile increases in heating pp.179–195.
phase and decreases in cooling phase, and the most NF P 94-150-1, 1999. Essai statique de pieu isolé sous un
obvious change was found near the middle of pile effort axial., pp.1–28.
length. This phenomenon is compatible with the Ng, C.W.W. et al., 2014. Centrifuge modelling of energy piles
case of a floating pile which transmits the head load subjected to heating and cooling cycles in clay. Géotech-
nique Letters, 4(October–December), pp.310–316.
to the ground through the skin friction. Pahud, D. & Hubbuch, M., 2007. Measured Thermal Perfor-
– The irreversible settlement related to thermal cycles mances of the Energy Pile System of the Dock Midfield
was larger than the effect of creep during thermo- at Zürich Airport. In Procedding European Geothermal
mechanical loading stage, and depends on the Congress 2007. Unterhaching, Germany, pp. 1–7.
mechanical load applied to the pile. Saggu, R. & Chakraborty, T., 2014. Cyclic Thermo-
Mechanical Analysis of Energy Piles in Sand. Geotech-
nical and Geological Engineering, 33(2), pp.321–342.
Stewart, M.A. & Mccartney, J.S., 2014. Centrifuge Modeling
REFERENCES of Soil-Structure Interaction in Energy Foundations. Jour-
nal of Geotechnical and Geoenvironmental Engineering,
Adam, D. & Markiewicz, R., 2009. Energy from earth-
140(4), pp.1–11.
coupled structures, foundations, tunnels and sewers.
Suryatriyastuti, M.E., Mroueh, H. & Burlon, S., 2013. A
Géotechnique, 59(3), pp. 229–236.
load transfer approach for studying the cyclic behavior
Akrouch, G.A., Sánchez, M. & Briaud, J.-L., 2014. Thermo-
of thermo-active piles. Computers and Geotechnics, 55,
pp.378–391.
clays. Acta Geotechnica, 9(3), pp. 399–412.
Wang, B. et al., 2014. Posttemperature Effects on Shaft
Amatya, B.L. et al., 2012. Thermo-mechanical behaviour of
Capacity of a Full-Scale Geothermal Energy Pile. Jour-
energy piles. Geotechnique, 62(6), pp. 503–519.
nal of Geotechnical and Geoenvironmental Engineering,
Bourne-Webb, P.J. et al., 2009. Energy pile test at Lam-
pp.1–12.
beth College, London: geotechnical and thermodynamic
Wang, C. et al., 2016. Model tests of energy piles with and
aspects of pile response to heat cycles. Géotechnique,
without a vertical load. Environmental Geotechnics,
59(3), pp. 237–248.
pp.1–11.
Brandl, H., 2006. Energy foundations and other thermo-
Yavari, N. et al., 2014. Experimental study on the mechan-
active ground structures. Géotechnique, 56(2), pp. 81–
ical behaviour of a heat exchanger pile using physical
122.
modelling. Acta Geotechnica, pp.385–398.
Brandl, H., 1998. Energy piles for heating and cooling
Yavari, N. et al., 2013. A simple method for numerical
of buildings. In 7th International confrence exihibition
modelling of mechanical behaviour of an energy pile.
Piling and deep foundations. Vienna, pp. 341–346.
Geotechnique letters, pp.1–6.
Di Donna, A. & Laloui, L., 2014. Numerical analysis of the
geotechnical behaviour of energy piles.
Goode, J.C., Zhang, M. & Mccartney, J.S., 2014. Cen-
trifuge modelling of energy foundations in sand. In
The 8th international conference on Physical modelling
in Geomechanics (ICPMG2014). Perth, Australia, pp.
729–735.
583
Long term monitoring of CFA energy pile schemes in the UK
F.A. Loveridge & W. Powrie

University of Southampton, Southampton, UK
T. Amis
GI Energy, UK
M. Wischy & J. Kiauk

Siemens
ABSTRACT: Energy pile schemes involve the use of structural foundations as heat exchangers in a ground
source heat pump system. Such schemes are attractive, as they reduce energy consumption compared with
traditional building heating and cooling systems. As energy prices increase and governments introduce subsidies
they are also proving increasingly economically attractive. Additionally, energy piles can contribute to reducing
the carbon dioxide emissions associated with a development. However, this approach to heating and cooling
building remains relatively novel and the lack of published long term performance data remains a barrier to
further implementation. Two issues remain to be addressed by long term monitoring. First, the need for a
database of operational energy piles schemes were the energy performance is proven over many years. Secondly,
availability of long term datasets of pile thermal behavior that can be used to validate design approaches and
tools and hence encourage less conservative design practices. This paper presents the initial results from a study
aimed at tackling these issues through long term instrumentation and monitoring of two energy pile schemes in
the United Kingdom.
1 INTRODUCTION data from building energy management systems to

provide the opportunity to assess both the whole sys-
Energy pile schemes involve the use of structural foun- tem performance and the pile thermal behaviour. The
dations as heat exchangers in a ground source heat latter is particularly important for understanding mod-
pump system. Such schemes are attractive, as they elling and design approaches and their application to
reduce energy consumption compared with traditional energy piles.
building heating and cooling systems.As energy prices This paper provides details of the two monitoring
increase and governments introduce subsidies they sites and some initial results from the schemes so far.
are also proving increasingly economically attractive. Lessons learnt from both the installation of instrumen-
Additionally, energy piles can contribute to reduc- tation and the resulting data are also considered.
ing the carbon dioxide emissions associated with a
development.
Energy piles have been in operation in Europe since 2 THE CRYSTAL, EAST LONDON
the 1980’s (Brandl, 2006). Some notable case studies
include Zurich Airport (Pahud & Hubbach, 2007), One The first scheme to be instrumented was Siemens’
New Change in London (Amis & Loveridge, 2014) landmark sustainable building in East London, known
and Keble College Oxford (Suckling & Smith, 2002, as The Crystal (Figure 1). The Crystal is a multi-use
Nicholson et al, 2014). However, despite an increase development and contains an interactive exhibition of
in constructed cases in recent years, relatively little sustainable technologies as well as office space and
performance data are available, especially for the long conference facilities. It has been designed to be an all-
term (Bourne-Webb, 2013). This holds back validation electric building and utilises solar thermal and ground
of thermal design approaches, and limits the value of source heat pumps to generate all the thermal energy
demonstration projects. needed by the development. Photovoltaics also gen-
Since 2011, the University of Southampton and GI erate electricity to reduce reliance on a more carbon
Energy have been working in partnership to develop intensive national supply.
monitoring sites for energy pile schemes. The project The source side of the ground source heat pump
has seen the instrumentation of energy piles beneath system comprises 160 pile heat exchangers and a field
two buildings in the United Kingdom. Temperature of 36 deep boreholes. The piles are 600 mm, 750 mm
sensors installed in the foundations are combined with or 1200 mm in diameter and were constructed using
585
Table 1. Ground conditions at the Crystal.
Strata Description Depth
Made Ground Fine to coarse brick and 3.3 m

concrete gravel; soft to firm
black sandy gravelly clay
Figure 1. The Crystal building in East London. Alluvium Very soft clayey silt, 6.3 m
sandy clay and peat
River Terrace Medium dense silty fine to 11.2 m
contiguous flight auger (CFA) techniques to approx- Deposits coarse sand and fine to coarse
imately 21m depth. Each pile incorporates a pair of gravel (mainly flint)
High Density Polyethylene (HDPE) U-pipes, which London Clay Stiff thinly laminated fissured 23.5 m
were inserted into the centre of the pile after the pile silty clay with silt partings
cage had been plunged into the concrete. The U-pipes Lambeth Silty fine sand and fissured 43.3 m
were then connected together in series and usually Group silty clay
joined into a single circuit with a neighbouring pile, Thanet Sands Very dense, slightly 56.1 m
before the pipework continued to the manifold cham- silty fine sand
ber and then on via larger header pipes to the plant Chalk Medium density >150 m
(Grade B3) chalk
room for connection to the heat pumps.
Each borehole is 150 m deep and contains a single
HDPE U-pipe. Backfill material is gravel through the
permeable strata encountered in the lower two thirds of Table 2. Depths of thermistors installed within the instru-
the hole and grout over the upper third (refer to Section mented pile at The Crystal.
2.1 below for geological information).
The Crystal has been operational since 2012, but it Depth Below Pile Cut Off Level (m)
has taken several years for some of the relevant build-
Thermistor Level Central String Outer Strings
ing monitoring data to come on stream. Hence it is
only now that initial interpretation of these data can 1 0.7 0.75
commence. 2 3.6 3.25
3 7.1 6.6
4 11.1 –
2.1 Ground conditions
5 15.1 –
The site is underlain by a sequence of London Basin 6 19.1 –
deposits. The piles are founded in the London Clay,
but also pass through a significant thickness of super-
ficial and man-made deposits (Table 1). The boreholes
damage occurred to buried cabling during construction
continue through the full sequence of strata and are
which unfortunately meant that data access to these
installed for approximately two thirds of their length
instruments was lost and could not be re-established.
in the aquifer formed by the Chalk and the overlying
All instrumentation was wired into a datalogging
Thanet Sands. As the site is located near to the conflu-
system located in the energy centre building adjacent
ence of the Thames and the River Lea in east London,
to The Crystal. The logger can be accessed remotely
the groundwater table is close to the ground surface,
to allow monitoring of the system from offsite. Tem-
at the base of the Made Ground.
perature data have been collected since the summer of
2012.
2.2 Pile and borehole instrumentation
One 1200 mm diameter pile near the north east corner
of the building was selected for monitoring. The pile
2.3 Building monitoring
was equipped with five thermistor strings. One of these
was attached to the central bundle of four pipes (two The Crystal is equipped with a comprehensive build-
U-tubes), themselves inserted into the pile attached ing energy management system (BEMS). Three heat
to a 40 mm steel bar for stiffness. The U-pipes, the meters record the heating and cooling energy deliv-
steel bar and the thermistor strings were installed to a ered to the building. Two more heat meters record the
depth of 20 m within the pile. The other four thermis- amount of heat exchanged with the ground, one for
tor strings were attached at equal spacings around the the pile ground loop circuit and one for the borehole
circumference of the steel reinforcing cage (Table 2). ground loop circuit. These data have been available
As the pile cage only extends to 8.5 m below the pile since the spring of 2013, except the heat meter for the
cut off level it was not possible to extend the outer borehole ground loop which was not communicating
thermistor strings over the full 21 m pile depth. correctly with the BEMS until 2015. Additionally, the
Thermistor strings were also installed in one of the ground loop heat meters require further adjustment to
borehole heat exchangers and two additional moni- differentiate between the direction of heat transfer to
toring boreholes (Loveridge et al., 2013). However, (or from) the ground.
586
Figure 3. Average operational pile temperatures since 2012
(data gap in the autumn of 2015 due to data logger
malfunction).
Figure 2. Undisturbed ground temperatures at The Crystal

site. Figure 4. Energy demand of The Crystal during 2014 and
2015. Winter months taken as January to April and November
A further heat meter was installed by the to December; summer months taken as May to October.
University of Southampton on the pipe circuit for the
instrumented energy pile during construction. How-
ever, this meter has also had substantial communica- years of monitoring. In the summer of 2015 the aver-
tion difficulties with the logging equipment and no age temperature of the thermistors installed on the pile
data are available from this meter at the present time. cage was approximately two degrees warmer than at
the start of monitoring in summer 2012, when it was
close to the undisturbed condition (Figure 2).
2.4 Results
The rise in pile temperature is likely to represent a
Initial results from the pile monitoring during and net transfer of heat from the building to the ground dur-
immediately following construction are included in ing the first three years of operation. However, when
Loveridge & Powrie (2013a). These data, along with the overall thermal energy demand of The Crystal is
further measurements from a thermal response test analysed, the heating and cooling profile is close to bal-
carried out in an instrumented borehole at the site anced (Figure 4), with annual demand approximately
(Loveridge et al., 2013), show that the undisturbed 550 MWh per year each for both heating and cooling.
ground temperature is approximately 12.8◦ C over the On the other hand, the instantaneous rate of energy
depth range relevant to the energy piles (refer to demand does differ between heating and cooling (Fig-
Figure 2). ure 5). The maximum power demand in heating is
For the purposes of initial analysis of performance, 399 kW, while in cooling it is 572 kW. This may be the
pile temperatures can be taken as approximately con- cause of the net heat injection to the ground. Another
stant with depth. Consequently, average operational factor that cannot be assessed yet is the relative con-
pile temperatures for the string attached to the pipes tribution of the piles and boreholes to the heating and
and those installed on the reinforcing cage are pre- cooling demand. For example, the piles could be tak-
sented in Figure 3. Using averages removes local ing a different heating/cooling ratio compared to the
variations due to any asymmetric installation of the borehole field. Finally, the weather during the period
pile cage and central pipes and allows the overall trends of assessment must be considered to place any trends
to be identified. However, it does obscure axial effects, in context. Table 3 shows that during the last four years
which are potentially important, but are outside the temperatures in the UK have been warmer that the long
scope of this paper. term average by several degrees. This may also have
The data show large fluctuations in temperature affected the rise in temperature seen within the instru-
at the centre of the pile, from approximately 7◦ C to mented pile during this period by triggering a higher
26◦ C. However, the temperature range at the steel cage, than normal cooling demand.
nearer the edge of pile, is much less, approximately Global coefficient of performance (COP) data for
10◦ C to 18◦ C. It can also be seen that there is an over- the ground source heat pump system at The Crys-
all rise in the temperature of the pile over the three tal can be taken as an initial indicator of the energy
587
Figure 5. Heating and cooling power required at The Crystal
since summer 2013.
Table 3. UK mean annual temperature anomalies compared

with 1961 to 1990 long term average.
Year Temperature Anomaly (◦ C)
2015 +0.9
2014 +1.6
2013 +0.5
2012 +0.4
Figure 6. Artist’s impression of the completed building
at 22 Station Road (source: http://www.cb1cambridge.eu/
22-station-road).
efficieny of the system. Current data suggest COPs
between 2.5 and 3.0 depending on the time of year.
However, the operational control system is still being Clay. Boreholes from the site describe the Gault as
optimized, hence these values are expected to increase initially a firm to stiff slightly sandy slightly silty
in the future. calcareous CLAY to approximately 6.5 m below pile
cut off level. Beneath this the Gault becomes a stiff to
very stiff laminated and fissured calcareous CLAY.
3 22 STATION ROAD, CAMBRIDGE The groundwater table at the site is relatively high,
with water strikes during borehole drilling rising to
22 Station Road (Figure 6) is the new building for approximately 2 m below pile cut-off level.
Mott MacDonald and Birketts in Cambridge. It forms
part of the extensive redevelopment of the zone around
Cambridge Station known as CB1. The building com- 3.2 Instrumentation
prises a basement carpark and a further six floors of
office space. A balanced circuit of up to six piles of 600 mm diam-
The building is founded on 81 CFA piles of 450 mm eter energy piles, each 20 m long, was instrumented
diameter and 68 CFA piles of 600 mm diameter. Pile using thermistor strings. Each of the six piles was
lengths are between 20 m and 25 m. Each pile is equipped with two four-thermistor strings attached
equipped with a single polyethylene pipe U-loop. As to the U-loop pipes and four two-thermistor strings
at The Crystal, the pipes were plunged into the centre attached to four of the six main bars on the reinforce-
of the pile following insertion of the steel reinforcing ment cage (Figures 7 & 8). Table 4 gives a summary
cage. In this case the pipes were attached to a 32 mm of the thermistor levels. As at The Crystal the steel
diameter steel bar for weight and stiffness. reinforcing cage did not extend the full depth of the
The U-loop pipes from individual energy piles pile. Therefore fewer thermistor levels were installed
were connected together to form a series circuit with on the steel cage compared with the central pipes.
the pipes from adjacent piles. Each circuit contains Heat meters were installed by GI Energy on the
between four and six piles and is connected to the six-pile circuit where it reached the manifold in the
header pipes at the manifold located in the building basement, to enable further monitoring of the perfor-
basement. The header pipes then run to the upper floor mance of the ground energy system as a whole and to
of the building, where the heat pumps and other plant record:
are located. • Thermal Energy Power (kW)
• Cumulative Thermal Energy delivered (kWh)
3.1 Ground conditions • Flow Rate (L/sec)
• Flow Temperature (◦ C)
The ground conditions at the site are Made Ground,
• Return Temperature (◦ C)
overlying River Terrace Deposits and Gault Clay.
Owing to the construction of a new basement and the All the GI Energy monitoring points and the Univer-
lowering of the original ground level, the piles were sity of Southampton thermistor strings were connected
constructed over their full length through the Gault via remote panels to the same monitoring system to
588
Figure 7. Detail of thermistor strings installed on the steel
cage at 22 Station Road.
Figure 8. Schematic arrangement of thermistor strings

within the instrumented piles at 22 Station Road.
Table 4. Depths of thermistors installed within the piles at Figure 9. Temperatures within the six instrumented piles at
22 Station Road. 22 Station Road at three dates during construction.
Height Above Pile Toe (m)

Thermistor proceeding the following month. By September 2014
Level Central Strings Outer Strings the instrument cables and polyethylene pipes were all
routed to the manifold located in the basement and
1 17.5–18.0 17.5–18.0 the building was being constructed around this. Sub-
2 13.5–14.0 13.5–14.0 sequent readings were taken in the autumn of 2015
3 7.5–8.0 7.5–8.0 as the cables were wired into the remote panels. By
4 1.5–2.0 1.5–2.0 this time the structure was largely complete, although
fit-out and commissioning work continued throughout
the remainder of the year.
allow web-based desktop reading of the data from any Spot readings from the thermistor strings at three
location. points during construction of the building are shown in
Figure 9. Shortly after construction of the piles (May
2014) there was an elevated temperature of 14◦ C to
3.3 Initial data
16◦ C in the lower part of the piles, probably reflecting
The remote monitoring system is not yet available as the residual heat of hydration of the concrete. At the
final commissioning of the system is ongoing. How- top of the piles the temperature readings were reduced,
ever, spot readings from the thermistors were taken at reflecting the time of year (late spring to early sum-
various points during construction (Figure 9).The piles mer) when summer air temperatures had not yet had
were cast during April 2014, with the slab construction a chance to be reflected in the near surface layers.
589
By September 2014 the upper part of the piles have
increased in temperature as the average air tempera-
ture also increased over the summer, but the values
recorded in the lower part of the pile have reduced as
the heat of hydration in the pile dissipates. The fol-
lowing spring, in April 2015, when it can be assumed
that the heat of hydration has fully dissipated, the tem-
perature near the base of the piles was around 13◦ C.
This can be taken as representative of undisturbed con-
ditions. Lower temperatures were recorded near the
ground surface, reflecting the cooler air temperatures Figure 10. Hourly cooling demand at the Crystal during two
in the preceding winter period. days in July 2015.
available at Station Road, it just would not have

4 DISCUSSION been possible within a tight construction programme
on a very constricted site to bring in additional
4.1 Lessons learnt from installations plant to achieve this goal. Consequently this addi-
tional information remains recommended for future
The Crystal instrumentation was arranged very close monitoring sites where possible.
to construction which left few options for optimising
the thermistor arrangements within the framework of
the thermal design of the system. By contrast, involve-
4.2 Key observations from operation
ment with the development of the new building at 22
Station Road commenced much earlier in the project The building at 22 Station Road has yet to experience
cycle. This allowed greater planning time and as a an operational period. However, operational data are
result the entire six pile circuit was instrumented rather now available for a number of years at The Crystal,
than just one of the piles within a circuit as at the Crys- although not all of the building monitoring equipment
tal.This additional planning also allowed integration of has been online for the entire duration. Nonetheless
the pile and building monitoring systems into a single some interesting observations can be drawn out.
remote access arrangement. As shown in Section 2.4 above, despite a nomi-
Both case studies illustrate the need to allow suffi- nally balanced energy requirement it is possible for
cient redundancy within any instrumentation scheme the peak power requirements to be uneven. This may
implemented within the framework of a live con- be responsible for the gradual increase in the pile
struction project. In only instrumenting one pile at temperatures over the last three years, however, other
the Crystal, it was fortunate that most of the ther- factors may also be relevant and additional data are
mistors have remained fully functional for over three required to analyse these. Nonetheless, the observa-
years. However, unfortunately, access to the bore- tion still demonstrates an important point with regard
hole heat exchanger instrumentation was lost due to to the design of ground source heat pump systems.
cable breaks, caused by construction which were sub- The data show how lumped demand can be misleading
sequently buried before they could be located and when real demand varies over short timescales. This is
rectified. Having six piles instrumented at Station illustrated in Figure 10, which shows some detail from
Road has meant the significance of some instrument Figure 5 during two days in July 2015. Consequently,
losses during pile breakout has been reduced. A further design of complex ground source heat pump systems
lesson regarding redundancy learnt from the Crystal should be carried out using hourly energy demand data
and applied to Station Road was the inclusion of two (kWh per hour) – effectively instantaneous power.
strings attached to the central pipes installed within Differences in predicted system capacities when
each pile. using longer duration rather than hourly demand data
One advantage of the building arrangement at the has recently been highlighted by Zhang (2016), who
Station Road site is that the manifold is located within used hourly, monthly and yearly demand data for a
the basement of the building, making it potentially Westminster wide district heating study and compared
much more accessible in the future. The correspond- the results. In this case the use of monthly or yearly
ing manifold at The Crystal is contained within a deep demand data was found to be non-conservative in
manhole in the grounds of the building. While this is terms of estimating the proportion of the city that could
accessible, additional health and safety considerations be included within the district heating scheme. The dif-
of the deep manhole make maintaining instrumenta- ference in predicted capacity between using monthly
tion in this area (the heat meter and a logger remote or yearly data and using hourly data was between 10%
panel) more challenging. and 15%.
At both sites it would have been beneficial to The second important observation from The Crystal
have included in-ground instrumentation to provide data is the extent of the temperature gradient across
additional temperature data beyond those measured the instrumented pile. This varies throughout the year
within the piles. Despite the additional planning time (Figure 11) and reaches a peak value of approximately
590
However, sometimes the apparent resistance value
is smaller than this and often the value is larger. The
larger values reflect periods when the thermal load is
reducing so that the temperature close to the pipes is
falling more quickly than at the pile edge. In these
cases as the power drops, the temperature difference
also drops but not as rapidly. This causes the apparent
increase in resistance.
The implications of the absence of a thermal steady
state within the pile are described by Loveridge &
Figure 11. Temperature difference between the thermistor Powrie (2013b). Assuming a steady state where none
strings installed on the pile cage and those on the central is present can lead to the over-estimation of the tem-
U-pipes. Positive difference indicates a cooler pile centre and perature changes that will occur within the pile and
hence heat extraction from the pile and ground. the ground. This means that the true capacity of the
associated ground source heat pump system will be
underestimated during design.
5 SUMMARY
This paper presents initial results from two energy

pile monitoring schemes in the UK. Such schemes
are essential to allow better understanding of the long
term benefits of using energy piles with ground source
heat pump schemes and also to allow verification of
Figure 12. Apparent thermal resistance of the pile between
appropriate and rigorous design approaches.
the pipes and the steel cage, calculated for two days in July Initial data from the first scheme are showing the
2015. importance of understanding the nature of the applied
thermal loads for such systems. This is in terms of both
the overall seasonal energy balance and the short term
8 degrees. Peak values typically occur at times of peak variation in demand.
demand. The additional superimposed variability is a The short term variation in demand also contributes
reflection of two factors, first the variable thermal load to the variable temperature observed within the pile.
(Figure 5 & Figure 10) and secondly the fact that the This causes fluctuation in the temperature difference
pile is not at a thermal steady state as assumed by many across the pile and hence also the dynamic thermal
design approaches. resistance. Such transient behavior must be accounted
The thermal load and pile temperature data can be for in design to prevent under-estimation of the ground
used to quantify the degree of steady state within the source heat pump system capacity.
pile. For clarity, considering only the same two days
in July (Figure 10), Figure 12 shows corresponding
changes in apparent thermal resistance. Thermal resis- ACKNOWLEDGEMENTS
tance is the ratio of the thermal power applied to the
pile and the temperature difference across the pile. Installation and maintenance of field instrumenta-
In this instance it has been calculated using the pile tion is time consuming and requires complete support
temperature sensors and hence is not the full pile resis- from the contractors and designers involved in the
tance, just the resistance between the outside of the schemes. At The Crystal, in addition to full support
pipes and the pile cage. In this case the term “appar- from Siemens, we are extremely grateful for assistance
ent” thermal resistance has been used since strictly and cooperation from Balfour Beatty Ground Engi-
speaking resistance is a steady state concept and within neering, Arup, Foundation Developments Limited and
this large diameter pile a steady state will not be IGS. At 22 Station Road we were supported by Mott
present. Nonetheless, consideration of the changes in MacDonald, Central Piling and Wates.
values of apparent resistance allows some interesting The fieldwork presented in this paper was funded
observations to be made. by the Engineering and Physical Sciences Research
The results show that the calculated apparent resis- Council (research grant number EP/H049010/1) and
tance in the pile concrete is far from constant (as would was further supported by Mott MacDonald, Cementa-
be the case if there was a thermal steady state). When tion Skanska, GolderAssociations & WJ Groundwater.
there is a sustained period of heat injection (Figure 10) The lead author is additionally funded by the Royal
the apparent resistance stabilises at approximately to Academy of Engineering, with support from Mott
0.1 mK/W, and this could be close to a steady state MacDonald,Arup, Cementation Skanska, WJ Ground-
value. Certainly it is not an unreasonable value for a water, GI Energy, Neo-Energy, Mimer Energy and the
pile of this size and type (e.g. SIA, 2005). British Geological Survey.
591
REFERENCES Nicholson, D., Smith, P., Bowers, G. A., Cuceoglu, F., Olgun,
C. G., McCartney, J. S., Henry, K., Meyer, L. L. &
Amis, T. & Loveridge, F. (2014) Energy piles and other ther- Loveridge, F. A. (2014a) Environmental impact calcu-
mal foundations for GSHP – Developments in UK practice lations, life cycle cost analysis, DFI Journal, 8 (2),
and research, The REHVA European HVAC Journal, 130–146.
January, 2014, 32–35. Pahud, D. & Hubbach, M. (2007) Measured Thermal Perfor-
Bourne-Webb, P. (2013) Observed response of energy mances of the Energy Pile System of the Dock Midfield
geostructures, In: Laloui & Di Donna (Eds), Energy at Zürich Airport, Proceedings European Geothermal
Geostructures, Wiley, London. pp. 45–77. Congress 2007, Unterhaching, Germany, 30 May–1
Brandl, H. (2006) Energy foundations and other thermo June 2007.
active ground structures, Geotechnique, 56 (2), SIA (2005) Utilisation de la chaleur du sol par des ouvrages
81–122. de fondation et de soutenement en beton, Guide pour la
Loveridge, F. & Powrie, W. (2013a) Performance of Piled conception, la realisation et la maintenance, Swiss Soci-
Foundations Used as Heat Exchangers, In: Proceedings ety of Engineers and Architects, Documentation D 0190.
of the 18th International Conference for Soil Mechanics (In French).
and Geotechnical Engineering, Paris, France, September Suckling, T. P. & Smith, P. E. H. (2002) Environmentally
2–5, 2013. Friendly Geothermal Piles At Keble College, Oxford, UK.
Loveridge, F. & Powrie, W. (2013b) Temperature response In: Proceedings of the Ninth International Conference on
functions (G-functions) for single pile heat exchangers, Piling and Deep Foundations, 2002, Nice, France. Deep
Energy, 57, 554–564. Foundations Institute, New Jersey, USA.
Loveridge, F., Holmes, G., Powrie, W. & Roberts, T. (2013) Zhang, Y. (2016) Application Potential of Shallow Geother-
Thermal response testing through the Chalk aquifer, Pro- mal Energy at City Scale. PhD Thesis, Department of
ceedings of the Institution of Civil Engineers Geotechnical Engineering, University of Cambridge.
Engineering, 166 (2), 197–210.
Met Office (2016) Climate summaries [online http://www.met
office.gov.uk/climate/uk/summaries; accessed 21st January
2016].
592
Numerical analysis of thermal cycling during a multi-stage

energy pile thermal response test
F. Cecinato & R. Piglialepre

University of Trento, Trento, Italy
F.A. Loveridge
University of Southampton, Southampton, UK
D. Nicholson
Arup, London, UK
ABSTRACT: Energy piles are emerging as convenient alternative to the more traditional Borehole Heat
Exchangers (BHEs) to provide heating/cooling to buildings, as they remove the need for special purpose exca-
vations and can accommodate more pipes, thus enhancing energy performance. However, their different aspect
ratio compared to BHEs requires different modelling tools and dedicated thermal response testing, to achieve
adequate thermal design. In this work, the results of an extended multi-stage Thermal Response Test (TRT)
carried out on a single energy pile installed in London Clay are presented in terms of both fluid temperature
data and concrete temperature, measured by vibrating wire strain gauges and optic fibre sensors. The results
are then explored in detail by means of a finite element numerical code, able to account for both convective
heat exchange in the fluid, between the fluid and the solids and transient heat diffusion in the concrete and the
ground. Analysis of the TRT field data shows that during the later stages of the test there is clear evidence of
cyclic changes in performance. Investigation of these effects using the numerical model raises the possibility
that there could be some alteration of the properties of the soil-pile contact during the test. Hypotheses for the
observed behaviour are tentatively put forward and discussed with work recommended to further investigate the
percieved phenomena.
1 INTRODUCTION pile are homogeneous and isotropic and that their

thermal properties are constant and independent of
Energy piles, serving the double function of founda- temperature.
tions and heat exchangers, have been proposed as a Thermal response testing is an in situ technique to
convenient alternative to borehole heat exchangers, determine the thermal properties of the ground and the
as they remove the requirement to make expensive thermal resistance of a borehole heat exchanger. It has
special purpose excavations. Furthermore, their com- also been applied to piles, although there are difficul-
paratively larger diameter means they can be expected ties with doing so for larger diameter piles (Loveridge
to have a greater energy capacity per drilled metre et al., 2014a). The same analysis methods and assump-
(Bozis et al., 2011). tions that are used for thermal design can be applied to
Approaches to the thermal design of energy piles the interpretation of thermal response tests, although
tend to be based on analytical (or empirical) methods in practice usually the simplest of analytical models,
developed for borehole ground heat exchangers. Such the line source, is adopted. This means that provision
methods assume that the dominant thermal process in of reliable thermal design parameters is also depen-
the ground is conduction and use analytical solutions dent on the assumptions of homogeneous and isotropic
to the diffusion equation to relate predicted tempera- conditions with temperature independent behaviour.
ture changes in the ground and the pile to the input of This paper will report on the case study of an
thermal loads in the form of the heating and cooling extended thermal response test of a small diame-
demand of the building. Typically transient solutions ter energy pile, which shows some indication that
are applied to determine the temperature changes in cyclic thermal performance could indicate tempera-
the ground (e.g Eskilson, 1987, Claesson & Javed, ture dependent behavior. The test results are explored
2011), while steady state solutions based on thermal by a combination of field data (Section 3) and numer-
resistance are used for the borehole or pile itself (e.g. ical simulation (Section 4), before the relevance and
Lamarche et al., 2010). In all these approaches it is significance of the findings are discussed and tenta-
assumed that the soil and the concrete that forms the tive conclusions drawn (Section 5). Before describing
593
this work the paper first presents a brief introduction
to thermal response testing (Section 2).
2 THERMAL RESPONSE TESTING
Thermal Response Testing (TRT) is an in situ investi-

gation technique developed for borehole heat exchang-
ers and subsequently extended for use with some
energy piles. The test aims to determine the thermal
conductivity of the ground and the thermal resistance
of the heat exchanger to provide input parameters for
thermal design.
In a typical test the ground heat exchanger is con-
nected, via its heat transfer pipes, to a number of
heaters and a circulation pump. Circulation of the Figure 1. The multi-stage thermal response testing, showing
heated fluid through the borehole or pile allows heat to heating power input at each stage.
be injected to the ground at constant rate. The temper-
ature change of the inlet and outlet fluid are monitored Ten days after grouting the pile, a multi-stage ther-
throughout the test and the results used with an ana- mal response test was carried out. Unusually the
lytical solution to the diffusion equation to calculate test comprised a number of different phases (Fig-
the ground thermal conductivity and heat exchanger ure 1) which were designed to allow investigation of
thermal resistance. As a minimum, tests include a heat the thermo-mechanical response of the pile (Ouyang,
injection stage, but additional information can also be 2014). After an initial circulation phase, a heat injec-
gained by continuing to circulate the fluid and moni- tion test (Stage 2) and recovery period (Stage 3) was
tor its temperature once the heaters have been switches followed by a heat extraction test (Stage 4) and recov-
off, i.e. during a recovery stage. ery period (Stage 5). Cyclic testing was then com-
There are several international and national guide- menced comprising two heat injection phases (Stages
lines for the test to encourage high quality testing 6 and 8) separated by heat extraction phases (Stages
and interpretation (Sanner et al, 2005; IGSHPA, 2007; 7 and 9). Each test stage followed directly from the
GSHPA, 2011). However, it must be recognised that preceding phase and measurements were maintained
the accuracy of the test depends on how well the throughout, except for a 4 day period in Stage 6 when
real in situ conditions reflect the assumptions that are repairs were being carried out to a faulty heating unit.
inherent in the analytical methods used in interpre- Consequently the heat injection of Stage 6 is only seen
tation. Well conducted tests carried out in boreholes at the latter part of this time period and there is an
in reasonably consistent ground conditions can be extended period of no recorded applied power prior to
expected to achieve an accuracy of around 10% (Witte, then (Figure 1). However, during the period when the
2013, Spitler & Gehlin, 2015). However, non-uniform power was not recorded, some heat was rejected to the
ground conditions can result in more significant errors system, as is shown by the concrete temperature data
developing (Signorelli et al, 2006). (refer to Figure 3 and Section 3.2).
3 CASE STUDY 3.1 Instrumentation & monitoring

The inlet and outlet heat exchanger fluid temperature
The thermal response test described in this study was Tin and Tout were measured throughout the test, except
carried out on a 300 mm diameter and 26.8 m length at the start of Stage 6 when the heating equipment
test pile constructed at a London development site was faulty. Moreover, the evolution of temperature
(Loveridge et al., 2014b). Beneath an initial concrete within the concrete Tc during the TRT was measured
slab, the pile was constructed through water saturated in two ways, namely (i) at four locations along the
London Clay over its entire length. The stratum was pile depth, by means of temperature sensors associated
described as firm to stiff grey clay and contained some with Vibrating Wire Strain Gauges (VWSG), and (ii)
layers of claystones. continuously along the pile depth by means of Optic
A single U-loop of heat transfer pipe was installed Fibre Sensors (OFS) placed at four positions (Fig-
in the hole to 26 m depth and backfilled with hard pile ure 2). Strain was also measured by both the VWSG
cementitious grout. The pipes were made from high and the OFS cables, however, consideration of this data
performance polyethylene material with an external is not within the scope of this study. Full description of
diameter of 32 mm and a wall thickness of 2.9 mm. these data and their interpretation are contained within
The pipes were installed separated by rigid spacers Ouyang (2014).
ensuring an even separation of the pipes and a centre
to centre spacing between the two legs of the U-tube of
3.2 Results
around 135 mm. The spacers also served as a housing
to mount numerous sensors which are described in The resulting fluid temperatures at the inlet and outlet
Section 3.1 below. in response to the applied cycles of heating (Figure 1)
594
Figure 3. Fluid and VWSG average concrete temperature
(at pile mid-height) data throughout the test.
Figure 2. Location of the temperature sensors within the

pile cross section.
are shown in Figure 3. The temperature increases in

response to heat injection and reduces in response to
heat extraction. It can also be seen that there is a bigger
difference between the inlet and outlet temperatures
when the applied heating power is greater. Figure 3
also shows as an example the average temperatures
recorded by the VWSGs at pile mid-height, which are Figure 4. Comparison between OFS and VWSG tem-
seen to have reduced amplitudes of variation to the perature measurements after application of moving average
fluid temperatures due to their greater offset from the filter to OFS data. The horizontal dotted lines denote the
pipes. approximate locations of spacers.
It can be also noticed in Figure 3 that during the
second part of stage 5 and the first part of stage 6 a certain degree of oscillation of OFS temperature
the concrete temperature varies unexpectedly, in a way measurements might have reflected the variable dis-
that suggests heat extraction, although this is neither tance of the cables from the pipes (i.e. changes of
reflected from the fluid temperatures nor from the position in different cross-sections along the pile).
applied power (Figure 1). A similar pattern is seen in
the OFS data, including on those OFS cables located on
4 NUMERICAL SIMULATION
the heat transfer pipes (Figure 2). This discrepancy is
the likely result of the heating system malfunctioning
4.1 Model description
in that period, leading us to presume that fluid temper-
ature measurements are not reliable during stage 5 and The numerical model (Cecinato and Loveridge 2015,
the beginning of stage 6. Cecinato et al. 2015) aims at realistically reproducing
Temperature measurements were also available the main processes behind the heat transfer phenomena
from within the concrete from the OFS placed at taking place in geothermal structures, namely thermal
approximately the same distance from the pipes. It convection between the fluid and the pipe wall, thermal
emerged that overall, the VWSG measurements are conduction in the grout/concrete, and thermal conduc-
more stable and consistent compared to OFS mea- tion in the ground. Convective heat transfer in the pore
surements, which appear more wavy. After applying a water is not considered. Hence, while the model is
moving average filter to the latter data, comparison of always applicable to low-permeability or dry geomate-
the two types of measurements showed adequate con- rials, it can only be applied to high-permeability water-
sistency. As an example, in Figure 4 VWSG and OFS saturated materials if the groundwater at a specific site
data are plotted versus depth for an instant at the end is known to be static.
of stage 6. It can be noticed that the two different OFS The transient heat convection-diffusion problem for
datasets tend to overlap in correspondence with the energy piles was solved using the Finite Element (FE)
locations of spacers. This could indicate that the OFS code ABAQUS to integrate 3D transient conduction
cables were not perfectly straight along the pile; hence through the solids, complemented by writing bespoke
595
Table 1. Material properties used in the simulation.
Material Parameters Values Units
Fluid Density 1000 kg/m3

Kinematic viscosity 1.00E–06 m2 /s
Specific heat capacity 4200 J/(kg K)
Mass flowrate 0.108 kg/s
Thermal conductivity 0.6 W/mK
Prandtl number 7
Concrete Density 2210 kg/m3
Pipes Thermal conductivity 0.385 W/mK
London Clay Density 1900 kg/m3
in the numerical model as a half domain exploiting

symmetry (Figure 5).
Validation of the model has been previously under-
taken using fluid data from Stages 2 to 5 of the case
study described above (Cecinato and Loveridge 2015,
Cecinato et al. 2015). The physical and thermal prop-
erties of the materials involved are given in Table 1.
Figure 5. FE mesh representing the pipes (schematised in
1D as a line of nodes), the pile and surrounding ground, with 4.2 Model development
temperature contours. Only half of the domain is considered,
to save computational time exploiting symmetry. Although in previous validation of the model it was
shown to perform well in reproducing the thermal
behaviour of the test pile, subsequent work had shown
that the inherent simplification of representing the
user subroutines to model the convective heat trans- pipes as 1D lines of nodes in the FE mesh might lead
fer at the fluid/solid interface and the temperature to less accurate calculations when pipes are placed
changes in the fluid along the pipe, represented within very close together (Loveridge and Cecinato, 2016). To
the FE mesh as lines of nodes, where the heat exchange investigate this potential effect, the numerical model
resulting from convection-diffusion in the pipes is was modified by changing both the FE mesh and the
concentrated. The 3D nature of the pipes (i.e. the user subroutines, to represent the exchanger pipes in
relevant diameter, in addition to length) is properly 3D. The scheme adopted involves representing each
accounted for via the user subroutines, by multiply- pipe in the pile cross-section with a set of nodes (2,
ing the heat flux corresponding to each pipe node by 4 or 8) distributed along the pipe’s circumference, so
the corresponding lateral surface area of each pipe that each node represents a part of the total pipe sur-
segment. face involved in the heat flux (Figure 6). On the other
To minimise computational time, while control- hand, in the original model a single node represents
ling the element aspect ratio and node spacing at key the whole pipe surface in a pile cross-section.
locations to warrant accuracy of heat exchange calcu- The modified model was then used to run the same
lations, the 3D FE mesh was created manually in an TRT simulations from Stage 2 to Stage 5, showing
axisymmetric fashion using 6-node linear triangular a significant improvement in the RMSE comparing
prism and 8-node linear brick diffusive heat trans- the simulated and measured outlet fluid temperature.
fer elements (Figure 5). The spacing of the nodes For example, for Stage 2, Table 2 shows the gen-
representing the ground was progressively increased eral decrease in RMSE, and hence increase in fit, as
towards the outer boundary, while the mesh was the number of pipe nodes is increased. This suggests
refined in the exchanger pipe and surrounding pile an increase in the accuracy of the simulation due to
areas. The size of the domain was determined by a more realistic representation of the heat flux spa-
numerical experimentation to be much larger than the tial distribution across the pipes. Additionally, more
area actually affected by heat transfer over the time symmetrical temperature contours further suggest a
range explored in this study. progressive improvement in the representative of the
To simulate the TRT case study, the inlet fluid heat fluxes (Figure 6).
temperature was prescribed as a function of time, as
a boundary condition for the analysis. At zero heat
4.3 Simulation of case study
flux an initial equilibrium temperature for both the
fluid and the concrete/ground conditions was spec- The model was now used to reproduce the entire tem-
ified. The TRT geometry was reproduced in detail perature history of the TRT case study over all stages
596
Figure 7. Comparison of measured and simulated outlet
fluid temperature during the TRT.
Figure 6. FE mesh representing a cross-section of the pipes

area with temperature contours during heat injection, for half
of the domain. The model was developed by representing
each pipe in the pile cross-section with a set of (a) 2, (b) 4
or (c) 8 nodes instead of a single node as originally proposed
(Figure 5).
Table 2. Improved model fit (Stage 2) when using increased

number of nodes to represent the fluid pipes.
No. Nodes 1 2 4 8
Figure 8. Comparison of measured and simulated out-
RMSE 0.240 0.202 0.169 0.196 let fluid temperature during the TRT, enlargement of later
stages 6–9.
of the test, using the 4 node 3D representation of Table 3. Model fit for the different test stages.
pipes. This provided an appropriate balance between
set up and computational time expended and the output Stage 2 3 4 5
accuracy. RMSE 0.169 0.205 0.504 N/A
The field measured fluid temperature was used as Stage 6 (2nd part) 7 8 9
RMSE 0.588 0.993 0.231 1.598
specified boundary condition and both the evolution
of the outlet fluid temperature (Tout ) and the concrete
temperate (Tc ) were used to assess fit of the model.
Based on the assessment of the OFS field data given (ii) generally at later TRT stages is observed, with spe-
in Section 3.2, only the VWSG data were used to cial regard to the heat extraction phases during thermal
assess Tc . cycling. A similar effect was observed by Loveridge
In Figure 7 the simulated and measured outlet fluid et al. (2014b) when fitting analytical solutions to the
temperature are reported for comparison, for all of same data. Those authors observed particular misfit
the TRT stages, leaving out the second part of stage of the analytical models in the last four stages of the
5 and the early part of stage 6 (due to the above test. It was suggested that this effect could be the
mentioned problems in that part of the TRT with results of the pile thermal resistance may not be con-
measurements reliability). The numerical simulation stant. It was hypothesised that this could be due to
effectively reproduces the field measurements for all increased contact resistance at the pile-soil boundary
considered stages of the TRT, however, it could be when the pile is cooled. However, the authors also
noticed that it does not approximate all stages with the showed that this hypothesis could not explain all the
same accuracy. In particular, the cyclic testing stages observed behaviours of the pile during the test.
(6 to 9) appear to be reproduced less precisely, as can In Figure 9 the VWSG-measured and simulated
be seen in an enlargement of these stages (Figure 8). (using the original 1D pipe scheme, to save computa-
To evaluate further the accuracy of the simulation, tional time yet providing adequate accuracy) concrete
the root mean square error (RMSE) of the residu- temperature values are reported, as an example, at
als was calculated, resulting in the values given in pile mid-height (13.8m depth) throughout the TRT.
Table 3. A tendency for the simulation accuracy to Simulation #1 was obtained using the measured inlet
worsen (i) for heat extraction phases (4, 7 and 9) and fluid temperature as boundary condition, and it can
597
Figure 10. Comparison of measured and simulated outlet
fluid temperature during stage 7.
Figure 9. Comparison of measured and simulated concrete
temperature during the TRT. Simulation #1 is obtained using
the measured Tout as boundary condition while Simulation
#2 is obtained with a corrected input during stages 5 and 6.
be seen that data are not adequately reproduced dur-

ing stages 5 and 6, during which the fluid temperature
measurements have been considered unreliable due a
temporary system breakdown. To be able to repro-
duce the evolution of Tc , the temperature input Tin
was modified during those stages generating a syn-
thetic temperature history, by means of numerical
back-analysis (Piglialepre 2016). This modified Tin
was used to run Simulation #2, which can adequately
reproduce Tc also during stages 5 and 6. The modified Figure 11. Comparison of measured and simulated outlet
Tin has a negligible effect on the subsequent stages fluid temperature during stage 9.
due to sufficient recovery time prior to the start of
heat injection in Stage 6. Table 4. Model fit for test stages 7 and 9 obtained with the
In general, the measured concrete temperature evo- original settings and with the altered interface settings.
lution is now correctly reproduced using Simulation #2
with synthetic input for Stages 5 and 6 (Figure 8). How- Stage 7 9
ever a worse fit during the last stages is still observed in
RMSE (original model) 0.993 1.598
both simualtions, with special reference to heat extrac- RMSE (altered interface) 0.844 0.703
tion stage 7 (field data were not available for stage 9).
This is consistent with what has been observed above
for the outlet fluid temperature. for this layer values of density, thermal conductivity
and specific heat equal to those of air.
A comparison of simulations of Tout obtained with
4.4 Cyclic effects
these settings and measurements for stages 7 and 9 is
A sensitivity study was carried out to preliminarily shown in Figures 10 and 11 respectively. In Table 4
investigate the possible reasons for the particular mis- the model fit in terms of RMSE is reported for stages
match between the outlet temperature simulations and 7 and 9 compared to values obtained with the orig-
data during heat extraction stages 7 and 9. The fact that inal simulation (cf. Section 4.3). It can be seen that
during those stages less heat than predicted is extracted model simulations now can better reproduce field data,
may suggest that, by virtue of differential contraction especially regarding stage 9, representing the second
between concrete and soil within the elastic regime thermal loading-unloading cycle. This result corrobo-
and/or due to possible thermo-plastic effects in the soil rates the conjecture that pile-soil contact might have
upon thermal cycling, contact at the pile-soil interface been reduced during stages 7 and 9, although further
might be reduced. This in turn could affect the lateral numerical analysis accounting for thermo-mechanical
bearing capacity of the pile. couplings would be needed to adequately support this
As a first-attempt analysis, the thermo-mechanical hypothesis.
pile-soil interaction was not modelled numerically (the
FE analysis was kept purely thermal), but its possible
effect was accounted for by changing the thermal prop- 5 DISCUSSION & CONCLUSIONS
erties of a thin layer of solid elements (i.e. a 1 cm thick
ring) in contact with the pile. Possible reduction of In this work, a recently proposed FE numerical model
contact between pile and soil was simulated by setting to interpret the thermal behaviour of energy piles was
598
further developed and validated against field data from Fleur Loveridge is funded by the Royal Academy
a multi-stage thermal response test. Numerical devel- of Engineering under their Research Fellow scheme.
opments, consisting in providing a 3D representation
of pipes (instead of the original 1D schematisation),
led to significant improvements in the model accuracy.
The comparison of simulations and field data REFERENCES
throughout the multi-stage TRT, both in terms of out-
Eskilson, P. 1987. Thermal analysis of heat extraction bore-
let fluid and concrete temperature, highlighted that the
holes. Doctoral Thesis, Department of Mathematical
simulations provide a very good fit in the early TRT Physics, University of Lund, Sweden (http://www.building
stages, and a worse fit at later stages. In particular, the physics.com/Eskilson1987.pdf).
model accuracy appears to worsen in correspondence Bozis, D., Papakostas, K., & Kyriakis, N. 2011. On the eval-
with the heat extraction phases. This suggests the pres- uation of design parameters effects on the heat transfer
ence of cyclic effects in the case study at hand. A efficiency of energy piles, Energy and Buildings, 43 (4),
similar effect was observed by Loveridge et al. (2014b) 1020–1029.
when fitting analytical solutions to the same data, and Cecinato, F. & Loveridge, F. A. 2015. Influences on the
by Ouyang (2014) when assessing the effect of pile- thermal efficiency of energy piles. Energy, 82, 1021–
1033.
soil interaction in the thermal exchange process of the
Cecinato, F., Loveridge, F., Gajo, A. & Powrie, W. 2015.
same TRT. A new modelling approach for piled and other ground
To further investigate cyclic effects, a sensitivity heat exchanger applications. In, XVI European Confer-
study was carried out by tentatively changing the ence for Soil Mechanics and Geotechnical Engineering,
thermo-physical properties at the soil-pile interface Edinburgh, GB, 13–17 Sep 2015. 6 pp.
during the last heat extraction phases (stages 7 and GSHPA 2011. Closed-loop Vertical Borehole Design,
9), to reproduce in a simplified manner the possi- Installation & Materials Standards Issue 1.0, Septem-
ble formation of an air gap, due to thermally induced ber 2011. Ground Source Heat Pump Association, Milton
(differential) contraction of the pile and soil material. Keynes, UK.
IGSHPA 2007. Closed-loop/geothermal heat pump sys-
This resulted in an improved fit of outlet tempera-
tems: Design and installation standards, International
ture data (with special regard to Stage 9, i.e. the last Ground Source Heat Pump Association/Oklahoma State
heat extraction cycle), suggesting that a reduction of University.
pile-soil contact may have been possible in this case. Javed, S. & Claesson, J. 2011. New analytical and numerical
However, this mechanism is not proven and should solutions for the short-term analysis of vertical ground
be corroborated by further investigation accounting heat exchangers. ASHRAE Transactions, 117(1): 3–12.
for thermo-mechanical couplings in a more rigorous Lamarche, L., Kajl, S. & Beauchamp, B. 2010. A review
manner. of methods to evaluate borehole thermal resistance
However, care must also be taken in extrapolating in geothermal heat pump systems. Geothermics, 39,
187–200.
this case to general practice for two reasons. Firstly, the
Loveridge, F. A. & Cecinato, F. 2016. Thermal performance
pile was not subject to mechanical load during the TRT. of thermo-active CFA piles. Environmental Geotechnics,
As highlighted by Ouyang (2014), the mechanical load doi: 10.1680/jenge.15.00023.
would allow the pile to form a better interface with the Loveridge, F.A., Brettmann, T., Olgun, C.G. & Powrie, W.
soil prior to thermal loading, thus preventing the free 2014a. Assessing the applicability of thermal response
contraction of the pile in the cooling cycle. Ouyang testing to energy piles. In, Global Perspectives on the
(2014) also suggest that the interface effects of “weak Sustainable Execution of Foundations Works, Stockholm,
interaction” could result from the unusually large size SE, 21–23 May 2014. 10 pp.
of the plastic spacers or the short time delay between Loveridge, F.A., Powrie, W. & Nicholson, D. 2014b. Com-
parison of two different models for pile thermal response
pile grouting and testing of the piles. The former could
test interpretation. Acta Geotechnica, 9 (3), 367–384.
lead to differential thermal expansion effects within Ouyang, Y. 2014. Geotechnical behaviour of energy piles.
the pile, while the latter could have resulted in the grout PhD thesis, University of Cambridge, UK.
not having cured completely, hence having different Piglialepre, R. 2016. Analisi numerica del funzionamento di
physical characteristics compared with a working pile. un palo geotermico. MSc thesis, University of Trento,
Nonetheless, the possible thermally induced weak- Italy.
ening of pile-soil contact is worth additional analysis Sanner, B., Hellstrom, G., Spitler, J. & Gehlin S. E. A.
considering the thermo-mechanical couplings rele- 2005. Thermal Response Test – Current Status and World-
vant to the suggested mechanisms. This will help Wide Application, In: Proceedings World Geothermal
Congress, 24–29th April 2005 Antalya, Turkey. Interna-
to understand whether or not there is a real risk of
tional Geothermal Association.
changing pile-soil interface conditions in operational Signorelli, S., Bassetti, S., Pahud, D. & Kohl, T. 2007. Numer-
scenarios. ical evaluation of thermal response tests. Geothermics, 36,
141–166.
Spitler, J. D. & Gehlin, S. E. A. 2015. Thermal response test-
ACKNOWLEDGEMENTS ing for ground source heat pump systems – An historical
review. Renewable and sustainable energy reviews, 50,
Francesco Cecinato acknowledges financial support 1125–1137.
from European Union FP7 project under contract num Witte, H. J. L. 2013. Error analysis of thermal response tests.
ber PIAPP-GA-2013-609758-HOTBRICKS. Energy, 109, 302–311.
599
An investigation of the potential thermal energy of geothermal tunnels

with focus on a case study in Stuttgart, Germany
P. Buhmann, B. Westrich & C. Moormann

Institute of Geotechnical Engineering, University of Stuttgart, Germany
A. Bidarmaghz & G. Narsilio

Department of Infrastructure Engineering, The University of Melbourne, Australia
ABSTRACT: Shallow surface geothermal energy is an essential contribution to the base load of heating energy.
The thermal activation of tunnels is an interesting alternative and an economically appropriate supplement of the
present hybrid systems in the field of ground engineering. An essential difference between the tunnel absorber
and structures such as activated piles and diaphragm wall elements is the use of heat fluxes from the earth, as
well as from the inside of the tunnel. Due to this circumstance, the tunnel absorber is assigned to the group
of the duo hybrid systems. The tunnel air temperature is essential for the heat flux inside the tunnel. Stuttgart
Fasanenhof and Jenbach are two geothermal tunnel plants which have been delivering dependable measurement
data of the subsoil temperature, the tunnel lining temperature and the tunnel air temperature since 2011. Based on
the selected operation modes and case studies, the measurement results are analyzed and discussed. In addition,
the possible geothermal potential of a geothermal tunnel plant is introduced.
1 GENERAL earth-facing concrete structures in the field of building

construction and civil engineering.
In Germany, approximately 12% of the primary power For decades, the thermal activation of structures
demand is covered by renewable energy (Nieder et al., of ground engineering such as piles (Knapp, 1992)
2015). In the heating sector, biomass accounts for and diaphragm walls (Brandl, 2013) has reached an
87.7%, solar thermal energy accounts for 5%, while economically effective level.
geothermal energy accounts for only 7.3%. The devel- Heat fluxes are used with systems like borehole heat
opment of solar thermal energy shows a steadily exchangers, between the absorber element (borehole
rising trend. This expansion stagnates the develop- heat exchanger, thermally activated foundation struc-
ment of classical geothermal energy absorber systems. ture) and the surrounding subsoil, for the extraction of
In 2014, the total volume of investment in renewable energy.A logical advancement is the thermal activation
energy in Germany amounted to 18.9 billion euros. of tubular components in the subsoil like infrastructure
Approximately 5% of this amount was used in the tunnels, sewers and other related components. The heat
area of geothermal energy, for both shallow and deep exchanger for these constructions is situated inside
geothermal energy. the load bearing concrete shell. With the activation
Shallow geothermal energy reaches a depth of up of sewers, a combination with absorbers situated on
to 400 m below the surface. With a geothermal gradi- the channel bottom is also possible. Besides, the large
ent of about 3◦ C/100 m, temperatures can reach 25◦ C; earth facing surface of these components are activated
nevertheless, the temperature often lies between 10– in contrast to classical geothermal energy absorbers;
15◦ C in the above locations. For the climate control of not only the heat flux from the absorber to the earth is
buildings, the necessary temperature rise is achieved active, but the heat flux between the absorber and the
by the application of a heat pump. inside of the tunnel of the thermally activated tube is
A classic absorber system in the field of shallow sur- also active.
face geothermal energy is the borehole heat exchanger. The geothermal potential and possible specific
Current studies show (Eicker and Thumm, 2012) extraction capacities of this new technology stand in
that the projected withdrawal lines in this absorber a context of different influences, which are to be con-
technology is often not achieved. In addition, the sidered during the design as well as the operation of
cost of manufacturing the deep-drilling makes heat the plant (Fig. 1).
production unprofitable. At present, 6 tunnels have been equipped with
An alternative to the classic absorber systems of absorber systems world-wide (Frodl et al., 2010; Islam
the shallow geothermal energy is the activation of et al., 2006; Markiewicz, 2004; Zhang et al., 2013),
601
Figure 1. Actions and interactions on tunnel geothermal Figure 3. Absorber system tunnel stuttgart fasanenhof.
plant.
The following explanations consider only the Stuttgart
Fasanenhof tunnel.
At the Stuttgart Fasanenhof tunnel, two separate
tunnel blocks of 10.0m each were equipped with
absorber technology. Therefore, meandering conduits
in which an absorber liquid circulates, beginning from
the tunnel ridge up to the escape in the tunnel walls
(Fig. 3) were fastened to the outside lining of the tunnel
by means of so-called splints.
The whole conduit length per activated block
amounts to 400m. The running after built-in inner
tunnel lining protects the conduit system against dam-
ages. An intensive measuring equipment carries out
the supervision of the tunnel air and tunnel lining
temperature with the aid of a total of 140 measuring
transducers. To capture the development of the subsoil
temperature, vertical and horizontal measuring probes
Figure 2. Subsoil conditions.
were installed in the area of the ridge and the elms
of the tunnel at discreet points; the development of
providing heating and cooling energy for very different the subsoil temperature was measured. Further, the
applications. inlet (ϑin ) as well as the outlet temperature (ϑout ) of
the primary circuit of the absorber system were mea-
sured. The volume flow rate (V̇) was also measured.
2 EXPERIMENTAL PLANTS With these data, it is possible to calculate the over-
all extracted heat flux (Q̇) by using the following
Two basically different experimental plants were avail- equation:
able to the authors for this research activity. They
include the Stuttgart Fasanenhof tunnel (Germany),
an approximately 380m long shallowly lying subur-
ban railway tunnel, as well as the 3470m long railway Also, it is possible to calculate the heat fluxe den-
Jenbach tunnel (Austria) running in the valley Inn. The sities from the absorber to the subsoil as well as from
tunnels differ in their geometrical dimension, their use, the absorber to the interior of the tunnel by using the
their geothermal absorber equipment as well as their following equation:
geological and hydrogeological conditions (Fig. 2).
In the Jenbach tunnel, which is situated in the high
hydraulically permeable “Innschotter” (gravel layers
of the river Inn), the heat transport in the porous media
subsoil is mainly driven by convection, while at the At the Stuttgart Fasanenhof tunnel, from 2011 till
Stuttgart Fasanenhof tunnel, the heat transport occurs now, numerous load scenarios were tested and evalu-
in the solid rock primarily based on conduction. ated. At this point, the interval operation mode from
The Stuttgart Fasanenhof tunnel is operated as a 2011/12 is introduced. Figure 4 shows the interval
pure test plant with given load scenarios, while the operation mode for a period of 7 days.
Jenbach tunnel serves a neighboring contractor’s yard For the simulation of the interval operation mode,
for climate control, covering a real user’s behavior. the volume flow rate (approx. 500 l/h) of an operation
602
Figure 4. Intervall heating mode 2011/2012.
of 2 months duration is held steady and the inlet tem-

perature of the primary circuit is regulated for 8 h each
day to approximately 0.5◦ C. During the remaining 16
hours, the inlet temperature of the primary circuit is Figure 5. Thermal energy demand profiles for heating
influenced by the tunnel climate, the subsoil temper- (continous line) and hot water (dotted line).
ature and the connected operation room. A steadily
rising tunnel air temperature which follows the outside
air temperature of the tunnel is observed. During the
day time, a heat extraction rate of 20 W/m2 is achieved.
This heat extraction rate does not influence the subsoil
temperature development significantly.
In the summer of 2016, the Stuttgart Fasanenhof
tunnel will be driven with a realistic chill demand
profile for the first time. This investigation offers an
essential contribution to the quantification of possible
heat extraction capacities. Figure 6. Numerical simulation model; dimensions: height:
50 m, width: 121 m; 37.000 three node elements.
3 REFERENCE LOAD PROFILES single-family dwelling of the Swabian-Franconian ter-

raced countryside and foothills of the Alps for January
The geothermal potential as well as the attainable heat and May. Seasonally, daily as well as hourly, a very
extraction rate of a geothermal tunnel plant strongly different thermal energy demand can be seen.
depend on the energy demand profile. In the case If the shown demand scenarios can be run by the
of an energy extraction rate taken from the absorber operation of the geothermal-heat pump-absorber sys-
higher than the heat fluxes from the subsoil to the tem, the geothermal tunnel plant gains the possibility
absorber, together with the heat fluxes from the interior of a thermal regeneration. For a geothermal absorber
of the tunnel to the absorber, there will be decreas- system that is particularly influenced by conductive
ing geothermal potential over the time of operation, heat transport, like the Stuttgart Fasanenhof tunnel,
considering a constant absorber temperature. Conse- this has an essential meaning.
quently the assessment of the geothermal potential
depends, beside the geological and hydrogeological
conditions, substantially on the operation modes of 4 CASE STUDY
the geothermal tunnel plant.
If no object-related heating or chilling demand Within the scope of a case study, the Stuttgart Fasa-
profiles are known, e.g. from comparative objects nenhof tunnel was analyzed for an exemplary heat-
or preliminary numerical studies, suitable simplistic ing energy demand of a single-family dwelling. The
demand profiles must be generated for an energetic simulations are based on one (Schneider, 2013) vali-
evaluation of a geothermal tunnel plant. This occurs dated and calibrated numerical 2D-model, which was
on the base (Verein Deutscher Ingenieure, 2008) and calculated using the software FeFlow 6.1.
the test reference years of the German weather ser- Basically, it is possible to consider the energy
vice (Bundesamtes für Bauwesen und Raumordnung, demand as a 2nd kind of boundary condition in the area
2014). Fig. 5 shows the generated thermal energy of the existing heat exchanger between the inner and
demand for heating and hot water of a representative outer tunnel lining. However, this boundary condition
603
Figure 8. Tunnel air temperature and surface temperature
outside tunnel; heat flux density in the area of the absorber.
like infrastructure tunnels, sewers and other related

components.
Two experimental geothermal tunnel plants offer
the possibility to investigate the interaction among the
tunnel air, the absorber system and the surrounding
subsoil. This forms the basis for validating numerical
simulation models.
The heat extraction capacity of a geothermal tunnel
plant is highly influenced by the energy demand pro-
file the plant is operated with as well as the subsoil and
tunnel air conditions. Extractions rates for short shal-
low laying suburban railway tunnels can be assumed
to be 50 W/m2 for long term heating modes.
Figure 7. Temperature field around the tunnel a) after a
cooling period, b) after a regeneration period [◦ C].
ACKNOWLEDGEMENTS
may be taken into account in the subsoil temperatures,
which lie beyond the juridically limiting values. The The authors would like to thank the German Federal
extracted heat flux by time would to be adapted to the Ministry for Economic Affairs and Energy for finan-
development of the subsoil temperatures. This form cial support; they would like to thank the Stuttgarter
of the iteration is currently not available. Hence, for Strassenbahn AG and Ed. Züblin AG for financial
the case study, a minimally allowed absorber tempera- and technical support for carrying out the described
ture was prescibed on the system for the heating mode large-scale pilot projects.
by 0◦ C. Fig. 7 shows the temperature field around
the tunnel after the heating mode, a), as well as the
regeneration time, b). REFERENCES
Consequently, the so ascertained maximum geother- Brandl, H. (2013), “Thermo-active Ground-Source Struc-
mal potential of the geothermal tunnel plant describes tures for Heating and Cooling”, Procedia Engineering,
the maximum heat flux density (q̇) per 1 square meter Vol. 57, pp. 9–18.
of the absorber surface for the analyzed energy demand Bundesamtes für Bauwesen und Raumordnung (BBR)
profile for a single-family dwelling. Considering a (2014), Testreferenzjahre von Deutschland für mittlere,
monovalent operation for the analyzed single-family extreme und zukünftige Witterungsverhältnisse, Offen-
dwelling, a maximum heat flux density of 50 W/m2 bach.
can be the base for dimensioning the heating system Eicker, U. and Thumm, F. (2012), “Energieeffizienz und
(Fig. 8). Wirtschaftlichkeit oberflächennaher Geothermie für das
Heizen und Kühlen von Nichtwohngebäuden”, Bau-
physik, Vol. 34 No. 1, pp. 11–18.
Frodl, S., Franzius, J.N. and Bartl, T. (2010), “Design and
5 CONCLUSIONS construction of the geothermal tunnel system in Jenbach
– Planung und Bau der Tunnelgeothermieanlage in Jen-
The actual developments in the geothermal market bach”, Geomechanics and Tunnelling, Vol. 3 No. 5, pp.
require a lowering of the heat production cost, so that 658–668.
geothermal energy can compete with other renewable Islam, M.S., Fukuhara, T., Watanabe, H. and Nakamura, A.
(2006), “Horizontal U-Tube Road Heating System using
energy sources in future. The thermal activation of Ground Tunnel Heat”, Journal of Snow Engineering of
earth-facing concrete structures in the field of build- Japan, Vol. 22 No. 3, pp. 229–234.
ing construction and civil engineering is a first step Knapp, C. (1992), “Energiepfähle: Ein interessantes System
in this direction. A logical advancement is the ther- zur Energiegewinnung”, Wohnen, 1992, pp. 6–7, available
mal activation of tubular components in the subsoil at: http://dx.doi.org/10.5169/seals-105919.
604
Markiewicz, R. (2004), “Numerische und experimentelle Instituts für Geotechnik der Universität Stuttgart, Institut
Untersuchungen zur Nutzung von geothermischer Energie für Geotechnik, Universität Stuttgart, Stuttgart, 2013.
mittels erdberührter Bauteile und Neuentwicklungen für Verein Deutscher Ingenieure (2008), Referenzlastprofile von
den Tunnelbau”, Institut für Grundbau und Boden- Ein- und Mehrfamilienhäusern für den Einsatz von KWK-
mechanik, Technischen Universität Wien, Wien, 09/2004. Anlagen No. VDI 4655, Beuth Verlag GmbH, Berlin.
Nieder, T., Bickel, P. and Musiol, F. (2015), Entwicklung Zhang, G., Xia, C., Sun, M., Zou, Y. and Xiao, S. (2013), “A
der erneuerbaren Energien in Deutschland im Jahr 2014: new model and analytical solution for the heat conduction
Grafiken und Diagramme unter Verwendung aktueller of tunnel lining ground heat exchangers”, Cold Regions
Datender Arbeitsgruppe Erneuerbare Energien-Statistik Science and Technology, Vol. 88 No. 0, pp. 59–66.
(AGEE-Stat),Stand Dezember 2015, Stuttgart.
Schneider, M. (2013), “Zur energetischen Nutzung von
Tunnelbauwerken – Messungen und numerische Berech-
nungen am Beispiel Fasanenhof ”, Mitteilung 68 des
605
Probabilistic analysis of a rock salt cavern with application to energy

storage systems, using subset simulation methodology
E. Mahmoudi, K. Khaledi, D. König & T. Schanz

Chair of Soil and Rock Mechanics, Ruhr-Universität Bochum, Germany
ABSTRACT: This paper presents an efficient approach to evaluate failure probability in geotechnical structures
dealing with rare failure events. Subset simulation is an appropriate method to substitute Monte Carlo simulation-
based approaches while it decreases computational costs of reliability analyses. This methodology combines the
idea of dividing a failure region into a sequence of nested failure regions and Markov chain Monte Carlo
sampling methodology. To accomplish this, a modified Metropolis-Hastings method is utilized to generate a
chain of random samples in intermediate failure regions from existing samples. Furthermore, subset simulation
is employed to evaluate the failure probability of an underground energy storage system. The surplus baseload of
the produced electrical energy by renewable resources can be converted to compressed air and stored in rock salt
caverns. The validation of the integrity and stability of these caverns is a prerequisite in the geotechnical design
process of them. The present study provides a reliability-based analysis of a typical renewable energy storage
cavern in rock salt. An elasto-viscoplastic creep constitutive model is applied to a finite element numerical model
of rock salt cavern to assess its behavior. The constitutive parameters are represented as random variables with
predefined PDF, mean, and standard deviation. The occurrence of dilation is considered as the mechanical failure
criteria of the system. Subset simulation is further validated by a comparison with a Monte Carlo simulation-based
analysis.
1 INTRODUCTION these uncertainties and evaluate the effect of them

on system predictions seems essential. Reliability
Due to the fact that the uncertainties in the rock and soil analysis assesses the uncertainty effects and their miti-
medium properties are unavoidable, a reliable design gation in the final design which leads to risk-informed
procedure in geotechnical engineering issues can not decisions.
rely merely on deterministic approaches. Besides the Nowadays, numerical simulations are designed
inherent randomness involved in some unpredictable with increasing complexity and higher expectation for
natural process, sources of uncertainties in modeling their reliable predictions. Subsequently, involved fail-
of physical phenomena may be classified into two ure events in such sophisticated designs are intended
major categories: to be rare, which makes its assessment by classical
•
methods, computationally prohibitive.
Aleatoric: this type of uncertainties may arise when
Setting up and solving a reliability-based analysis
the estimation of the parameters and their statistical
problem is usually conducted by Monte Carlo Simula-
measures are inaccurate. Also called as statistical
tion (MCS) methods. However, in case of rare failure
errors may be mostly due to the lack of available
events, the MCS-based method needs a large number
data and assigning inappropriate statistical features
of system evaluations, which drastically increases the
to the input variables.
• Epistemic: the second type of errors in imperfect
computational burden, especially for complex mod-
els. To overcome this drawback, subset simulation
modeling correspond to the inaccuracies of model-
method is employed to decrease the required number
ing and simulation. A mathematical equation which
of model executions. In this approach, a small fail-
idealizes natural phenomena in the framework of
ure probability is broken down into a series of larger
constitutive model, idealized boundary conditions
conditional probabilities of some chosen intermediate
and geometry, or even ignoring some interaction
failure events.
may be origins of these systematic errors.
The efficiency of subset simulation approach is
Thereupon, in addition to an adequately accurate com- shown in this paper by probabilistic analysis of the
putational model, the stochastic analyzes approaches stability of a rock salt storage for renewable energy.
are needed as well. In other words, carrying out prob- Besides all advantages of producing energy from
abilistic analyzes, as substantial tools which consider renewable resources, their production profile may not
609
be compatible with the consumption profile, which Subsequently, the failure probability is computed by
can question their functionality. Storing pressurized
gas, compressed by surplus of the produced electri-
cal renewable-based energy, in rock salt cavities can
overcome this disadvantage.
Rock salt formation has proper features, as negli-
gible permeability, high compressive strength, as well where fX (x)dx is the joint probability density function
as healing properties. Although an underground stor- of X (the vector of variables).
age plant is generally safer and more stable than same In order to solve the integral in Eq. 2, many
facilities on the ground, a poorly designed or operated mathematical techniques have been proposed in the lit-
plant may cause major accidents (Berést and Brouard erature. These methods can be classified into two main
2003). Therefore, the stability of a rock salt cavity in categories on the basis of their underlying approaches,
an underground storing plant are the most important namely analytical approximation (e.g., direct integra-
issues in their geomechanical design process. In order tion method, Gaussian approximation, perturbation,
to investigate how reliable the safety of a rock salt cav- first/second – order reliability or Taylor series meth-
ity is, a no dilation criterion is defined. Exceeding this ods) and stochastic (e.g., importance sampling and
criterion may cause initiation of cracks in the rock salt MCS-based). For more details about these method-
which leads to failure in the cavern’s sealing. ologies, see Ang & Tang (2007) and Au & Wang
In this regard, a finite element numerical simulation (2014). Nevertheless, despite the employed mathe-
of a rock salt cavern, based on an elasto-viscoplastic matical approach i.e., deterministic or stochastic, the
creep constitutive model is conducted to predict the involved computational effort grows drastically when
behavior of surrounding rock salt. The prediction of the failure event corresponds with small probability.
the rock salt cavity behavior in the computational
model relies mainly on the governing constitutive
model. Therefore, the mechanical properties of rock 2.2 Subset simulation approach
salt media play a key role in the numerical design,
Direct Monte Carlo which uses statistical averaging
but determining their values is based on rare test data
over random samples generated from the probability
sets. Moreover, the natural variability of rock salt in
density function of the parameters to evaluate PF , is
the field, accessibility, difficulties to run in-situ exper-
a well-known and robust procedure to address every
iments, and lack of adequate experimental set ups for
complex model. It is employed widely in the geotech-
conducting full-scale tests result in significant lev-
nical field to conduct reliability analyzes, (Tang et al.
els of uncertainties. Therefore, utilizing probabilistic
1976, Phoon and Ching 2014, Miro et al. 2015). How-
approaches developed in Wang (2011), Mollon et al.
ever, since the number of the numerical simulation
(2013) and Phoon & Ching (2014) seems a necessity.
runs required to achieve a given probability of failure
In this context, the contributed parameters of the gov-
(pF ) is proportional to 1/pF , for a small probabil-
erning constitutive model are considered as random
ity failure a huge number of numerical simulations
variables with predefined statistical measures.
is needed. Subset simulation is an advanced MCS
This study describes the methodology of conducting
method that combines conditional probability and
subset simulation in Sec 2. A modified and efficient
Markov chain Monte Carlo (MCMC) method to cal-
version of Metropolis-Hastings method is also pre-
culate small values of probabilities by a few number of
sented in this section. After a brief description of
the deterministic model executions. This methodology
the numerical simulation model of a rock salt cav-
was developed by Au & Beck (2001). Subset simula-
ern in Sec. 3, the employed constitutive model and the
tion method is based on a simple idea that the failure
statistical measures of governing input variables are
probability of a rare event can be represented as the
introduced. Sec. 3 follows with presenting the results
products of a number of more likely conditional fail-
of conducting subset simulation. The article is ended
ure probabilities. Different steps of conducting subset
with conclusions in Sec. 4.
simulation method, presented by Au & Wang (2014),
are illustrated in Fig. 1.
Consider a failure event F defined by the condi-
2 METHODOLOGY tion Gx ≤ 0, where Gx is the performance function,
and let s1 , . . . , sm , . . . , sZ be a sample of Z realizations
2.1 Reliability analysis of a vector s, where s is composed of K random vari-
Reliability analysis evaluates the likelihood of contra- ables. In the subset simulation method, the space of
vening the system stability or functionality criteria. In uncertain parameters is divided into l levels with an
reliability analysis, the space of the system response is equal number of Zs realizations (s1 , . . . , sm , . . . , sZs )
divided into failure and safe regions (Tang et al. 1976). in each level. A sequence of nested failure regions
These regions are separated by the limit state surface, F1 , . . . , Fj , . . . , Fl of decreasing size are defined where
Gx = 0. With a given performance function Gx , the F1 ⊃ . . . ⊃ Fj ⊃ . . . ⊃ Fl = F. An intermediate fail-
failure event is defined as ure region, Fj can be defined by G < yj , where yj
is an intermediate failure threshold whose value is
larger than zero. Thus, there is a decreasing sequence
610
Figure 2. Flowchart of modified metropolis-hastings sam-
pling method.
value of the prescribed value of nth component in the

ascending order,
Figure 1. Flowchart of subset simulation method.
of positive failure thresholds y1 , . . . , yj , . . . , yl corre-

In the next step, the parameter sets corresponding to
sponding to F1 , . . . , Fj , . . . , Fl , respectively. It should
those Zs × P(Fj+1 |Fj ) realizations, which are located
be noted that, y1 > . . . > yj > . . . > yl = 0. An inter-
in the failure zone of Fj , are used as seeds in MCMC
mediate level, j contains a safe region and a failure
analysis. Seeds generate Zs new samples for the next
region defined with respect to a given failure thresh-
level of subset simulation. This procedure to generate
old yj . The failure probability corresponding to the
new levels of intermediate failure regions is repeated
intermediate level of j is calculated by
up to reaching yj ≤ 0.
The failure probability of the failure region F,
denoted as PF , is calculated from the sequence of
conditional failure probabilities
here IFj (sm ) is a non-negative scalar function, which is

equal to one if the system output, related to the vector
of sm , is located in the failure region with respect to
Fj , otherwise IFj (sm ) = 0. The first Zs realizations are where l is the number of levels required to reach the
generated using MCS methodology according to a tar- limit state surface.
get PDF. In order to compute the failure probability in
this study, a same prescribed conditional failure prob- 2.2.1 Modified metropolis-hastings
ability P(Fj+1 |Fj ) is considered for all levels and yj Markov chain Monte Carlo methodology generates
in each level is evaluated, separately. In the following, efficient random samples according to an arbitrar-
the value of performance functions for each sample ily given probability distribution. This method has
are calculated and sorted in an ascending order. Sub- been utilized in statistical estimation problems, like
sequently, the ratio between the number of realizations Bayesian system identification or rare events simu-
for which G < yj and the number of samples Zs is equal lations. One of the well-known algorithms to carry
to the prescribed value of the conditional failure prob- out MCMC methodology is Metropolis-Hastings
ability, yj will be the intermediate threshold (Ahmed (Metropolis et al. 1953, Hastings 1970). This algo-
and Soubra 2012). In other words, yj is equal to the rithm proceeds by generating new sequences of the
611
Markov chain from a proposal distribution conditional 3.2 Material model
(q(.|x(h) )) on the current samples x(h) . Then, it accepts
Since the stress-strain response of rock salt is governed
or rejects these new samples with a certain accep-
by its mechanical properties, the numerical analysis
tance probability which is based on the current and the
should be based on an appropriate constitutive model.
proposed state. Fig. 2 shows the steps of a modified
In this paper, an elasto-viscoplastic creep model is
Metropolis-Hastings algorithm, proposed by Santoso
employed to describe the time-dependent behavior of
et al. (2011). In order to obtain the next sample x(h+1)
rock salt considering dilatation and creep behavior.
from x(h) , a candidate sample x(∗) from a proposal PDF
Under the small strain assumptions, the total strain
is generated, firstly. Then, B factor is calculated,
rate is defined using the equation below,
The elastic strain rate, ε̇el is defined using the

If B ≤ 1, then x(h+1) = x(∗) , otherwise the first step generalized Hooke’s law. The viscoplastic component
is repeated to generate a new candidate. Afterward, of strain rate, ε̇vp is described by utilizing a non-
G(x(h+1) ), i.e., the performance function for x(h+1) is associated flow rule, which has been developed by
evaluated. In case G(x(h+1) ) < yj , which means x(h+1) Desai & Varadarajan (1987) based on the viscoplas-
is located in Fj domain, x(∗) is accepted as the new tic model of Perzyna (1966). The creep strain rate, ε̇cr
ij
sample. Otherwise, x(∗) is rejected and x(h+1) = x(h) . which represents the steady state creep behaviour of
According to the suggestion made by Santoso et al. the rock salt is defined on the basis of the viscosity
(2011), in this study, the target probability distribution of Maxwell model (stress-dependent) which controls
is assumed to be Gaussian and the proposal PDF was the steady state strain rate of the model (Heusermann
chosen to be uniform. et al. 2003).This constitutive model is developed by the
authors and explained in detail in a companion paper.
3 EXAMPLE ANALYSIS 3.3 Random variables

3.1 Deterministic numerical model In order to conduct reliability analysis, the statisti-
cal measures of the relevant uncertainties should be
Figure 3 shows a typical geological profile of a rock identified for each input parameter. The uncertain
salt strata where a storage cavern may be excavated. input variables are characterized by their statistical
The rock mass is idealized by dividing it into two moments, e.g., the mean value and standard deviation,
homogeneous layers, i.e. rock salt and cap rock. In besides their relevant probability density functions.
this work, the depth of cavern’s roof is assumed to be Based on the results of a conducted and documented
equal to 800 m. The non-saline material above the rock sensitivity analysis by the authors in a companion
salt formation, i.e. cap rock with a specific weight of paper, six mentioned parameters in the Table 1 have
γcr = 21 KN/m3 , simulated as a uniform load at the top the most influence on the outputs variability. All the
of the rock salt column. The shape of the cavity after regarded uncertain parameters are assumed to vary
excavation is considered as a cylinder with a height between upper and lower bounds according to their
of 233 m and a diameter of 75 m. Fig. 3 illustrates the statistical measures. The minimum and maximum val-
geometry, boundary conditions and discretized finite ues of the uncertain parameters are chosen based
element mesh of this cavern. More details about this on former experiences (Hansen et al. 1984, Desai
simulation are presented by the authors in (Khaledi and Zhang 1987, Sane et al. 2008, Guo et al. 2012,
et al. 2016). Heusermann et al. 2003) and engineering judgment. It
In order to model the solution mining procedure, should be noted that the probability distribution of the
the whole excavation and debrining process are sim- input variables is assumed to be in accordance with a
plified by reducing the internal pressure of cavern Gaussian distribution, the most common distribution
in a time duration about three and a half years. To encountered in engineering applications.
accomplish this, the geostatic stress, which is assumed
to be initial stress condition, is reduced gradually to
3.4 Failure criteria
the maximum internal pressure of the cavern. The
stored gas is also simulated by applying the equivalent The deviatoric stress, which can be induced by differ-
internal pressure to the cavern’s boundary. The oper- ent internal pressures and in-situ stresses of the rock
ation protocol is assumed to be cyclic (one cycle per around the cavities, leads to creep deformation and
day). Hence, the internal pressure of the cavity after dilatant behavior, as well (Cristescu & Hunsche 1998).
excavation is set to 10 MPa, afterward it decreased to Because of dilatation, the mechanical properties of the
7 MPa during 12 hours, and then it is kept constant for host rock can change drastically, which may initiate
12 hours. The cyclic loading simulation is followed micro cracks. In some cases, a micro crack network
by raising the internal pressure gradually to 10 MPa can provide possible path ways for the stored product in
within 12 hours time interval. In this study, 100 cycles the cavity to mitigate out. Therefore, the dilatant zone
of charging/discharging the cavern are simulated. in the rock surrounding the cavern should be avoided.
612
Figure 3. Representative stratigraphy of salt deposit (a), geometry and boundary conditions of the salt cavern model (b).
Table 1. Statistical characteristics of material parameters.
Parameter Description Dimension Boundaries COV (%)
E Elastic modulus MPa [19000 , 25000] 10

N Flow exponent [-] [2.4 , 3.5] 5
a1 Hardening parameter [-] [0.3e−4 , 0.35e−3] 5
γ Ultimate parameter [-] [0.085 , 0.11] 5
η Hardening parameter [-] [0.7 , 0.9] 5
η∗M Maxwell coefficient MPa s [1e12 , 3e12] 10
√
The mathematical formulation of dilatancy boundary J2 is the value of the second invariant of devia-
which determines the dilatancy domain in the stress toric stress. If DoU , transcended one it corresponds
space entitled as compression/dilatancy (C/D) bound- to accordance of dilatancy around the cavern.
ary. Those loading conditions leading the stress state As previously mentioned, the model response space
to locate above the C/D line (namely dilatant zone) is is separated through the limit state surface, into safe
regarded as unsafe states, thereupon the C/D line is and failure regions. In this study Gx , the performance
considered as the failure criterion. function, is related to no dilation failure criterion.
Over the years, various constitutive models based Based on that, the following function is defined,
on various empirical investigations or different rhe-
ological models, define slightly different C/D bound-
aries (Mahmoudi et al. 2015). In this study, Desai dila-
tancy boundary (Desai and Zhang 1987) is employed. In Eq. 9, Gs is the performance function related to the
To investigate the occurrence of dilatancy around degree of utilization against dilatation, the subscripts u
the cavern, the second variant of the stress state of the and s indicate ultimate resistance and calculated values
surrounding rock salt is evaluated to define a quan- of the system, respectively.
tity as degree of utilization (DoU ), based on Bond &
Harris (2008) and DIN 1054:2003-01,
3.5 Results
In the following, results of conducting subset simu-
lation method to evaluate the failure probability of
the above mentioned rock salt cavern are presented.
here, J2dil indicates the distance of the CC/D Firstly, the required number of model evaluations in
boundary from the isotropic condition in π-plane and each level is determined. Afterward, a comparative
613
Figure 4. Coefficient of variation of failure probability
versus the number of samples generated per level. Figure 5. Comparison between the obtained PF computed
by applying MCS and subset simulation on the bottom of
cavern.
study is conducted in order to verify the obtained
outcomes by subset simulation.
intermediate failure probabilities. Thereupon, the final
3.5.1 Sufficient number of model evaluation PF s are calculated by subset simulation using 3,000
As we consider the intermediate failure P(Fj+1 |Fj ) realizations. These results show that, although the
amount in all intermediate levels to be the same, the computational burden required in subset simulation
corresponding intermediate thresholds depend on the approach is significantly less compared to MCS, the
generated conditional samples and vary in different differences between the obtained PF s are negligible.
runs. To ensure that the variations in the sample sets
make no significant differences in the obtained value
of PF , the coefficient of variation of the probability
failure value (denoted as COVPF ) can be a proper mea- 4 CONCLUSIONS
sure of how accurate is the conducted analysis. For
more details about the procedure of evaluating COVPF , This paper presents a computationally affordable reli-
see Au and Beck (2001). ability analysis method, called as subset simulation.
In order to obtain the optimal number of required Subset simulation can address small probabilities
sample sizes, different Zs are applied to calculate the encountered in the practical reliability assessment of
probability of failure, and for each case, the COVPF complex systems which calculating them with classi-
is computed. Fig. 4 represents the considered num- cal MCS-based methods is computationally expensive.
ber of realizations in each level of subset simulation A modified Metropolis-Hasting approach is employed
versus its corresponding COVPF . It demonstrates the to generate adaptive samples in a sequences of failure
dependency of the coefficient of variation of PF on regions. Moreover, reliability-based analysis is con-
the number of samples. Based on this figure, raising ducted on a typical renewable energy storage cavern
Zs from 1000 to 4000 makes no important change in in rock salt. In this regards, constitutive parameters are
the COVPF , thereupon, the number of samples per level represented as random variables and the probability of
in the present study is chosen to be equal to 1000. failure in the stability of system is evaluated consider-
ing no dilation criterion. Afterward, the efficiency and
3.5.2 Validation by comparing with MCS method accuracy of subset simulation was justified by con-
The efficiency of subset simulation methodology is ducting a comparative calculation. It should be noted
examined by comparing its results with the estimated that this case study is a synthetic one, nevertheless,
PF s using an MCS-based analysis. The failure prob- the proposed method can be applied analogously to
ability of the rock salt cavern’s vicinity, considering realistic problems.
no-dilation criteria, is calculated by conducting both
methodologies. The comparison here is conducted for
a point located on the bottom of the cavern, and the fail- REFERENCES
ure probability is computed against several prescribed Ahmed, A. & A.-H. Soubra (2012). Extension of subset sim-
DoUu . Fig. 5 illustrates the obtained results, where in ulation approach for uncertainty propagation and global
the MCS-based approach 100,000 samples are used. In sensitivity analysis. Georisk: Assessment and Manage-
subset simulation method the corresponding PF is esti- ment of Risk for Engineered Systems and Geohazards 6(3),
mated by generating samples through three levels of 162 – 176.
614
Ang, A. H.-S. & W. H. Tang (2007). Probability Concepts salt. Journal of Computational and Applied Mathemat-
in Engineering: Emphasis on Applications to Civil and ics 293, 128–138.
Environmental Engineering, 2nd Edition. John Wiley & Mahmoudi, E., K. Khaledi, D. König, & T. Schanz (2015).
Sons Ltd. Numerical simulation of deep and shallow energy stor-
Au, S.-K. & J. L. Beck (2001). Estimation of small fail- age systems in rock salt. In T. Schanz and A. Hettler
ure probabilities in high dimensions by subset simulation. (Eds.), Aktuelle Forschung in der Bodenmechanik 2015,
Probabilistic Engineering Mechanics 16, 263–277. pp. 69–83.
Au, S.-K. & Y. Wang (2014). Engineering risk assessment Metropolis, N., A. W. Rosenbluth, M. N. Rosenbluth, A. H.
with subset simulation. John Wiley & Sons Ltd. Teller, & E. Teller (1953). Equation of state calculations
Berést, P. & B. Brouard (2003). Safety of salt caverns by fast computing machines. The Journal of Chemical
used for underground storage. Oil & Gas Science and Physics 21, 1087–1092.
Technology 58, 361–384. Miro, S., M. König, D. Hartmann, & T. Schanz (2015).
Bond, A. & A. Harris (2008). Decoding Eurocode 7. Taylor A probabilistic analysis of subsoil parameters uncer-
& Francis. tainty impacts on tunnel-induced ground movements with
Cristescu, N. & U. Hunsche (1998). Time Effects in Rock a back-analysis study. Computers and Geotechnics 68,
Mechanics. John Wiley & Sons Ltd. 38–53.
Desai, C. & A. Varadarajan (1987). A constitutive model for Mollon, G., D. Dias, & A.-H. Soubra (2013). Range of the
quasi-static behavior of rock salt. J. Geophys. Res 92, 445– safe retaining pressures of a pressurized tunnel face by
456. a probabilistic approach. Journal of Geotechnical and
Desai, C. & D. Zhang (1987). Viscoplastic model for geologic Geoenvironmental Engineering 139, 1954–1967.
material with generalized flow rule. Int. J. numer. Anal. Perzyna, P. (1966). Fundamental problems in viscoplasticity.
Met 11, 603–627. Rec. Adv. Appl. Mech 9, 243–377.
Guo, Y., C. Yang, & H. Mao (2012). Mechanical properties Phoon, K.-K. & J. Ching (2014). Risk and Reliability in
of jintan rock salt undercomplex stress paths. Int. J. Rock. Geotechnical Engineering. CRC Press.
Mech. Min 56, 54–61. Sane, S., C. Desai, J. Jenson, D. Contractor, A. Carlson,
Hansen, F. D., K. D. Mellegard, & P. E. Senseny (1984). Elas- & P. Clark (2008). Disturbed state constitutive modeling
ticity and strength of ten natural rock salts. In H. R. J. of two pleistocene tills. Quaternary Science Reviews 27,
Hardy and M. Langer (Eds.), The Mechanical Behavior of 267–283.
Salt,Proc. of the First Conferenc, pp. 53–70. Santoso, A., K. Phoon, & S. Quek (2011). Modified
Hastings, W. K. (1970). Monte carlo sampling methods using metropolishastings algorithm with reduced chain cor-
markov chains and their applications. Biometrika 57, relation for efficient subset simulation. Probabilistic
97–109. Engineering Mechanics 26, 331–341.
Heusermann, S., O. Rolfs, & U. Schmidt (2003). Nonlinear Tang, W. H., M. S. Yucemen, & A. H. S. Ang (1976).
finite element analysis of solution mined storage cav- Probability based short-term design of slope. Canadian
erns in rock salt using the lubby2 constitutive model. Geotechnical Journal 13, 201–215.
Computers and structures journal 81, 629–638. Wang, Y. (2011). Reliability based design of spread soun-
Khaledi, K., E. Mahmoudi, M. Datcheva, D. König, & dations by monte carlo simulations. Geotechnics 61,
T. Schanz (2016). Sensitivity analysis and parameter iden- 677–685.
tification of a time dependent constitutive model for rock
615
Effect of degree of saturation on thermal conductivity of CLSM used

for a horizontal ground coupled heat pump system
T.M. Do, Y.S. Kim, C.H. Lee & M.Q. Dang

Department of Civil and Environmental Engineering, Chonnam National University, South Korea
ABSTRACT: The aim of this paper is to investigate the effect of degree of saturation on thermal conductivity
of coal ash based CLSM for an use as backfill material of a Horizontal Ground Coupled Heat Pump system
(HGCHP). Initially, a family of coal ash based CLSM mixtures with a variable range of water to binder (W/B)
and water to solid (W/S) ratios, different kinds of binders (e.g., cement and Cementless Binder (CB)) were
systematically tested in accordance with applicable ASTM Standards to determine general properties of CLSM
including bleeding, flowability, fresh unit weight, initial setting time, and unconfined compressive strength.
Subsequently, thermal conductivity was measured by using thermal needle probe conforming ASTM D 5334 at
various degrees of saturation of CLSM specimens.As a result, the general engineering properties of coal ash based
CLSM satisfied the specification of ACI 229R. In addition, a sensitive relationship between degree of saturation
and thermal conductivity of prepared CLSM mixtures was achieved, the higher the degree of saturation, the
higher the thermal conductivity. This finding is very important when determining a proper thermal conductivity
with respect to various degrees of saturation to play a significant role in achieving optimum performance and
full potential toward the use of the given CLSM mixtures as a backfill for trenches of HGCHP. Finally, a thermal
conductivity prediction equation was proposed as a function of degree of saturation, thermal conductivity under
fully saturated and dry states.
1 INTRODUCTION these coal ashes as their mixture components. The

American Concrete Institute (ACI) defines a CLSM
Based on the standpoint that the warm earth and as a self-leveling, self-compacting, and cementitious
groundwater below the surface provides a free renew- material primarily used to replace conventional back-
able source of energy year-round to heat and cool fill soil and structural fillings that result in unconfined
an average suburban residential home, Ground-Source compressive strengths of 1,200 psi (8.3 MPa) or less
Heat Pump (GSHP) systems, transforming this Earth (ACI 229R, 1999). CLSM should not be considered
Energy into useful energy to heat and cool buildings, as a type of low-strength concrete, but rather as a
have progressively been evolved in Korea. Significant structural backfill. CLSM is known by many differ-
energy savings can be achieved through the use of ent names such as flowable fill, controlled density fill,
GSHPs in place of conventional air-conditioning sys- unshrinkable fill, flowable mortar, soil-cement slurry,
tems and air-source heat pumps. In general, GSHP and plastic soil-cement. There are various inherent
systems exchange heat with the ground, often by advantages of using CLSM instead of compacted fill
means of GCHP that use the ground as a heat source in these applications. These benefits include reduced
and sink, either with vertical or horizontal ground heat labor and equipment costs (due to self-leveling proper-
exchangers. In these systems, a backfill material is ties and no need for compaction), faster construction,
used to fill up as well as provide heat transfer medium and the ability to place material in confined spaces
between heat exchanger and surrounding ground (e.g., (ACI 229R, 1999). If future excavation is anticipated,
soils or rocks) and also control groundwater move- the maximum long-term compressive strength should
ment to prevent contamination of water supply. For that generally be less than 2.1 MPa (i.e., general fill) (ACI
purpose, it is ideal to apply Controlled Low Strength 229R, 1999). It is advantageous due to the relatively
Material (CLSM) as a backfill material to oppose low strength itself of CLSM. Another advantage of
the use of conventional grouts. Meanwhile, with a CLSM is that it possibly contains by-product materi-
generation of 3,245 MW thermal power plants every als, thereby reducing the demand on landfills, where
year, large quantities of coal ashes are being pro- these materials may otherwise be deposited and con-
duced during incineration and stored at Honam area in tributing towards the sustainable development (Razak
Korea. The disposal of these wastes will be a big chal- et al., 2009). During the past decade, several successful
lenge in the near future for Korea to decrease harmful applications of coal ashes in CLSM have been reported
environmental effects. Interestingly, CLSM can use (Gabr and Bowders, 2002; Lee et al., 2013; Katz
617
and Kovler, 2004). Other industrial by-products have
been used in CLSM production such as foundry sand
(Siddique and Noumowe, 2008) blast furnace slag
(Muhmood et al., 2009) and others. The most com-
mon ones of these studies are the ones that consider
the recycled and by-product materials as ingredients
for a production of CLSM. However, thermal conduc-
tivity of CLSM under various degrees of saturation has
never been closely evaluated in the literature.
In addition, from a standpoint of economics and
sustainable development, it has been a trend of CLSM
production without using cement since the produc-
tion of Portland cement not only consumes limestone, Figure 1. Particle size distribution of ponded ash.
clay, coal, and electricity, but also releases waste gases
such as CO2 , SO3 , and NOx . In Korea, cementless Table 1. Physical properties of ponded ash.
binder (CB) which is manufactured with granulated
blast-furnace slag and other alkali activators is a new Properties Ponded ash
alternative material to cement for the production of
CLSM (Mun et al., 2007). The reactions of pozzolanic Maximum dry density (kg/m3 ) 1279
materials (e.g., fly ash and ponded ash in CLSM Optimum Moisture Content (OMC) (%) 17
mixture) and slag coupled with some activators in Natural water content (%) 15.57
CB are believed to help in hardening and strength Specific gravity 2.15
development of CLSM. Water absorption (%) 4.65
Fineness modulus 3.37
Since the HGCHP is often less expensive to install
Particles <75 µm (%) 4.91
than the vertical one, but requires a larger land area, Color Black
it can be preferably used in Honam area for residen-
tial and small commercial buildings. In this study,
CLSM made with coal ashes and cementless binder also important in CLSM mixtures with the hydration
was considered to be used as a backfill material for process that enables CLSM to be cohesive and hence
the HGCHP. Moreover, due to the fact that direct mea- harden to develop strength. In this study, Ordinary
surement thermal conductivity of this material under Type I Portland cement conforming to ASTM C150
various degrees of saturation is practically impossi- was employed. In addition, Cementless binder was
ble, the evaluation of effect of degree of saturation on also employed in this investigation. It is non-sintering
thermal conductivity of CLSM plays a key role to the cement and manufactured by adding phosphogypsum
efficiency maintenance of the HGCHP systems. and waste lime to granulated blast-furnace slag as
sulfate and alkali activators (Mun et al. 2007). The
reactions of pozzolanic materials (e.g., fly ash and
2 EXPERIMENTAL STUDY ponded ash in CLSM mixture) and slag coupled with
some activators in CB are believed to help in hardening
2.1 Materials and strength development of CLSM. Fly ash used in
Conventional CLSM mixtures usually consist of water, CLSM can improve the fluidity or flowability of mix-
Portland cement, fly ash and aggregates (ACI 229R tures with its fineness and spherical shaped particles.
1999). Fine aggregate is considered as a material with In this study, Class F Fly ash produced from Cogener-
particles in a size range from 4.75 mm (No. 4 sieve) ation plant in Honam area conforming to ASTM C618
to 0.075 mm (No. 200 sieve), commonly up to 80– was used with its specific gravity of 2.3.
85% (ACI 229R, 1999). In the present study, ponded
ash originated from cogeneration plants in Honam area
2.2 Mixture proportions and test procedures
was used as a fine aggregate in a production of CLSM.
Ponded ash was first dried in an oven at 105◦ C until Table 2 shows the mix proportions of various coal
constant mass, and then sieved through a 9.52 mm ash based CLSM mixtures. CLSM mixtures were pro-
size sieve to eliminate unnecessary large particles and duced with binders (Portland cement or cementless
approach the particle size of natural fine aggregates. binder), Class F fly ash, ponded ash, and water. At first,
The physical properties of ponded ash are detailed in all mixtures were evaluated for bleeding, flowability,
Table 1. The particle size distribution curve of ponded fresh density, setting time and unconfined compressive
ash is shown in Fig. 1. Ponded ash was classified as a strength in accordance with applicable ASTM Stan-
poorly graded sand (SP) based on Unified Soil Classi- dards (2004): C 232, D 6103, C 138, C 403, D 4832,
fication System (USCS) (ATSM D2487 2004) with the respectively. Flowability is the property that governs
classification parameters of Cu (20.25), Cc (0.94). The the self-leveling ability of CLSM. In this study, flowa-
fineness modulus of ponded ash was 3.37. In addition bility of CLSM was measured according to ASTM D
to fine aggregate, other ingredients such as cement, 6103 with an open-end flow cylinder with 150 mm
supplementary cementitious materials and water are of length and 75 mm in diameter. The cylinder was
618
Table 2. Mix proportions of various CLSM mixtures.
Unit weight (kg/m3 )
Ratios Binders
Ponded
Mix code W/B W/S ash Cement CB Fly ash Fresh density (kg/m3 ) Cement content (%)
WB1C 1.2 – 1067.03 79.52 0.00 280.74 1857.97 4.28

WB2C 1.15 – 1070.63 93.91 0.00 281.74 1878.29 5.00
WB3C 1.1 – 1065.02 107.48 0.00 280.28 1882.33 5.71
WB1CB 1.2 – 1066.56 0.00 79.49 280.62 1857.15 0.00
WB2CB 1.15 – 1069.85 0.00 93.85 281.54 1876.93 0.00
WB3CB 1.1 – 1063.00 0.00 107.28 279.75 1878.76 0.00
WS1C – 0.27 1097.24 96.29 0.00 288.70 1880.76 5.12
WS2C – 0.3 1070.63 93.91 0.00 281.74 1878.29 5.00
WS3C – 0.33 1044.02 91.62 0.00 274.68 1873.70 4.89
WS1CB – 0.27 1096.98 0.00 96.27 288.63 1880.32 0.00
WS2CB – 0.3 1069.85 0.00 93.85 281.54 1876.93 0.00
WS3CB – 0.33 1043.99 0.00 91.62 274.67 1873.63 0.00
WB-C: Water to binder (CLSM made with cement); WB-CB: Water to binder (CLSM made with cementless binder);
WS-C: Water to solid (CLSM made with cement); WS-CB: Water to solid (CLSM made with cementless binder).
initially placed on a levelled nonporous plate and then

filled with fresh CLSM. The cylinder was lifted slowly
within 5s allowing the fresh CLSM mixture to spread
freely on the plate. The diameters of spread in two
perpendicular directions were measured. The average
diameter of spread represents the flowability of the
mixture. Bleeding is commonly used to quantify the
degree of subsidence for CLSM and was measured in
this study according to ASTM C 232. A 150 × 150 mm Figure 2. Thermal conductivity devices (ASTM D5334).
cylinder mould was filled with fresh CLSM mixture
and then covered with a plastic para-film to prevent of 25◦ C and 100 % of relative humidity immediately
evaporation. The volume of bleed water which accu- adjacent to the cylinders until desirable testing times
mulated at the surface of the CLSM was recorded until of 3 days, 7 days and 28 days as presented in ASTM
two successive readings showed no further bleeding. D 4832.
The bleeding was expressed as a percentage with the Finally, thermal conductivity was measured by
volume of bleed water over the initial volume of CLSM using thermal needle probe, conforming ASTM D
mixture. The fresh density of CLSM was obtained by 5334 as shown in Fig. 2.
measuring the weight of the fresh CLSM mixture filled The thermal needle probe used consists of a heating
with a 100 × 200 mm cylinder container. Initial set- element and a temperature measuring element and is
ting time was performed by penetration resistance test inserted into the specimen. A known current and volt-
according to ASTM C 403. age is applied to the probe and the temperature rise
A container of 150 mm in diameter and 150 mm in with time is recorded over a period of time. The tem-
depth was first filled with fresh CLSM mixture. Before perature decay with time after the cessation of heating
starting the penetration test, bleed water was removed can also be included in the analysis to minimize effects
from the surface of the CLSM mixture by means of a of temperature drift during measurement. The thermal
pipet or suitable instrument. In this study, the needle conductivity is obtained from an analysis of the time
with 16 mm2 in area was selected and then inserted into series temperature data during the heating cycle and
contact with the CLSM sample until the penetration cooling cycle, if applicable. CLSM specimens for ther-
reached a depth of 25 mm. The penetration resistance mal conductivity tests were prepared at 0%, 20%, 40%,
was calculated by dividing the recorded force by the 60%, 80%, and 100% degree of saturation. Fully sat-
bearing area of the needle. The period of time from urated specimens were tested first at 28 days since the
the initial contact of binder and water to the time of complete saturation was achieved throughout the cur-
reaching a penetration resistance of 2.74 MPa was cho- ing condition (i.e., CLSM specimens were immersed
sen as the initial setting time. Unconfined compressive in water immediately after de-moulding). Specimens
strength test was further performed with 50 × 100 mm tested at 0% saturation were dried in a drying oven at
cylindrical specimens moulded in plastic moulds and 50 ± 2◦ C and weighed at 24-h intervals until the loss
then stored under the curing storage condition with in weight did not exceed 1% in a 24-h and then cooled
a curing chamber that can maintain the temperature to room temperature (Benazzouk et al., 2008). The
619
Table 3. All general properties of CLSM mixtures mix code.
Ratios Compressive strength (kPa)

Bleeding Flow Initial set
Mix Code W/B W/S 3 days 7 days 28 days (%) (cm) (hours)
WB1C 1.2 – 154.13 278.76 523.16 2.630 27 9.09

WB2C 1.15 – 250.79 469.16 827.05 2.062 25 8.57
WB3C 1.1 – 471.18 614.47 1095.62 2.080 24 7.82
WB1CB 1.2 – 415.83 480.16 656.11 3.684 28 18.07
WB2CB 1.15 – 556.46 805.01 1021.14 3.271 26 17.03
WB3CB 1.1 – 633.88 1117.2 1582.2 3.160 24.5 16.83
WS1C – 0.27 487.15 827.45 1339.04 1.724 23 7.65
WS2C – 0.3 250.79 469.16 827.05 2.062 25 8.57
WS3C – 0.33 224.42 331.19 631.11 3.257 29 10.32
WS1CB – 0.27 1088.12 1310.58 1935.82 2.914 24 16.79
WS2CB – 0.3 556.46 805.01 1021.14 3.271 26 17.03
WS3CB – 0.33 366.68 513.87 826.75 4.678 30 22.47
WB-C: Water to binder (CLSM made with cement); WB-CB: Water to binder (CLSM made with cementless binder);
WS-C: Water to solid (CLSM made with cement); WS-CB: Water to solid (CLSM made with cementless binder).
partially saturated specimens were obtained by plac- made with cement. Table 3 also shows the unconfined
ing them on paper towels and air-dried at either room compressive strengths in average for various CLSM
temperature or with the aid of warm air blowers with mixtures in experimental program. There were incre-
occasional turning to prevent concentration of water ments in compressive strength of all CLSM mixtures
in any particular section of the specimens. During the due to the hydration process of cementitious materi-
drying process, CLSM specimens were periodically als which gradually completed with curing ages. The
weighed until they had reached a previously calculated 3-day strength and 7-day strength varied in ranges of
weight corresponding to the prescribed degree of sat- 0.15–1.08 MPa and 0.28–1.31 MPa, respectively. At
uration at which time they were immediately tested 28 days age, the strength was 0.52–1.94 MPa, con-
thermal conductivity. forming to strength requirement of CLSM in ACI 229.
Moreover, the compressive strength of mixtures made
with CB showed a higher strength than that of ones
3 RESULTS AND DISCUSSION made with cement in the CLSM mixtures. The reac-
tion of pozzolanic materials (e.g., fly ash and ponded
3.1 General properties of CLSM ash) activated by lime in CB are believed to form the
C-S-H gel (Lee et al., 2013). Additionally, in CB, the
Test results of flowability, bleeding, initial setting time presence of phosphogypsum generated the shortage of
and compressive strength of coal ash based CLSM (SO3 )−2 ion that completely converted Ca2+ and Al3+
made with cement and CB are summarized in Table 3. ions contained in the granulated blast-furnace slag into
As expected for all CLSM mixtures, flowability, rang- ettringite. Residual Ca2+ and Al3+ ions also could be
ing from 230 mm to 300 mm, satisfied the standard reacted with water to create C4AH13 (Mun et al., 2007).
requirement of flowable fill reported in ACI 229. In These gels and ettrigite were then filled up the pores
addition, it was found that flowability of CLSM mix- in the specimens and eventually help in hardening and
tures made with CB was slightly higher than that of strength development of CLSM.
ones made with cement, regardless of W/B or W/S
ratios in mixtures, probably because of the presence
of slag, known as the workability improvable material
3.2 Effect of degree of saturation on thermal
(Sheen et al., (2013), in CB. The excess water added to
conductivity of CLSM
maintain the required flowability of CLSM comes out
as bleed. The bleeding values of all prepared mixtures The thermal conductivity of a material is the quan-
varied in the range of 1.72–4.68%, all of which con- tity of heat transmitted through a unit thickness in a
formed to the bleeding requirement of CLSM of less direction perpendicular to a surface of unit area, due
than 5% at 2 hours. As similar as flowability, due to the to a unit temperature gradient under given conditions
high amount of slag in CB, bleeding of CLSM made (Sengul et al., 2011). The experimental thermal con-
with CB was higher than that of one made with cement ductivity values of coal ash based CLSM measured
for all mixtures. Therefore, CLSM is considered sta- under various degrees of saturation are shown in Figs.
ble with bleeding. Initial setting time was found in the 3–6. Generally, high thermal conductivity is needed
range of 7.65–22.47 h, below the maximum allowable to provide the heat transfer capacity of CLSM. It was
limit of 36 h as required for general CLSM. Further- observed that thermal conductivity depends strongly
more, it was revealed that initial setting time of CLSM on the degree of saturation, regardless of W/B and
mixtures made with CB was relatively longer than that W/S ratios or binders used, the higher the degree
620
Figure 5. Thermal conductivity of CLSM made with CB
Figure 3. Thermal conductivity of CLSM made with cement with respect to degrees of saturation and W/B ratios.
with respect to degrees of saturation and W/B ratios.
Figure 6. Thermal conductivity of CLSM made with CB

Figure 4. Thermal conductivity of CLSM made with cement with respect to degrees of saturation and W/S ratios.
with respect to degrees of saturation and W/S ratios.
was achieved with a lower degree of saturation. Based
on this understanding, a multi-regression statistic was
of saturation, the higher the thermal conductivity. In employed from data collected from thermal conductiv-
addition, thermal conductivity of all prepared CLSM ity tests to develop the prediction equation for CLSM.
specimens made with cement and CB showed large It was then proposed as the equation 1.
reductions, ranging from 67.15–69.75% and 55.75–
71.88%, respectively, between fully saturated and dry
states. Thermal conductivity showed highest values at
the fully saturated states since the pores would be fully where K = thermal conductivity of CLSM under cer-
filled with water. Under the dry states, water might be tain degree of saturation; S = Degree of saturation; and
fully replaced by air. Since the thermal conductivity of Ks, Kd = thermal conductivity of CLSM under fully
air (0.024 W/mK) is much smaller than that of water saturated and dry state.
(0.6 W/mK). This prediction equation was further verified by
Therefore, the large reductions in total thermal comparing with the measured results in experimental
conductivity of CLSM mixtures in these cases are program as shown in Fig. 7. As a result, the pre-
expected. In addition, thermal conductivity of CLSM dicted thermal conductivity values were in relatively
made with CB was acceptable when comparing to that good agreement with the measured ones. This means
of conventional grout made with bentonite. This find- it is possible to predict thermal conductivity of CLSM
ing is important when the proposed CLSMs are con- mixtures at any partially saturated state by using Ks
sidered to be used for an economic backfill material and Kd.
of HGCHP. Finally, a thermal conductivity prediction
equation was proposed as a function of degree of satu-
ration, thermal conductivity under fully saturated (Ks) 4 CONCLUSIONS
and dry states (Kd). Making this equation targeted to
determine whether Ks and Kd can be used to pre- The effect of degree of saturation on thermal conduc-
dict the thermal conductivity of CLSM specimens at tivity of coal ash based CLSM was evaluated in this
any partially saturated states that reflect field condi- study. At first, all general properties of CLSM includ-
tions. As a general trend, a lower thermal conductivity ing bleeding, flowability, fresh density, initial setting
621
ASTM C138. 2004. Standard Test Method for Density (Unit
Weight), Yield, and Air Content of Concrete. Annual book
of ASTM Standards, 4(2), PA 19428-2959, US.
ASTM C403. 2004. Standard Test Method for Time of Setting
of Concrete Mixtures by Penetration Resistance Annual
book of ASTM Standards, 4(2), PA 19428-2959, US.
ASTM C618. 2004. Standard Specification for Fly Ash And
Raw Or Calcined Natural Pozzolan For Use As A Mineral
Admixture In Portland Cement Concrete Annual book of
ASTM Standards, 4(2), PA 19428-2959, US.
ASTM C232. 2004. Standard Test Method for Bleeding
of Concrete Annual book of ASTM Standards, 4(2), PA
19428-2959, US.
ASTM D2487. 2004. Standard Practice for Classification of
Soils for Engineering Purposes (Unified Soil Classifica-
tion System) Annual book of ASTM Standards, 4(2), PA
19428-2959, US.
ASTM D6103. 2004. Standard test method for flow consis-
tency of controlled strength material (CLSM)Annual book
of ASTM Standards, 4(2), PA 19428-2959, US.
ASTM D4832. 2004. Standard test method of preparation
Figure 7. Comparison of the predicted thermal conductivity and testing of controlled low strength material (CLSM)
and the measured thermal conductivity. test cylinders Annual book of ASTM Standards, 4(2), PA
19428-2959, US.
ASTM D5334. 2004. Standard Test Method for determination
time and unconfined compressive strength were veri- of thermal conductivity of soil and soft rock by thermal
fied if they conformed to the requirements of CLSM. needle probe procedure Annual book of ASTM Standards,
The thermal conductivity values of coal ash based 4(2), PA 19428-2959, US.
CLSM were then measured under various degrees Benazzouk, A., Douzane, O., Mezreb, K., Laidoudi, B.,
of saturation. Results indicated that thermal conduc- Que’neudec, M. 2008. Thermal conductivity of cement
tivity depends strongly on the degree of saturation, composites containing rubber waste particles: Experi-
regardless of W/B and W/S ratios or binders used, the mental study and modeling. Constr Build Mater 22(4):
higher the degree of saturation, the higher the thermal 573–579
Gabr M.A. & Bowders J.J. 2002. Controlled low-strength
conductivity. Moreover, it can be found the signifi-
material using fly ash and AMD sludge. J Hazard Mater,
cant reduction in thermal conductivity of both CLSM 76 (2): 251–263.
made with cement (up to 69.75%) and one made with Katz A. & Kovler K. 2004. Utilization of industrial by-
CB (71.88%) between fully saturated and dry states products for the production of controlled low strength
since air is the poorest conductor compared to the materials (CLSM). Waste Management 24(5): 501–512.
water due to its molecular structure. Moreover, the Lee, N.K., Kim, H.K., Park, I.S., Lee, H.K. 2013. Alkali-
multi-regression analysis allowed a development of activated, cementless, controlled low-strength materials
statistically predictive equation for the thermal con- (CLSM) utilizing industrial by-products. Constr Build
ductivity of CLSM based on input of the degree of Master. 49: 738–746
Muhmood, L., Vitta, S., Venkateswaran, D. 2009. Cementi-
saturation along with thermal conductivity under fully
tious and pozzolanic behavior of electric arc furnace steel
saturated and dry state. Finally, it is worth noting that slags. Cem Concr Res. 39 (2): 102–109.
a feasible CLSM made with CB (i.e., without using Mun KJ, Hyoung WK, Lee CW, So SY, Soh YS 2007. Basic
Portland cement) can be used as a backfill material properties of non-sintering cement using phosphogypsum
for trenches of HGCHP. and waste lime as activator. Constr Build Mater 21: 1342–
1350.
Razak HA, Naganathan S, Hamid SNA. 2009. Performance
appraisal of industrial waste incineration bottom ash as
ACKNOWLEDGEMENTS controlled low-strength material. J Hazard Mater, 172(2):
862–867.
This research was supported by a grant (No 16-RDRP- Sengul, O., Azizi, S., Karaosmanoglu, F., Tasdemir, M.A.
B076564-03) from Regional Development Research 2011. Effect of expanded perlite on the mechanical prop-
Program funded by Ministry of Land, Infrastructure erties and thermal conductivity of lightweight concrete.
and Transport of Korean government. Energy and Buildings, 43 (2/3): 671–676.
Siddique, R. & Noumowe, A. 2008. Utilization of spent
foundry sand in controlled low-strength materials and
REFERENCES concrete. Resous, Conser Recylc. 53(1–2): 27–35
Sheen YN, Zhang LH, Le DH 2013. Engineering properties
ACI Committee 229. 1999. Controlled low strength materials of soil-based controlled low-strength materials as slag par-
(ACI 229R-99). American Concrete Institute, Farmington tially substitutes to Portland cement. Constr Build Master.
Hill, MI, USA 48(1–2): 822–829.
ASTM C150. 2004. Standard specification for Portland
cement Annual book of ASTM Standards, 4(2), PA 19428-
2959, US.
622
Evaluation of temperature controlled oedometer test apparatus
Y. Li, M. Karstunen, M. Karlsson & J. Dijkstra

Department of Civil and Environmental Engineering, Chalmers University of Technology,
Gothenburg, Sweden
ABSTRACT: Internationally, geothermal foundations have started to become more common, and their thermal
efficiency is being improved. In Sweden, however, energy piles have not been widely used. This is mainly due to
the remaining uncertainties in the effect of the thermal cycles on the soft soil response. Under repeated generation
and dissipation of excess pore pressures, the microstructure and consequently mechanical properties of soft soils,
especially of sensitive (quick) clays that are common in Sweden are significantly alerted. Temperature controlled
laboratory tests are required to study the thermal response of soft sensitive clay. This paper introduces a modified
oedometer apparatus for the investigation of the temperature dependence of the mechanical properties of clay,
with special focus on the creep rate of sensitive clay.
1 INTRODUCTION very limited and not systematic. In most constitutive

models for soft clay, creep rate depends on the apparent
Geothermal energy gains popularity in the trend of preconsolidation stress. It is generally accepted that
sustainable development. After the temperature con- the preconsolidation stress decreases with tempera-
trolled tests conducted by Campanella and Mitchell ture, which according to the current models would lead
(1968) for soil behaviour at different temperatures to an increasing creep rate. It is, furthermore impor-
during sampling and transportation, more and more tant to study the temperature dependency of creep
laboratory and field tests have been carried out. Exam- without additional stress history change from thermal
ples include studies of soil around nuclear waste consolidation.
(Hueckel & Baldi 1990), energy piles (Bourne-Webb, In this paper two systems to study the effect of
Amatya, Soga, Amis, Davidson, & Payne 2009) and thermal cycling on the soft soil response under 1D
heat storage (Gabrielsson, Lehtmets, & Moritz 1997) compression is discussed. In the first system, the room
where temperature has a significant influence on the temperature will be used as a slow controlled heat-
physical and mechanical properties of the soil. The ing and cooling source without generating excess pore
studies demonstrate that the preconsolidation stress water pressures. A water bath and pump will be used
(Tidfors & Sällfors 1989), the volumetric strains for the second fast control system. The proposed test
(Towhata, Kuntiwattanku, Seko, & Ohishi 1993), the system will be introduced and simulated to study the
stiffness (Cekerevac & Laloui 2004), the excess pore performance of the adapted test cells as a function of
pressures (Abuel-Naga, Bergado, & Bouazza 2007) room temperature change, as well as the effect of a
and the hydraulic properties (Delage, Sultan, & Cui water circulating system.
2000) can change under thermal loading.
In most geothermal structures with competent
ground, the geotechnical function (stiffness and sta-
bility) is not significantly altered. However, in Scan- 2 THERMAL CONSOLIDATION TEST
dinavia, soft sensitive clays are widely spread. For soft
clays with some natural structure, even a small temper- Temperature loading will affect both the soil proper-
ature induced strain can effect the structure of the soil ties and the test apparatus. For a study of long term
particle arrangement. Consequently, the stiffness will soil creep, the effect of cycles with limited tempera-
decrease due to temperature fluctuation, and therefore ture range on the test system requires addtional care
more settlement could be expected. in the design. To study the dependency on temperature
Previous experiments indicate that thermal loads of creep behaviour, long term temperature controlled
produce an acceleration of the mechanical creep phe- laboratory tests are needed.
nomena (Burghignoli, Desideri, & Miliziano 1992, Creep in soil is defined as the on going deforma-
Towhata, Kuntiwattanku, Seko, & Ohishi 1993, tion under constant effective stress. In most cases, the
Akrouch, Sánchez, & Briaud 2014). The amount creep rate is obtained from laboratory tests with an
of data available on creep tests for the appropriate open drainage boundary, such as oedometer tests or
temperature range and the soil type (sensitive clay) is drained triaxial tests with slow strain rate, to get the
623
strain-rate related parameters. To study the temper-
ature dependency of creep, modification of conven-
tional laboratory tests is needed.
Compared to traditional laboratory tests, in thermal
controlled tests an additional variable temperature is
introduced by using heating or cooling devices. There
are many heating systems, for example water circulat-
ing system to exchange heat between the water bath
and the water in the tube (Cekerevac & Laloui 2004)
or to pump hot water in the water bath (Abuel-Naga,
Bergado, & Bouazza 2007), and a heating coil or other
kind of heating wires around the test cell (Delage, Sul-
tan, & Cui 2000), mostly for temperature ranges from
22◦ C to 80◦ C.
During the heating process, excess pore pressures
would be generated, because of the difference in the
thermal expansion of the water and the soil particles. Figure 1. Excess pore pressure generated during heating
It means that under a constant external load, effec- with different rate.
tive stresses would change due to temperature. As
already discussed, creep rate is stress-history depen-
dent. According to the plastic multiplier in most
constitutive models, creep rate depends on the scal-
ing function (F), as shown in Eq. (1), where F is a
function that includes preconsolidation stress, g is the
viscoplastic potential function, σij is the current stress
state and γ is the viscoplastic multiplier.
During cyclic thermal loading, corresponding

effective stress would also change. In this case, the
stress history would be different even with the same
external stress. The heating or cooling rate should be Figure 2. Temperature change rate in temperature room
sufficiently slow to avoid excess pore pressures to
study the temperature effect on pure creep. temperature is increasing proportionally with time, the
In the tests done by Abuel-Naga, Bergado, Bouazza, generated excess pore pressures varies with with dif-
& Ramana (2007), it could take 25 minutes for the heat ferent heating rate. The hydraulic conductivity adapted
to reach the sample. In the drained triaxial tests by in this calculation is sufficiently large for the dissipa-
Burghignoli, Desideri, & Miliziano (2000), in which tion of excess pore pressure, even under the heating
the heating rate was controlled within 3◦ C/h, the rate of 10◦ C/h. Therefore, the maximum heating rate
excess pore pressures would not exceed u/p = 0.02. needs to be decided according to the permeability of
A general function of excess pore poressure generated the soil.
during thermal loading was proposed by Campanella
and Mitchell (1968) for normal consolidated clay:
3 TEST SET UP
3.1 Heating with room temperature

in which α is a coefficient from laboratory test, T The accuracy of temperature control can be up to
is temperature change, u is excess pore pressure and ±0.1◦ C for the water tank controlling system, and can
p is the mean effective stress. In this equation, the reach to ± 0.01◦ C for the temperature measurement.
heating rate is not considered. Even if the heating and For the temperature regulated room in this research,
cooling process is controlled by the room temperature, the rate of changing temperature is not constant. In the
heating rate should be slow enough to keep the stress measurements it ranges from 2◦ C/h to 7◦ C/h, and the
state stable for low permeability sensitive clay. This cooling rate is controlled between −7◦ C/h to −3◦ C/h,
is because the pore pressure generated during heating as illustrated in Fig. 2, where R(◦ C/h) is heating rate.
depends strongly on the hydraulic conductivity of soil, To evaluate the temperature transfer in this heating
which ranges from 10−11 m/s to 10−8 m/s. With high system, a simple model of oedometer cell is analysed
hydraulic conductivity, pore pressures dissipate faster using finite elements. The outside wall of test cell is
than they accumulate. As illustrated in Fig. 1, when the modeled as plastic, the base plate is assumed to be
624
Table 1. Basic parameters for oedometer cell.
Density Young’s modulus Thermal conductivity Heat capacity Expansion coefficient

◦ −1
Component kg/m3 MPa W/(m·K) J/(kg·K) C
Filter stone 2300 5 × 104 1.7 800 1.2 × 10−5

Brass 8500 1.2 × 105 109 380 1.87 × 10−5
Plastic wall 1300 1.4 × 103 0.50 167 8 × 10−5
Saturated sample 1800 10 1.2 800 1.2 × 10−5
Water 1000 2.2 × 103 0.563 4200 3.5 × 10−5
Figure 3. Temperature distribution inside oedometer cell Figure 4. Temperature transfer inside the cell.
after 3 hours of heating.
brass. The basic mechanical and thermal properties of

the materials used are listed in Table 3.1.
In this calculation, the sample is heated by changing
the room temperature. The initial temperature of the
cell is set to be 4◦ C. The temperature on the outside
boundaries are set to a new temperature 25◦ C, with
the heating rate of 7◦ C/h. The temperature measured
in water will not be the temperature in the soil sam-
ple: this is due to the difference between the thermal
properties of cell components, as shown in Fig. 3.
In this system, temperature in soil sample changes
faster than the water in the bath as shown in Fig. 4. Figure 5. Sketch of flow configuration.
According to Fig. 2, even if the room temperature
changes at the highest rate, the heating and cooling rate
of soil sample will not exceed the maximum heating which can transport water with a maximum rate of
and cooling rate 7◦ C/h. 0.4 m3 /h.
3.2 Heating with water tank 3.3 Accuracy and error

A safe heating-rate range that would not change the Due to the temperature fluctuation, the oedometer
stress history is required for the investigation of pure components such as the ring, the filter stone, the bot-
creep for the studied soft clay. On the other hand, to tom plate and the loading cap would be influenced
capture the excess pore pressures generated by the by the temperature. Therefore, the deformation of soil
heating, a circulation system with water tank and pump sample cannot be only evaluated by measuring the dis-
is used to change the temperature with a faster rate, placements at the loading cap. Tests without sample
as illustrated in Fig. 5. This system can control tem- using the pumping system as shown in Fig. 5, and a
perature change with higher rates than room heating calculation should be done before the real tests to eval-
system. In the pumping system, temperature of water uate and calibrate the influence of temperature on each
is controlled in storage tank and circulated by a pump part of the oedometer test.
625
3.3.1 Displacement 4 TEST WITH SOIL SAMPLE
The displacement can be affected by every part of the
loading system, especially when the temperature of To test the experimental range of the heating rate for
the room is used to cool down and heat up the sample. sensitive soft clay, one single test under thermal and
The whole loading frame, as well as the oedometer mechanical loading was done as illustrated in Table 2.
cell, could deform under the temperature change. Only excess pore pressure and displacement are mon-
The average inner diameter of the oedometer ring itored. The temperature was only increased after the
is 50.2 mm and the height is 20 mm. In theoretical cal- excess pore pressures due to the mechanical loading
culation, the inner diameter would change 0.01 mm were dissipated. With water tank, the heating rate can
under 20◦ C temperature change. If the volume of the be controlled from 5◦ C/hour to 85◦ C/hour as shown in
sample is assumed to be constant, this expansion of Fig. 7.
ring would cause 0.008 mm change in the height of
the soil sample, which is 0.04% strain. In the on going
oedometer test, the oedometer ring is constrained by 4.1 Pore pressures during heating and cooling
screws and the top ring. For this reason, the expansion
In order to control the pore pressures, the heating rate
of the brass in the air is different from the brass in
should be properly regulated. Based on Fig. 8, the gen-
the cell.
erated pore pressure can not just be expressed with
The same applies to the loading piston rod. The
Eq. 2. The heating or cooling rate should also be con-
length can be extended during heating and shortened
sidered in the formulation. As shown in Fig. 8, the
during cooling. Because of the constraint along the
generated excess pore pressure varies with the initial
piston rod, the real extended length would be smaller
stress state. In this trial test, the excess pore pressure is
than the theoretical values. The additional displace-
high at high stress state, even though the heating rate
ment of the loading system should be corrected for in
is much lower at the high stress state.
the analyses. A heating test without soil sample was
During cooling process, the cooling rate is crit-
performed to calibrate the transducers and oedometer
ical for stress states that are larger than the
cell components. A linear relation between tempera-
preconsolidation stress of the soil sample. Negative
ture and additional displacement can be adapted for
out set-up according to the calibration result shown in
Table 2. Test procedure with water tank.
Fig. 6.
Vertical load Temperature
◦
Loading stage kPa C
3.3.2 Pore pressures
A pore pressure transducer is placed outside the water 1 12.5 20
bath and was connected to the bottom of the sam- 2 12.5 30
ple. Therefore, the water pressure transducer would be 3 25 30
influenced by the temperature change earlier than sam- 4 25 40
ple, mainly due to the water expansion inside the tube 5 50 40
6 50 50
connecting the soil and the transducer. In the test with- 7 100 50
out soil sample, excess pore pressure fluctuated within 8 100 60
1 kPa. Therefore, the excess pore pressure caused by 9 100 50
transducer under thermal loading can be neglected. 10 200 50
11 200 40
Figure 6. Displacement of loading system under tempera-

ture change during calibration. Figure 7. Heating rate of water bath in oedometer cell.
626
ACKNOWLEDGEMENTS
The research presented in this paper is funded by For-

mas (the Swedish Research Council for Environment,
Agricultural Sciences and Spatial Planning) as part
of a project on Geothermal foundations on soft soils
(243-2013-771).
REFERENCES
Abuel-Naga, H., D. Bergado, A. Bouazza, & G. Ramana
(2007). Volume change behaviour of saturated clays under
drained heating conditions: experimental results and con-
stitutive modeling. Canadian Geotechnical Journal 44(8),
942–956.
Figure 8. Excess pore pressure in heating process. Abuel-Naga, H. M., D. T. Bergado, & A. Bouazza (2007).
Thermally induced volume change and excess pore water
pressure of soft bangkok clay. Engineering Geology 89(1),
144–154.
Akrouch, G. A., M. Sánchez, & J.-L. Briaud (2014). Thermo-
clays. Acta Geotechnica 9(3), 399–412.
Bourne-Webb, P., B. Amatya, K. Soga, T. Amis, C. Davidson,
& P. Payne (2009). Energy pile test at lambeth college,
london: geotechnical and thermodynamic aspects of pile
response to heat cycles. Géotechnique 59(3), 237–248.
Burghignoli, A., A. Desideri, & S. Miliziano (1992).
Deformability of clays under non isothermal conditions.
RIG 4, 92.
Burghignoli, A., A. Desideri, & S. Miliziano (2000). A labo-
ratory study on the thermomechanical behaviour of clayey
soils. Canadian Geotechnical Journal 37(4), 764–780.
Campanella, R. G. & J. K. Mitchell (1968). Influence of
temperature variations on soil behavior. Journal of Soil
Figure 9. Negative pore pressure in cooling process. Mechanics & Foundations Div.
Cekerevac, C. & L. Laloui (2004). Experimental study
pore pressures can generate under cooling. Effective of thermal effects on the mechanical behaviour of a
stress will increase under cooling, which can lead to clay. International journal for numerical and analytical
additional irrecoverable strains when initial stress is methods in geomechanics 28(3), 209–228.
Delage, P., N. Sultan, & Y. J. Cui (2000). On the ther-
larger than the preconsolidation stress, as shown in mal consolidation of boom clay. Canadian Geotechnical
Fig. 9. Journal 37(2), 343–354.
Gabrielsson, A., M. Lehtmets, & L. Moritz (1997). Heat Stor-
age in Soft Clay: Field Tests with Heating (70 OC) and
5 DISCUSSION AND CONCLUSIONS Freezing of the Soil. Swedish Geotechnical Institute.
Hueckel, T. & G. Baldi (1990). Thermoplasticity of satu-
A systematic study on the effect of temperature cycles rated clays: experimental constitutive study. Journal of
(4◦ C to 25◦ C) on the long-term response of natural Geotechnical Engineering 116(12), 1778–1796.
and reconstituted high plasticity, high sensitivity clay Tidfors, M. & G. Sällfors (1989). Temperature effect on
preconsolidation pressure. ASTM Geotechnical Testing
in one dimensional compression is ongoing. These Journal 12(1).
preliminary results show the feasibility of the pre- Towhata, I., P. Kuntiwattanku, I. Seko, & K. Ohishi (1993).
sented test set up to study pure creep as well as pure Volume change of clays induced by heating as observed in
creep with thermal consolidation as a function of consolidation tests. Soils and foundations 33(4), 170–183.
heat cycles. Experimental errors have been discussed
and the necessity of an advanced system analysis is
demonstrated.
627
Numerical modelling of the cyclic behaviour of rock material in the context

of underground pumped storage hydroelectricity
B. Cerfontaine, F. Collin & R. Charlier

Geomechanics and Engineering Geology, University of Liege, Belgium
ABSTRACT: The use of old mines as a reservoir in the context of pumped storage hydroelectricity rises
the question of their stability especially upon cyclic loading due to filling and emptying the mines. This paper
presents the formulation of a new constitutive law able to represent the inherently cyclic behaviour of rocks at low
confinement. The main features of the behaviour evidenced by experiments in the literature depict a progressive
degradation and deformation of the material with the number of cycles. A constitutive law based on a boundary
surface model is developed. It represents the brittle failure of the material as well as its progressive degradation. It
also encompasses a fatigue resistance under which no degradation of the material holds. Parameters are calibrated
with respect to laboratory experiments from the literature. A parametric study is carried out to highlight the role
of different parameters.
1 INTRODUCTION the rock behaviour upon both monotonic and cyclic

loading.
Over the last decade, the part of renewable energy has Experimental results are firstly presented to intro-
considerably grown in the energy mix. However solar duce and detail the main features of the rock behaviour.
and wind energy particularly suffer from a crucial A triaxial formulation of the constitutive law is then
drawback. The period of maximum production does detailed. Finally a comparison with experimental data
not necessarily coincide with the period of maximum is provided together with a parametric study illustrat-
consumption (Førsund 2015). The pumped-storage ing how the parameters influence the results.
hydroelectricity stores energy as gravitational poten-
tial energy. Water is pumped from a lower reservoir
into a higher reservoir when electric power is avail- 2 EXPERIMENTAL REVIEW
able at a low cost (off-peak of consumption) (Steffen
2012). During the peak of energy consumption, tur- 2.1 Monotonic behaviour
bines produce energy at a higher cost, ensuring the
economic viability of the system. The overall allows a Failure of brittle rock is mainly driven by the gen-
storage of the energy. eration of micro-cracks and cracks within the rock
The idea arose to reuse disused mines, pits and sample. An example of such a behaviour is depicted
quarries as lower reservoirs for pumped-storage. in Figure 1 for a simple compression experiment.
Ensuring the overall stability of mine galleries and Following the pioneering work of (Bieniawski 1967a,
quarries is one of the main challenges for engineers. Bieniawski 1967b) the brittle behaviour of rocks is
Indeed the numerous pumping and filling phases of classically delineated into five phases for simple
the old galleries modify continuously the stress and compression tests.
pore water distributions within the surrounding rock. (I) The first phase corresponds to the closure of the
This results into a fatigue loading of the rock mate- microcracks initially existing within the rock sam-
rial whose resistance has been shown to be lower than ple. Acoustic emissions (AE) recorded in (Cai,
the monotonic resistance (Royer-Carfagni & Salvatore Kaiser, Tasaka, Maejima, Morioka, & Minami
2000, Erarslan & Williams 2012). 2004) correspond to sound waves emitted when
This papers aims at developing a constitutive law to microcracks or fractures are created or propa-
reproduce the fatigue behaviour of rock materials at gated. During this first phase, the number of AE is
low confinement, representative of the overall prob- almost equal to zero. It lasts up to a vertical stress
lem. The formulation proposed hereinafter is based on equal necessary to ensure the crack closure σcc .
the bounding surface model framework already devel- (II) The second phase is purely elastic. There is no
oped for sands (Taiebat & Dafalias 2008), assuming initiation of cracks nor propagation. There are still
the material is rate independent. The original model is no AE and the volumetric deformation is equal to
modified to take into account the salient features of the elastic one.
629
Figure 1. Stress-strain diagram exhibiting the differ-
ent stages of the monotonic behaviour of brittle rocks, Figure 2. Brittle failure in Lac du Bonnet granite, (Martin
(Martin 1997, Cai, Kaiser, Tasaka, Maejima, Morioka, & 1997).
Minami 2004).
(III) From a vertical stress equal to σi , microcracks

initiate or propagate and are mainly parallel to
the applied stress. They induce a volumetric dila-
tion. The number of acoustic emissions rises up
but remains limited. This phase is termed “stable
crack growth”.
(IV) During the last phase, the coalescence of all
microcracks takes place at σd around 0.7 to 0.8
times the peak strength (Bieniawski 1967a, Mar-
tin 1997, Cai, Kaiser, Tasaka, Maejima, Morioka,
& Minami 2004). The dilation strongly increases
as well as the number of AE.
(V) Finally failure is reached at σf . The rock sample
may exhibit shear bands or macrocracks.
Figure 3. Monotonic and cyclic simple compression (con-
The brittle behaviour of rock is not limited to stant amplitude) tests on Lorano marble, data traced from
the simple compression experiments. It is observed (Royer-Carfagni & Salvatore 2000).
also when the sample is subjected to a confinement
as shown in Figure 2. It is known that the ductil-
ity increases with confinement (Liu & He 2012).
The failure may become totally ductile at very high
confinement. However this paper will focus on low
confinement behaviours. Many other features are not
illustrated here but are relevant for the formulation
of a constitutive law such as the material (Attewell
& Farmer 1973) or load-induced anisotropy (Benz &
Schwab 2008).
2.2 Cyclic behaviour

The typical behaviour of a sample of Lorano marble is
illustrated in Figure 3 for both monotonic and constant
cyclic amplitude experiments. The monotonic failure
is brittle and presents more plastic deformation lat- Figure 4. Comparison of monotonic and cyclic indirect ten-
erally than vertically. The cyclic experimental results sile strength results of Brisbane tuff, (Erarslan & Williams
highlight an accumulation of plastic deformation cycle 2012).
after cycle. The increment of deformation increases
progressively and cycles are more open at the end. load. It is observed in (Erarslan & Williams 2012) that
A cyclic indirect tensile test on Brisbane tuff is pro- the cyclic resistance is systematically lower than the
vided in Figure 4. This test consists in a ramp signal, monotonic one. Moreover the modes of failure are
i.e. the load oscillates between an increasing average different. The monotonic samples present a straight
630
Figure 6. Description of the different surfaces of the model
(compression side): α back-stress tensor, My yield surface,
Mb bounding surface, Mh hardening surface, Mpc decohesion
Figure 5. S-N curve for indirect tensile strength of Brisbane surface.
tuff, (Erarslan & Williams 2012).
3.2 Yield, hardening and other surfaces
crack while cyclically loaded samples are surrounded
by debris and crushed rock material. This suggests that The model is defined by different surfaces in the p−q
much more microcracks are generated. plane. Only the surfaces on the compression side are
It is finally observed that upon constant amplitude presented in Figure 6 for clarity. Their common apex
cycles, a threshold exists below which there is no is located on the p-axis at coordinate −pc which is a
failure of the material even upon a large number of function of the cohesion c and the friction angle φ
cycles (Erarslan & Williams 2012). Figure 5 presents
the number of cycles as a function of the amplitude
applied. When the amplitude is below 70% of the
monotonic resistance, no failure was observed even
where φ is the friction angle.
after more than 500000 cycles. This threshold may be
The yield surface has an opening of 2 My . Its
defined as the fatigue resistance of the material.
mathematical formulation reads
3 TRIAXIAL FORMULATION OF A
CONSTITUTIVE LAW
where α is termed the back-stress tensor of the surface.
3.1 Triaxial notations It describes the variation of its position and is limited
by
In triaxial conditions, only the diagonal components
of the deformation and stress σ tensors are different
from zero. Two stress invariants describes the stress
state of the material
where Mb is the slope of the bounding surface and

δb is the ratio between the compression and extension
openings of the bounding surface. The slope Mb is
where σ1 and σ3 are respectively the vertical and lateral related to the friction angle through
effective stresses applied to the sample. The stresses
are gathered into the vector sT = [q, p] where T is
the transpose operator. A reduced deviatoric stress ξ is
defined according to
The yield surface is then a function of two internal
variables: α and pc . The bounding surface is described
by
where pc will be further defined. Similarly strain
invariants are defined according to
and depends only on the internal variable pc .

The last two surfaces are termed hardening (opening
Mh ) and decohesion of (opening Mpc ) surfaces. They
where 1 and 3 are the vertical and lateral defor- are homothetic to Equation (11).
mations. They are assembled into the vector eT = The purpose of the hardening surface is to allow a
[q , v ]. brittle failure of the material. Indeed in the bounding
631
Such a definition allows different stiffness in loading
and unloading.
The isotropic hardening rule simply reads
Figure 7. Difference between ductile and fragile failure.

where Ac is a parameter ruling the rate of decohesion,
surface models (Taiebat & Dafalias 2008), the plastic the operator X = 0.5 (|X | + X ) is the Mc Cauley
modulus is a function of the distance between the yield brackets and
and the bounding surface α + My − Mb . When those
surface are very close, the plastic modulus tends to
zero and the failure is very ductile as shown in Figure 7.
If the plastic modulus is computed with respect to the
hardening surface Mh ≥ Mb , the bounding surface is The hardening laws are written in compact form
reached before the plastic modulus is set to zero and such that
the failure is brittle.
The decohesion surface of slope Mpc defines the
fatigue resistance of the rock as a function of its
confinement. If the stress path lies within the deco-
hesion surface, there is no decrease of the cohesion 3.3.3 Plastic multiplier
variable pc . Assuming the classical additive decomposition of
deformation between plastic and elastic components
3.3 Yield surface (Simo & Hughes 1998), the triaxial stress-strain rela-
tion is written
The yield surface is activated in both monotonic
and cyclic loading, before failure. It is subjected to
kinematic (on α) and isotropic (on pc ) hardening.
where
3.3.1 Flow rule
The flow rule is associated in the deviatoric direc-
tion and non-associated in the volumetric direction,
which is classical in many cyclic constitutive laws. The
variations of the plastic components of deformation is where G is the shear modulus and K the bulk modulus.
defined such that The consistency condition is then written
where λ̇ is the variation of the plastic multiplier and Inserting Equations (12), (20) and (21) into Equation
Ad is a parameter of dilatancy. (23) leads to the following expression of the plastic
multiplier
3.3.2 Hardening of internal variables
The kinematic hardening law of the yield surface reads
where N is a parameter that rules the stiffness

and brittleness of the material, and h is defined as
(Taiebat & Dafalias 2008) where
where b0 is a material parameter. The definition of the and

αh variable depends on the loading or unloading
632
Table 1. Parameters of the model.
Elasticity My [−] E [GPa] ν [−]

0.1 70 0.16
Bounding Mb [−] pres [MPa] cb [−]
1.4 5 1
Decohesion Mpc [−] Ac [−] Ac2 [−]
0.5 20 500
Hardening Mh [−] Nexp [−] b0 [−]
1.5 2 3 · 104
Volumetric Ad0 [−]
−1.8
Table 2. Initial internal variables.
α0 [−] pc0 [MPa]

0 34
Figure 8. Comparison of monotonic simple compression
simulation and experimental results, stress deviator vs. axial
deformation.
3.4 Bounding surface
When the yield surface reaches the bounding one, the
second mechanism is activated. The kinematic harden-
ing of the yield surface is stopped since it has reached
its limit, i.e. α̇ = 0. However the cohesion continues
to degrade but at a higher rate. The cohesion degrada-
tion is similar to Equation (17) but the parameter Ac is
replaced by Ac2 such that
Moreover the derivatives of f y in the plastic multiplier

definition, Equation (24), are replaced by derivatives
with respect to f b .
Figure 9. Comparison of monotonic simple compression

4 EXAMPLES OF SIMULATIONS simulation and experimental results, volumetric vs. axial
deformation.
Comparison with experimental results are based on
results published in the literature for the Lorano mar- The brittle failure is well captured in terms of both
ble material. The simple compression monotonic and peak resistance (∼ 85MPa) and maximum deforma-
cyclic stress paths are presented in (Royer-Carfagni tion (∼ 0.22%). The behaviour of the material was set
& Salvatore 2000). Elastic and failure parameters identical in compression and extension since no data
are extrapolated from (Cattaneo & Labuz 2001, are available in extension.
Stavropoulou, Liolios, & Exadaktylos 2004, Ferrero, The post-peak behaviour of the samples was not
Migliazza, & Spagnoli 2009). investigated in (Royer-Carfagni & Salvatore 2000).
Parameters and initial internal variables resulting However according to other studies the unconfined
from the calibration process are provided in Tables 1 failure is probably brittle (Cattaneo & Labuz 2001). A
and 2. The following sections compares experimental steep post-peak behaviour was assumed with a slow
results and numerical simulations. A parametric study but non zero cohesion residual parameter. It is shown
is also carried out to highlight the main role of some in Figure 8 for different values of parameter Ac2 .
parameters. The dilatancy parameter is calibrated with respect
to the volumetric deformation results presented in
Figure 9. Only the pre-peak part of the results is exhib-
4.1 Comparison with experimental data
ited to prove the good agreement of simulation results
Results of monotonic simulations are compared with with experiments.
experimental data in Figures 8 and 9. Elastic param- The evolution of the pc internal variable is provided
eters as well as friction angle and cohesion were in Figure 10. The pre-peak variation of the cohesion pc
estimated from the literature. Indeed triaxial tests is very limited. It steeply decreases afterwards due to
at different confinement pressures would have been the activation of the boundary surface. Therefore this
necessary. is a consequence of the failure and not a cause.
633
Figure 10. Evolution of the pc internal variable during the Figure 12. Evolution of the pc internal variable during the
simple compression test. cyclic simple compression test.
Figure 11. Comparison of cyclic simple compression sim- Figure 13. Influence of the Mh parameter.
ulation and experimental results (envelope only), stress
deviator vs. axial deformation. cohesion is the cause of the failure. This is consistent
with the most diffusion and progressive mechanism of
Since pc is almost not decreased at failure, the cohe- failure. The post-peak behaviour is not modelled since
sion at peak can be used to estimate the initial value the simulation is stress-driven.
of pc . This is true especially if the degradation rate Ac The non-linear unloading part of the cycles is not
is low. reproduced. Indeed, this feature is probably due to the
It was impossible to reproduce the cyclic experi- re-opening of crack within the sample. The crack clo-
ment with the same set of parameters due to the large sure phase is not captured by the model. Therefore the
variability of experimental results. Indeed it can be re-opening phase is not. However since the opening-
observed that monotonic test and the first compression closure of the cracks as well as friction dissipate
of the cyclic test diverge in Figure 3. Results of a sim- energy, modelling this phase could be an improvement
ulation are provided in Figure 11 assuming Mh = 1.7, of the model.
b0 = 105 , Mb = 1.6 and Ac = 150.
The experimental result exhibit a failure of the sam-
4.2 Brittleness and stiffness
ple after almost 30 cycles. The numerical simulation
leads to failure after 25 cycles. Numerical results well The brittleness of the failure depends on two parame-
capture the accumulation of deformation. However the ters: Mh and the exponent N . Results are provided for
deformation is a bit too high. The difference between different values of these parameters. The decohesion
lateral and axial deformation is also well reproduced. parameter Ac is set to zero to simplify the analysis, i.e.
The opening of the last cycles is increasing which was there is no degradation of cohesion before failure. Both
required. parameters modifies the plastic deformation reached
Failure is finally reached due to the progressive at failure, as shown in Figures 13 and 14. When Mh
alteration of the cohesion (pc internal variable), as tends to Mb , the original bounding surface concept
depicted in Figure 12. In this case, degradation of the is recovered and the yield surface tends towards the
634
Figure 16. Influence of the Ad parameter.
Figure 14. Influence of the N parameter.
Figure 17. Influence of the Mpc parameter.
Figure 15. Influence of the b0 parameter.
bounding one without reaching it, the deformation is

infinite.
When only the exponent N is modified, the plastic
deformation increases at failure. However the defor-
mation is always finite. This parameter also increases
the divergence from linearity since the plastic multi-
plier is a function of |α − αh |N . Therefore as soon as
|α − αh | < 1, the loss of stiffness is higher for higher
exponents.
The b0 parameter directly influences the stiffness
of the elastoplastic behaviour. For the present exam-
ple, the elastic limit is overpassed for a stress equal to
20MPa. Divergence of the elastoplastic behaviour is
more obvious for lower b0 , as shown in Figure 15. Figure 18. Influence of the Ac parameter.
4.3 Dilatancy
Figures 17 and 18. Mpc is a threshold beyond which
The parameter Ad classically rules the plastic volumet- degradation of the cohesion starts. Therefore the lower
ric behaviour. It is dilative if Ad < 0 and contractive this value, the higher the degradation caused at each
otherwise as shown in Figure 16. cycle. This defines the fatigue resistance of the rock.
The Ac parameter rules the degradation rate of
the cohesion parameter. It matters only when the
4.4 Decohesion
stress path overcomes the degradation limit Mpc . This
Parameters Mpc and Ac regulate the degradation of the parameter indirectly influence the number of cycles
cohesion. Results of cyclic loading are provided in before failure that can be afford by a rock sample.
635
5 CONCLUSION Bieniawski, Z. (1967b). Mechanism of brittle fracture of rock:
PART II Experimental studies. International Journal of
This paper presents the mathematical formulation of Rock Mechanics and Mining Sciences 4, 407–423.
a new model describing the cyclic behaviour of rock Cai, M., P. Kaiser, Y. Tasaka, T. Maejima, H. Morioka, &
M. Minami (2004, jul). Generalized crack initiation and
materials. It is based on the boundary surface frame-
crack damage stress thresholds of brittle rock masses near
work developed for sands. It reproduce the main underground excavations. International Journal of Rock
features of the monotonic and cyclic behaviours of Mechanics and Mining Sciences 41(5), 833–847.
rock material at low confinement, Cattaneo, S. & J. Labuz (2001). Damage of marble from
cyclic loading. Journal of materials in civil engineer-
• brittle failure of the material,
ing 13(December), 459–465.
• post-peak behaviour, Erarslan, N. & D. J. Williams (2012, nov). Investigating the
• accumulation of plastic deformation upon cyclic Effect of Cyclic Loading on the Indirect Tensile Strength
load, of Rocks. Rock Mechanics and Rock Engineering 45(3),
• progressive opening of the cycles, 327–340.
• degradation of the cohesion, Ferrero, A., M. Migliazza, & A. Spagnoli (2009). The-
• fatigue resistance of the material. oretical modelling of bowing in cracked marble slabs
under cyclic thermal loading. Construction and Building
A comparison with experiments exhibits that the Materials 23(6), 2151–2159.
model well reproduces the monotonic results. However Førsund, F. (2015). Hydropower economics, Volume 217.
additional triaxial experiments should be necessary Springer.
to calibrate more accurately all the parameters. The Liu, E. & S. He (2012). Effects of cyclic dynamic loading
cyclic behaviour is qualitatively captured. However the on the mechanical properties of intact rock samples under
simulation leads to excessive plastic deformation. confining pressure conditions. Engineering Geology 125,
81–91.
Further developments could be interesting. The
Martin, C. D. (1997, oct). Seventeenth Canadian Geotechni-
damaging of elastic properties or modification of the cal Colloquium: The effect of cohesion loss and stress
plastic modulus may improve the model. A confine- path on brittle rock strength. Canadian Geotechnical
ment dependency of the elastic parameter may also Journal 34(5), 698–725.
be introduced. The modelling of the hysteresis of Royer-Carfagni, G. & W. Salvatore (2000, oct). The char-
the cycles could be of interest. However all of these acterization of marble by cyclic compression loading:
modifications require more experimental results to be experimental results. Mechanics of Cohesive-frictional
calibrated which reduces their applicability to actual Materials 5(7), 535–563.
projects. Simo, J. & T. Hughes (1998). Computational inelasticity.
Springer-Verlag, New-York.
Stavropoulou, M., P. Liolios, & G. Exadaktylos (2004).
Calibration of the Triaxial Hyperbolic Mohr-Coulomb
REFERENCES Elastoplastic Model Parameters on Laboratory Rock
Mechanics Tests. International Journal of Geomechan-
Attewell, P. & W. Farmer (1973). Fatigue behaviour of rock.
ics 12(December), 618–631.
International Journal of Rock Mechanics and Mining
Steffen, B. (2012). Prospects for pumped-hydro storage in
Sciences 10, 1–9.
germany. Energy Policy 45, 420–429.
Benz, T. & R. Schwab (2008). A quantitative comparison
Taiebat, M. & Y. Dafalias (2008). SANISAND: simple
of six rock failure criteria. International Journal of Rock
anisotropic sand plasticity model. International Journal
Mechanics and Mining Sciences 45, 1176–1186.
for Numerical and Analytical Methods in Geomechan-
Bieniawski, Z. (1967a). Mechanism of brittle fracture of
ics 32(March 2007), 915–948.
rock: PART I Theory of the fracture process. Interna-
tional Journal of Rock Mechanics and Mining Sciences 4,
395–406.
636
A small geothermal project of low enthalpy in Bío Bío sand
V. Hernández & D. Böttcher

BBSolutions, Providencia, Chile
F.A. Villalobos
Laboratory of Geomaterials, Faculty of Engineering, Catholic University of Concepción, Chile
ABSTRACT: There are currently very high levels of air pollution in many cities in the centre and south of Chile
due to mostly the intensive use of burning wood heating systems. It is found that the cheapest choice in terms of
heating efficiency and price per kWh is a shallow geothermal system. Consequently, a small pilot geothermal
project of low enthalpy was designed and built in order to provide an economical and non polluting heating
system in the town of Santa Juana, Chile. A small 3.92 kW geothermal heat pump was connected with a net of
shallow soil heat absorbers, which were buried 1.5 m in a non-plastic medium dense silt. Thermal properties of
this fine deposit of the Bío Bío River, were obtained. Preliminary monitoring results have shown that this shallow
geothermal system is able to sustain the interchange circulating water temperature at night to keep around 19◦ C
indoors when the outside temperature is around 10◦ C.
1 INTRODUCTION Silva & Arcos (2011), Schueftan & González (2013,

2015), among others.
The current environmental situation in the south of It is worth pointing out that changing the current
Chile, especially during autumn and winter, is very polluting practice of household heating for a non-
serious due to high levels of air pollution. There are polluting one will save also thousands of millions of
cities such as Talca, Chillán, Temuco, Valdivia, Osorno dollars spent every year on breathing and lung health
and Coyhaique where the high level of particulate mat- programmes. Moreover, it would save the indiscrimi-
ter PM comes mainly from combustion of wood fuel nate cut down of local native forest, which can take
in household stoves. For instance, more than 93% of hundreds of years to grow. Other advantage it will
PM from firewood burning in residential sectors of be the elimination of the fire danger when house-
Temuco corresponds to PM2.5 , which has the most hold stoves and chimneys are out of control burning
serious health effects in human beings (Schueftan & housings, and even wounding and killing people. Fur-
González, 2013). Although there are governmental thermore, wood combustion requires constant and
strategies to reduce the air pollution, they try only to regular supervision, from delivery, cubic meters of dry
mildly ameliorate and react to the problem rather than storage to the insertion of previously chopped pieces
avoid it from its origin (DS 39, 2014). Recommen- into the stove several times during each day.
dations such as the use of drier wood and brand new It will be shown that one of the best possible solu-
stoves are not really solving the problem (Schueftan & tions for this environmental problem is the use of a
González, 2013, 2015). heating system which is not only clean but it uses
PM2.5 limit levels as day average in µg/m3 have energy stored in shallow ground, which is provided by
been stipulated as: good: 50; fair: 80; warning: 110; the sun, hence renewable. It has to be pointed out that
pre-emergency: 170; emergency: 250; health emer- this geothermal system does not need a hot ground
gency: beyond 250. Available data from measurements as in very deep borehole systems. In the following,
since 2003 until 2012 shows that there has been a an example case of low enthalpy geothermal system
steady increase of PM2.5 and PM10 in the cities men- is presented. The idea is to show the possibility of
tioned above. For instance, in 2012 Temuco had worse developing this system in a small scale project with-
records than Beijing in China, with 7 (Temuco)/2 out requiring large excavations such as expensive well
(Beijing) warnings, 10/6 pre-emergencies and 9/7 systems or large areas for pipe systems.
emergencies, totalling 26/15 exceeding limit levels
with a peak value of 956/886 µg/m3 . This air condition
is leading to serious health problems in the population, 2 LOW ENTHALPY GEOTHERMAL PROJECTS
especially babies and old people. More details of the
current air pollution situation coming from household Low enthalpy geothermal projects are relatively new
heating in Chile can be found in Celis et al. (2004), in Chile with no so many examples. They have been
637
Figure 2. Grain size distribution of soil around the collec-
tors.
But in other closed circuit, the fluid of the heat pump

towards indoors can be distributed from 25 to 45◦ C
in the same way as a central heating will do. That
is, through floor, wall or ceiling radiators or even for
Figure 1. Sketch of shallow heat from the ground (Geolog- shower hot water.
ical Survey from Ireland 2015). Similarly like in any other heating system it is
important that the building insulation is sufficiently
efficient to avoid letting the heating to leave quickly
expensive and mostly based on boreholes for large the rooms. Roof, windows, walls and floors should
buildings for schools, hotels and universities. slow heating from dissipating to the exterior. It will be
The aim of this project is to make accessible to much shown in the following that despite the semidetached
more people a non-polluting heating system, using housing not having adequate insulation, the proposed
renewable energy, therefore, reducing the extensive geothermal system works anyway.
use of polluting burning wood. In addition, the solution
has to be economically affordable and therefore low
3 SANTA JUANA PILOT PROJECT
inversion and installation costs should be required as
well as low or no maintenance cost. In fact, geothermal
3.1 Ground properties
heating turns out to be the cheapest option according to
the price of heating per kW as shown in Table 1, where Santa Juana is a town located 48 km south of Concep-
prices are presented for different options of heating ción, Chile. It lies next to the south bank of the Bío
fuel and type of heater. Bío River. The ground where the collectors extract the
Some of the advantages of a geothermal system are heat in Santa Juana corresponds to a fine material with
the possibility of invert heating to cooling, no nega- around half of the particles having sizes smaller than
tive environmental impact, no need for large physical 0.075 mm. Figure 2 shows the grain size distribution
space for the installation and equipments, no visual of the fine material where the row of collectors are
impact, low or nil maintenance cost and high energy buried.
efficiency and COP around 4 (e.g. Inalli & Esen, 2004; The specific gravity of the soil particles is
Omer, 2008). The latter means that a heat pump needs Gs = 2.82, revealing the presence of basalt, the max-
1 kW of electricity to obtain around 3 kW from the imum and minimum void ratios are emax = 1.09
ground to generate 4 kW of thermal energy (hot water), and emin = 0.63 and the water content is w = 16%.
resulting in a coefficient of performance COP = 4. Although the soil surrounding the collectors is not
Figure 1 shows a sketch of extraction of heat using saturated and the soil particles are very fine, the soil
shallow collectors with depths between 0.8 and 1.6 m coefficient of permeability k can be estimated by using
instead of deep wells of up to 100 m, which can be the well known Hazen formula based on an effec-
prohibitively expensive due to complicated boring, tive diameter as d10 = 0.031 mm, resulting for a loose
installation, operation and maintenance. The fluid cir- soil in k = 10−5 m/s. For a dense soil, k = 4 · 10−6 m/s
culating in the buried collectors is water with ethylene for d15 = 0.035 mm. Using the Kozeny-Carman for-
glycol to avoid freezing. mula with a grain shape factor of 6.4 for worn grains
This fluid when passes through the heat pump is (see Figure 3), results in not so different values,
compressed to in this form extract the heat coming k = 1.8 · 10−5 m/s for a void ratio of 0.9 (40% of rel-
from the ground. ative density) and k = 10−5 m/s for a void ratio of 0.7
During autumn and winter the shallow ground tem- (85% of relative density). Due to the large percentage
perature can fluctuate between 5 and 15◦ C, and after of fine particles the soil classifies as low plasticity silt
passing through the heat pump the fluid temperature ML according to the USCS. But in fact, they can be
drops significantly and it is put back in the collectors treated as a fine sand since the particles do not have
to recover extracting again the ground temperature. plasticity at all as can be observed in Figure 3.
638
Table 1. Cost of heating for different fuel and type of heater (Jahnke, 2014).
Price heating,
Fuel, heater type Energy price, $1 Net energy, kWh Efficiency, %2 $/kWh
1 kg wood 20% humidity, old stove 150 4.28 60 58.4

1 kg wood 50% humidity, old stove 150 2.42 60 103.3
1 kg wood 20% humidity, modern stove3 150 4.28 80 43.8
1 kg wood 50% humidity, modern stove 150 2.42 80 77.5
1 kg wood 20% humidity, old chimney 150 4.28 15 233.6
1 kg wood 20% humidity, old salamandra4 150 4.28 35 100.1
1 kg sawdust pellets, modern stove 215.25 4.50 85 56.3
1 kg liquid gas without gas release outside 1044.44 12.87 100 81.2
1 kg liquid gas with efficient gas release 1044.44 12.87 80 101.5
1 kg liquid gas with not so efficient gas release 1044.44 12.87 60 135.3
1 kg catalytic gas without gas release outside 1088.89 12.87 100 84.6
1 lt ethanol without gas release outside 1400 5.49 100 255.1
1 lt kerosene without gas release outside 730 9.58 100 76.2
1 lt kerosene with efficient gas release 730 9.58 80 95.3
1 lt kerosene with not so efficient gas release 730 9.58 60 127.0
Electric radiator, kWh normal price 81.33 1 100 81.3
Electric radiator, kWh overconsumption price 112.34 1 100 112.3
Heat pump, kWh normal price 81.33 1 300 27.1
Heat pump, kWh overconsumption price 112.34 1 300 37.5
1
$ is for Chilean pesos, $720 = 1 U$, $760 = 1 Euro.
2
energy effectively used for heating, considering losses.
3
modern double chamber air-tight steel stove.
4
steel and iron old model of wood heater and cook stove.
Geologically speaking, the sand has been trans-

ported by the Laja and Bío Bío Rivers from the Antuco
volcano erupted materials 200 km upstream in the
Andes Mountains. This alluvial deposit is composed
mainly of basalt and feldspar, for that reason its colour
is predominantly grey, although there are also parti-
cles with other colours such as brown and white as
previously described by Ayala et al. (2015) for coarser
grain size distributions. Figure 3 shows an amplified
photograph of the soil grains extracted at 1 m depth
next to the collector’s trench.
The Santa Juana fine soil thermal conductivity
λ was measured using a manual thermal analyser
KD2 PRO (DECAGON, 2008; ASTM D 5334, 2008).
This instrument also allows the measurement of the
thermal resistivity ρ, heat capacity C, thermal dif-
fusivity d and soil temperature T as shown in Table
2. The measured average thermal conductivity of the
soil was 0.551 W/m/K with a standard deviation of
0.067 W/m/K. This range of values is later on used to
design the geothermal system, for example the number Figure 3. Image of soil grains of Santa Juana (courtesy of
of absorbers and the heat pump capacity as a function Mauro Poblete from LGM, UCSC).
of the air temperatures encountered. No analysis of the
variation of λ nor the other geothermal properties have
been performed in this preliminary study. It is a matter absorbers or collectors as shown in Figure 4(a). In
of further studies the assessment of the effect of vari- addition, the land surface needed for the installation
ations due to for instance, saturation degree, humidity of absorbers is much smaller than for usual pipe type
and density on the soil geothermal properties. absorbers (e.g. series, parallel, slinky), being around
a seventh of the usual land surface. Figure 4(b) shows
that the width and depth of the trench excavated should
3.2 Heat absorbers
be 0.7 m and 1.5 m respectively. In cases where the
What makes this geothermal heating system techni- excavation sides are unstable, for safety reasons, a
cally and economically possible is the use of shallow shore system should be implemented.
639
Table 2. In situ measurements of soil thermal parameters
(cega.ing.uchile.cl and BBSolutions).
λ, W/m/K ρ, mK/W C, J/g/K d, mm2 /s T, ◦ C
0.577 173.4 – – 10.58

0.573 174.6 – – 10.78
0.531 188.3 1.541 0.345 10.98
0.453 220.9 1.447 0.313 11.24
0.610 164.1 – – 10.40
0.610 163.8 – – 10.54
0.500 200.2 1.474 0.339 10.35
0.519 172.8 1.643 0.316 10.80
0.675 148.1 – – 10.44
0.464 215.6 1.819 0.255 10.44
17.62 5.675 77.43 0.228 0.29*
*frozen soil.
Figure 5. a) Front view of the semidetached housing of

the Santa Juana project and b) excavation with geothermal
collectors at the back.
and filled with a mixture of water and glycol, which

is the heat transmitter and transports the energy from
the ground to the heat pump. For more details see the
Handbook (2014).
A semidetached housing of 50 m2 divided in two
floors without a proper thermal insulation was used
as the pilot project as shown in Figure 5(a). Addition-
ally, Figure 5(b) shows the excavation dug in Santa
Juana with 30 collectors, which corresponds to 3 kW
of power. Note that the land surface needed by the
collectors is 16 m2 , which is much less than other hor-
izontal type of absorbers such as pipes, which will need
around 140 m2 .
3.3 Geothermal heat pump

Figure 4. a) Collector model used (www.geocollect.de) and A geothermal heat pump, also called Ground Source
b) trench dimensions with collectors’ position.
Heat Pump GSHP, is a system that transfers heat from
the ground for heating as well as to the ground for
The backfilling of the trench should be carried out cooling and also can supply warm water to buildings.
carefully to not damage the collectors or the pipe con- Figure 6(a) shows a general diagram of a geothermal
nections. Each collector module is around 0.9 m long, heat pump. The cycle starts when ground heat is cap-
0.35 m high, resulting in approximately 0.63 m2 of tured by a buried absorption circuit as explained in the
flat interaction with the surrounding soil. This collec- previous section. Then the circulating fluid reaches
tor design allows an efficient absorption of the ground the heat pump where by means of applying high pres-
energy due to a much larger absorption surface than the sures in a compressor this fluid vaporises, increasing
flat 0.63 m2 . The absorbers are made of polypropylene significantly the temperature. This higher temperature
640
Figure 7. Mat radiation system installed below the floor and
ceiling (www.ewktec.com).
Figure 8. Infrared photography of inclined ceiling (courtesy

of Francecs Ferraro from UCSC).
Figure 6. a) General diagram of a heat pump and b) heat

pump installed in Santa Juana project (www.neura.de).
is transferred to another fluid inside the building dis-

tribution network or second internal closed cycle, for
example in wall radiators, underfloor or ceiling radi-
ators. To close the external cycle, the vapour flows to
a condenser where then passes through an expansion
valve that expands the fluid, sending it back cool to the
absorption circuit where the cycle starts again.
The horizontal closed ground loop arrangement Figure 9. Average maximum and minimum temperatures in
Santa Juana (http://www.worldweatheronline.com).
used in this project brings the fluid that captures the
geothermal energy from the ground to the electric
heat pump shown in Figure 6(b). This small 3.92 kW conductivity (seeTable 2). For these conditions the sys-
heat pump system makes possible that the circulat- tems works properly until soil reaches −13◦ C, below
ing water-glycol mix with a consumption of 0.97 kW that temperature the soil is unable to have enough time
of electricity can result in a COP = 4.0. This value to recover and transfer heat to the absorption system.
needs to be verified by measurements of heating load This was determined following recommendations by
divided by the sum of the power input to the compres- VDI 4640 (2010). However, further research is needed
sor, water-antifreeze solution circulating in the pump to assess the buried depth effect on the panel type
and condenser. absorber as a function of soil thermal properties and
During six months of monitoring a geothermal heat- number of collectors.
ing system with pipe absorbers, Inalli & Esen (2004)
found that the average COP value increased around
3.4 Radiation system
0.2 for 2 m depth buried pipe collectors respect to other
buried at 1 m. In this project 30 collectors were chosen A system of mat radiation under the floor and in the
for 1.5 m depth according to the available soil thermal ceiling has been adopted as shown in Figure 7. The
641
Figure 10. Typical output from monitoring system temperatures.
system is cheap and uses lower temperatures of radia- pilot project in Santa Juana aims to offer a solution for
tion, around 25◦ C. It makes the warmed fluid to flow not only a cheaper and efficient way of heating but also
in parallel small pipes. It is flexible and easy to install to reduce the extensive use of polluting burning wood
even in inclined surfaces as the ceiling for example risking the lives and health of people and saving extinc-
(see Figure 8). A more uniform heat radiation is gen- tion of local native forest. The geothermal properties of
erated, giving a better comfort when compared with the fine soil in Santa Juana are adequate for extracting
fixed wall radiators which irradiate higher tempera- heat for the averaged temperature conditions encoun-
ture heat, but tend to send it to the ceiling along the tered. Research is needed to study possible variations
walls. of these property values due to variations in saturation
Figure 8 shows an infrared photography of an degree, soil density and soil humidity. The combina-
inclined part of the ceiling. Note the lines showing tion of specific components and devices made possible
the presence of warm fluid at 25◦ C in the small pipes the set up of an efficient low enthalpy geothermal sys-
of the thin mat inside the inclined ceiling. tem for a small scale project. The Santa Juana project
opens the possibility for the development of similar
3.5 Preliminary results geothermal systems around the centre and south of
Figure 9 shows the air average maximum and lower Chile. However, it is suggested that like in this case,
temperatures in Santa Juana. It can be seen that the there should be always, at least at the beginning, a pilot
autumn and winter months between May and August project with enough time for monitoring to take into
are the ones when more heating is needed. Below account possible thermal ground differences as well as
zero temperatures are rare and summer months of environment and construction conditions.
December to February would need cooling.
Figure 10 shows temperature results obtained from
sensors installed in the geothermal system. The mon- ACKNOWLEDGEMENTS
itoring system has been installed to measure the
temperature of circulating water within the absorption The authors would like to acknowledge the funding
and heat pump piping as well as outside and indoor provided for this project by the German Federal Min-
temperatures. Preliminary results have shown that the istry for Economic Co-operation and Development
shallow heat extraction geothermal system can sustain (develoPPP.de) and the German Society for Interna-
the interchange circulating water temperature down to tional Collaboration (giz.de). The authors would also
−6◦ C at night to keep around 19◦ C indoors when the like to thank for the support to this project to Juan Pablo
outside temperature is around 10◦ C. Sanzana, Paulo Orostegui from OITEC and Mauri-
These are preliminary results and more extreme cio Poblete, Mauro Poblete and Francecs Ferraro from
temperature conditions are needed to assess the UCSC.
geothermal system performance. However, what is
clear from the data in Figure 10, it is that the system
recovery capacity is quite fast. REFERENCES
ASTM D 5334 2008. Standard test method for determination
4 FINAL COMMENTS of thermal conductivity of soil and soft rock by thermal
needle probe procedure. ASTM International, USA
A new low enthalpy geothermal heating system is pro- Ayala, J., Villalobos, F. &Alvarado, G. 2015. Travel time mea-
posed for Chilean small housing developments. This surements of shear waves in Bío Bío sand. Proceedings of
642
the Sixth International Symposium on Deformation Char- Jahnke, E. 2014. Evaluación económica de alternativas de
acteristics of Geomaterials, Buenos Aires, IOS Press, calefacción año 2014. http://www.territorioverde.cl/energia/
470-478 calefaccion.pdf
Celis, J.E., Morales, J.R., Zaror, C.A. & Inzunza, J.C. 2004. Omer, A.M. 2008. Ground-source heat pumps systems and
A study of the particulate matter PM10 composition in the applications. Renewable & Sustainable Energy Reviews
atmosphere of Chillán, Chile. Chemosphere 54, Issue 4, 12, 344–371
541–550 Schueftan, A. & González, A.D. 2015. Proposals to enhance
DECAGON 2015. KD2 PRO thermal properties analyser. thermal efficiency programs and air pollution control in
Operator’s manual, USA south-central Chile. Energy Policy 79, 48–57
DS 39 2014. Establece norma de emisión de material par- Schueftan, A. & González, A.D. 2013. Reduction of firewood
ticulado para los artefactos que combustionen o puedan consumption by households in south-central Chile associ-
combustionar leña y pellet de madera. Ministerio del ated with energy efficiency programs. Energy Policy 63,
Medio Ambiente, Gobierno de Chile 823–832
Handbook 2014. GeoCollect-Erdwärme-Absorber-System. Silva, A. & Arcos, D. 2011. Aplicación del programa AER-
http://www.architekten24.de/mediadb/news/18507/ MOD para modelar dispersión de PM10 emitido por
Handbuch-Erdwaerme-Absorber-System1.pdf equipos de calefacción a leña en la ciudad de Constitución.
Inalli, M. & Esen, H. Experimental thermal per- Obras y Proyectos 9, 4–10
formance evaluation of a horizontal ground-source VDI 4640 2010. Thermische Nutzung des Untergrundes –
heat pump system. Applied Thermal Engineering 24, Grundlagen, Genehmigungen, Umweltaspekte. Beuth
2219–2232 Verlag, Berlin
643
Determination of interface temperatures at scaled borehole heat exchanger

physical models using distributed temperature sensing
K.S.O. Seibertz & T. Vienken

Department of Monitoring and Exploration Technologies, UFZ – Helmholtz-Centre for Environmental Research,
Leipzig, Germany
H. Shao
Department of Environmental Informatics, UFZ – Helmholtz-Centre for Environmental Research,
Leipzig, Germany
P. Dietrich
Umwelt – und Ingenieursphysik, Eberhard-Karls University, Tübingen, Germany
Leipzig, Germany
F. Händel
Institut für Grundwasserwirtschaft, Technische Universität Dresden, Dresden, Germany
Leipzig, Germany
ABSTRACT: The ongoing energy transition in Germany has led to an increased use of the shallow subsurface
for heating, cooling, and energy storage. This approach represents a viable alternative to conventional, e.g. fossil
fuel derived, thermal energy provision. To ensure a sustainable geothermal usage and operation, especially for
large scale systems, various models can be used to predict induced temperature anomalies in the vicinity of the
Borehole Heat Exchanger (BHE) analytical and numerical. However, only a few physical models exist that allow
comparison of predicted with actually measured temperature data.
To close this knowledge gap, laboratory to medium scale experiments were undertaken to improve the under-
standing of the thermal behaviour within and at the BHE’s interface to the subsurface. To measure temperature
distribution that was generated by thermal exaggeration, fibre optic Distributed Temperature Sensing (DTS)
was used.
The possibilities offered by DTS, were used to measure spatial high-resolution BHE integral temperatures at
the BHE/subsurface-interface as well as temperature evolutions along the interface of inlet and outlet pipes to
grouting. Therefore, a 1:3 scaled physical BHE model wrapped with optical fibre was used to measure the BHE’s
integral temperature. A second installation consisting of a five meter long BHE (to scale) using DTS wrapped
around inlet and outlet pipes was used to measure the temperature evolution at the interface of the pipelines to
grouting.
This study provides continuously measured, high spatial resolution (between 0.03 to 0.1 m at laboratory and
medium scale respectively) temperature data along the BHE’s physical models, under different experimental
set-ups. Results allow us to get a thorough insight of the temperature evolution of BHEs.
1 INTRODUCTION a heat transfer medium. The space between pipes and

subsurface/aquifer, is commonly filled with a special
The demand on renewable or alternatives to conven- (thermo-)cement and called grouting. This is to ensure
tional, fossil energy resources in course of the energy an optimal coupling of the BHE to the subsurface as
transition led to increased usage of the subsurface as well as to protect the subsurface against heat transfer
source or for storage of thermal energy (Morrone et al. medium leakage (Self et al. 2013, Omer 2008).
2014, Blum et al. 2010). Beside the commonly used numeric modelling (Li
One of the more common systems for geother- et al. 2014, Kolditz et al. 2012, Trefry and Muffels
mal energy usage is the Borehole Heat Exchanger 2007) and field scale tests (Acuña and Palm 2013, Fujii
(BHE). BHEs are closed loop systems, circulated by et al. 2009) there are only a few approaches in doing
645
such research on BHEs at laboratory scale (Eslami- simulating the heat transfer medium, which in field
nejad and Bernier 2012, Beier et al. 2011). From the applications is often a Glycol. The pipes from the
experiences with field implementations of fibre opti- heat transfer medium reservoir to the BHE were in
cal Distributed Temperature Sensing (DTS) on BHEs both cases isolated to reduce heat loss. The columns
for, e.g. thermal response test (TRT) (Acuña and Palm were isolated, too. The DTS device N4386-A (Agilent
2013, Fujii et al. 2009), it was deducted, that DTS is Technologies 2007) allows temperature measurements
a suitable measurement technique for observing tem- down to a inherit spatial resolution of one meter. The
perature evolution at the outer boundary of BHEs as already mentioned wrapping was used to achieve the
well as from within a BHE. desired spatial resolutions. Temperature traces were
To contribute in laboratory studies on BHEs as well recorded every five minutes for each depth interval,
as testing the viability of DTS at scaled BHEs, two resulting in 66 and 100 temperatures every single time
physical models of BHEs were built and temperatures step, respectively.
measured a) at interface BHE/subsurface as well as Due to there was no interest in absolute tem-
b) at the interface of inlet and outlet pipe/grouting perature values offset calibration was not installed.
during hydration of the cement and thermal exaggera- Nevertheless, the DTS system was used in dual-ended
tion. Temperature measurements are provided by DTS. mode to avoid the need for attenuation calibration. As
DTS was hereby chosen because it allows continu- part of the temperature analysis the data was filtered
ous, spatial and temporal high resolution temperature using a moving average low pass filter to reduce high
measurements (van de Giesen et al. 2012, Selker frequency measurement noise.
et al. 2006) but is rarely tested on laboratory scale
experiments up to now.
2 METHOD The temperature changes from the water bath exper-

iment are presented in Fig 1a. The heating was done
The two presented physical models were built over a period of 168.5 h. After this period of time tem-
according to two German guidelines issued by the perature equilibrium was reached, and the experiment
Sächsisches Landesamt für Umwelt, Landwirtschaft was stopped intentionally. The temperature evolution
und Geologie and Niedersächsisches Umweltminis- at the interface of BHE and surrounding water shows
terium (Niedersächsisches Umweltministerium 2006, decreasing temperatures with depth. This leads to an
Sächsisches Landesamt für Umwelt, Landwirtschaft overall temperature increase of 13.3 K at measurement
und Geologie (LfULG) 2015). The first model, built interval 0.13 to 0.17 m at top of column, while the bot-
for temperature measurements at the BHE/subsurface tom of the column only shows an increase of 10.9 K.
interface was scaled 1:3 according to the physical The presented temperature values result from averaged
dimensions provided in the guidelines, having a length temperature values over the last measured 100 values
of two meter. Water-saturated gravel was used as from equilibrium state.
backfill between pipe and subsurface, which was sup- The according results from the sand-column exper-
ported by a plastic tube. The supporting tube was iment, presented in Fig. 1b, show a different tempera-
also the mounting point of the optical-fibre, which is ture evolution behaviour than expected from the water
needed for DTS measurements. Therefore, the fibre column experiment. The initial (averaged baseline)
is wrapped onto the supporting pipe allowing for temperature within the sand-filled cylinder started
integral, BHE surrounding temperature measurements at a linear temperature distribution, having a slight
with a vertical spatial resolution of 0.03 m. The sub- decrease from the top of the column of around 18◦ C
surface was simulated in two different ways: a) by a to the bottom of the column of around 17.5◦ C (com-
free water column and b) by water-saturated sand. pare to Fig. 1b). Even with more linearly distributed
The second model, built for grouting and starting temperatures the effect of decreased tempera-
pipe/grouting interface DTS measurements during tures towards the bottom of the column is present.After
cement hydration, and thermal exaggeration, was built 122.5 h of heating the maximum temperature change
using the physical dimensions of the mentioned guide- occurs at 1.4 m with a temperature change of 9.9 K.
lines. The model is five meter long (vertically oriented) The maximum temperature change at top of the col-
and has a slightly larger diameter then recommended. umn (still covered with sand) is 8.6 K. At the bottom
As grouting thermally enhanced cement for BHEs was of the column the maximum temperature change is
used (SCHWENK Spezialbaustoffe GmbH & Co. KG around 9 K. This leads to a curved temperature change
2014). This time the optical-fibre is mounted at the response as presented in Fig. 1b.
inlet and outlet pipes (via wrapping) allowing for a Both experiments show lower temperatures at the
temperature measurement every 0.1 m at the interface bottom of the column during thermal exaggeration.
between the two pipes and grouting inside the model This can easily be explained by the larger area of possi-
BHE. This BHE was built with the intend to later on ble heat loss at the BHE’s probe foot. Close to the probe
focus on studying thermal short-circuiting behaviour. foot the half space that allows temperature exchange
For both systems a water cycle was installed, allow- with the surrounding material is large in comparison to
ing for pumping water through the pipe system, more upper points within the BHE. This also explains,
646
why this temperature evolution is present at the top the experiment is situated in (comp. Fig. 2, isolation
of the column filled with water saturated sand (comp. was installed after cement hydration). The exothermic
Fig. 1). However the absence within the water bath hydration process, that lead to a rise in temperature
points towards a convection cell transporting heated in comparison to the ambient conditions, is detected
water upward within the free water column (comp. within the first days of hydration. Around the fourth
Fig. 1), which seems to be strongly inhibited within metre from the top, the peak temperature is reached
the second run. with a delay of around 24 h. This can be due to a
The results of the first two experimental series prove change in cements quality or the processing speed of
DTS is viable measuring integral temperatures around the cement as it is stated in Jäger (2013).
BHEs. From this result it was decided to upgrade To validate the proper operation of the heat
the experimental set-up by constructing the to scale, exchange medium circuit a test run was conducted
five metre long BHE. The DTS temperature measure- during a full isolation state of the BHE. This allowed,
ments were started during cement casting to allow also for analysing, if the overall behaviour of the to
an insight into the signal evolution during the hydra- scale BHE is comparable to the 1:3 scaled physi-
tion process. The temperature evolution after cement cal model. The heat source used during the test run
casting is strongly influenced by ambient temperature was the waste heat of the installed pump. During full
changes introduced due to air conditioning of the hall isolation this leads to a strong increase in tempera-
ture as well as thermal short circuiting (comp. Fig.
3), in the end temperatures were reaching close to
70◦ C. The test run was stopped after 180 h of system
up-time, because the water cycle started to eject water
at the connection between BHE pipes and pipes to the
reservoir. The resulting saturation of the stone-wool
at the lowest parts of the BHE reduced the isolation-
capability and therefore increased the cooling from
the fourth metre on, detected, by lower temperatures
(comp. Fig. 3 191 h).
In comparison to the results from the 1:3 scaled,
physical BHE model the temperature evolutions of the
five metre long BHE, especially close to the probe foot,
are similar. As one would expect the temperatures are
lowered for the reasons explained, and are comparable
to real BHE data, too (e.g. Acuña and Palm (2013);
Jäger (2013)). The to scale, five metre long, physi-
Figure 1. Relative temperature change from baseline mea- cal BHE model, hereby allows for a detailed insight
surement of 1:3 scaled BHE at equilibrium state: a) after into the temperature evolution at the pipes/groutings
168.5 h of heating (water subsurface) and b) after 122.5 h of interface providing information on the thermal short
heating (saturated sand subsurface), sediment surface equals
circuiting between the upward and the downward pipe.
water surface.
Figure 2. Temperature evolution after cement casting within the five metre long BHE at minus one, four and five (probe
foot) metre. The Black lines shows the ambient conditions measured at the halls floor.
647
Figure 3. Temperature evolution during the test run of the water cycle for different times at the interface of inlet and outlet
pipes and grouting.
Those BHE internal measurements are therefore con- environment as well as the lack of need for special
sidered as beneficial over the integral temperature permissions, e.g. for drilling. On the other hand Beier
measurement of the 1:3 scaled physical model, as et al. (2011) already showed how complex the task of
the interface temperature distribution did not allow providing a good physical simulation of the subsurface
for thermal short-circuit detection. In the presented for to scale physical BHE models is, what can limit the
experimental run from the five metre long BHE, the laboratory approach on BHEs.
observed thermal short circuiting between the upward The provoked and detected thermal short circuiting,
and downward pipes apparently lead to a complete of the to scale five meter long BHE during the test run
thermal short-cut, shown by equal temperatures at inlet will be object of additional studies on thermal short
and outlet pipes. The main reason for this is seen in circuiting of BHEs and be presented in the upcom-
the reduced thermal dissipation from the BHE due to ing special issue of Environmental Earth Sciences on
being isolated instead of being connected to the sub- ANGUS+.
surface. However another possible explanation can be,
that there is a difference in temperature between inlet
and outlet pipes, but this is lower then the DTS’s tem- ACKNOWLEDGEMENT
perature resolution of 0.1◦ C and therefore could not
be detected. This study was funded by German Federal Envi-
ronmental Foundation – DBU as part of their PhD
scholarship programme as well as by the joint research
4 CONCLUSIONS AND OUTLOOK
project ANGUS + (Bundesministerium für Bildung
und Forschung, Förderkennzeichen 03EK3022B). We
The results of a laboratory to medium scaled exper-
thank Manuel Kreck for helping us to carry out our
imental series on physical models of borehole heat
work in this investigation.
exchangers (BHE) were presented. Therefore a 1:3
scaled and a five meter long to scale physical model of
a BHE were build. The physical models were success- REFERENCES
fully used for measuring BHE/subsurface interfaces,
integral temperatures during thermal exaggeration as Acuña, J. & B. Palm (2013). Distributed thermal response
well as for measuring the temperature evolution along tests on pipe-in-pipe borehole heat exchangers. Applied
the interface of inlet and outlet pipe/grouting during Energy 109(0), 312–320.
hydration and induced thermal exaggeration. Agilent Technologies (2007). Agilent N4385A/N4386A
With this experiment it was possible to show that Distributed Temperature System User’s Guide.
laboratory experiments with BHEs can be done on Beier, R. A., M. D. Smith, & J. D. Spitler (2011). Reference
data sets for vertical borehole ground heat exchanger mod-
multiple scales while at the same time having unrivaled els and thermal response test analysis. Geothermics 40(1),
spatial resolution in comparison to man power and time 79–85.
need to install the sensors and ensure their placement Blum, P., G. Campillo, W. Münch, & T. Kölbel (2010). CO2
by the use of DTS. Hereby one of the main benefits savings of ground source heat pump systems – A regional
of a laboratory experiment is the controllability of the analysis. Renewable Energy 35(1), 122–127.
648
Eslami-nejad, P. & M. Bernier (2012). Freezing of geother- Sächsisches Landesamt für Umwelt, Landwirtschaft und
mal borehole surroundings: A numerical and experimen- Geologie (LfULG) (2015). Erdwärmesonden Informa-
tal assessment with applications. Applied Energy 98, tionsbroschüre zur Nutzung oberflächennaher
333–345. Geothermie.
Fujii, H., H. Okubo, K. Nishi, R. Itoi, K. Ohyama, & SCHWENK Spezialbaustoffe GmbH & Co. KG (2014,
K. Shibata (2009). An improved thermal response test 12). Füllbinder® EWM Hydraulischer Spezialmörtel für
for U-tube ground heat exchanger based on optical fiber Erdwärmesonden.
thermometers. Geothermics 38(4), 399–406. Self, S. J., B. V. Reddy, & M. A. Rosen (2013). Geothermal
Jäger, C. (2013). Effizienzsteigerung und Qualitätssicherung heat pump systems: Status review and comparison with
von Erdwärmesonden. bbr-Sonderheft Geothermie. other heating options. Applied Energy 101, 341–348. Sus-
Kolditz, O., U.-J. Görke, H. Shao, & W. Wang (Eds.) tainable Development of Energy, Water and Environment
(2012). Thermo-Hydro-Mechanical-Chemical Processes Systems.
in Porous Media (1 ed.), Volume 86 of Lecture Notes in Selker, J., L. Thévanez, H. Huwald, A. Mallet, W. Luxemburg,
Computational Science and Engineering. N. Van de Giesen, M. Stejskal, J. Zeman, M. Westhoff,
Li, Y., J. Mao, S. Geng, X. Han, & H. Zhang (2014). Evalua- & M. Parlagne (2006). Distributed fiber-optic temper-
tion of thermal short-circuiting and influence on thermal ature sensing for hydrologic systems. Water Resources
response test for borehole heat exchanger. Geother- Research 42.
mics 50, 136–147. Trefry, M. G. & C. Muffels (2007). FEFLOW: A Finite-
Morrone, B., G. Coppola, & V. Raucci (2014). Energy and Element Ground Water Flow and Transport Modeling
economic savings using geothermal heat pumps in differ- Tool. Ground Water 45(5), 525–528.
ent climates. Energy Conversion and Management 88(0), van de Giesen, N., S. C. Steele-Dunne, J. Jansen, O. Hoes,
189–198. M. B. Hausner, S. Tyler, & J. Selker (2012). Double Ended
Niedersächsisches Umweltministerium (2006, 12). Leitfaden Calibration of Fiber Optic Raman Spectra Distributed
Erdwärmenutzung in Niedersachsen. Temperature Sensing Data. Sensors 12(5), 5471–5485.
Omer, A. M. (2008). Ground-source heat pumps systems
and applications. Renewable and Sustainable Energy
Reviews 12(2), 344–371.
649
Underground nuclear power plants
P. Duffaut
French Committee on Rock Mechanics, Paris, France
ABSTRACT: Opening of the minisymposium on Underground Nuclear Power Plants, second only conference
on this theme after the Hannover one in March 1981. Early plants in Siberia, Scandinavia, France and Switzerland
were presented there as well as a summary of many projects studied in the USA and Europe. The author had
begun working long before with EDF, the French Electricity Authority when many hydro power plants were built
into underground rock caverns. The first nuclear accidents urged him to propose underground siting of reactors
be studied again in order to escape any radionuclides dissemination.
1 INTRODUCTION
I am glad to welcome you all at this minisympo-

sium on UNPP, Underground Nuclear Power Plants.
To my knowledge only one international meeting has
yet focused before on UNPP, in 1981 at Hannover,
Germany, held at the headquarters of BGR, the Geo-
logical Survey of Germany. I had the chance to attend
it (Duffaut, 1981, Bender, 1981). Here, the city Kiel
is famous in Europe diplomatic history thanks to the
“entrevue de Kiel” (Kiel interview, or encounter) when
Tsar Alexander met Emperor Guillaume II about one
century earlier, on June 8, 1892. A spectacular detail
was the tsar delivered a toast in French, which was
the diplomatic language at the time. Today the world
scientific language is English, I’ll try to use it for you.
To begin, I must tell you how and why I engaged
on lobbying for underground siting of nuclear power
plants, or more realistically of sensitive parts of
them, beginning with nuclear reactors. Graduated
mining engineer in 1948, I joined EDF, the French
national electricity company, at the time of hydro
plants that were made of dams, tunnels and sometimes
underground caverns hosting turbines and alternators. Figure 1. Perspective of the Chooz A caverns, close to
Underground siting was chosen for a number of rea- Meuse river, a few km downstream of the border of Belgium;
sons, either lack of space or the more, for pump and operated 1968–1991; the reactor is in the cavern at left, the
storage plants, position below the lower reservoir level pool and auxiliaries are at right; the generating machines are
in a surface building on the river side; the Chooz B plant has
to escape turbine cavitation.
been built later on the other side, inside a meander.
To a young mining engineer, underground construc-
tion was something fully normal.
I was glad when the first PWR reactor was placed by
EDF into purposely dug rock caverns along the river Among reasons of its underground location, EDF
Meuse, rather close to the Belgium border (Fig. 1), did not rely on the vertical tendons of the prestressed
under the joint venture SENA (Société Electronu- reactor buildings, menaced by segregation hazards
cléaire des Ardennes), with Electrabel, a company inside the grout. An extensive program of tests
from Belgium. I remember the visit of the caverns at (followed by the author) was launched at the Aix-en-
the end of their excavation, among many French and Provence EDF laboratory to study better grout matter
foreign caverns for hydro plants. and grouting procedure, which were used later in all
653
PWR reactors built and operated in France. (Chooz A works. Really many such heavy constructions as well
has been the first PWR in France and the prototype as all underground structures meet geotechnical prob-
of the whole nuclear park after the first UNGG plants; lems of soil or rock mechanics. So I suggested a place
its commercial production began April 1967 and lasted be offered to underground nuclear plants. That is why
to 1991). we are here today! I am just sorry to be myself the more
Unfortunately this first UNPP was to be also the last a rock mechanics engineer, with rather few personal
one, as well in France as abroad …; I had no longer works in soil mechanics!
any hope of a brilliant career in designing, building
and operating underground caverns for nuclear plants;
so I left EDF to join BRGM, the French Geological
2 GEOTECHNICAL PROBLEMS OF ENERGY
Survey, from which I had the chance to attend the
PRODUCING PLANTS
only international conference ever organized on UNPP,
held at Hannover BGR headquarters (the German Geo-
Early coal fired plants long met only classical founda-
logical Survey). So I have been aware of previous
tion problems, easily solved through improved tech-
underground plants operated in Scandinavia, France
nologies. When their power increased from 250 to
and Switzerland, but not of the Soviet ones which
600 MW, the dimensions and the loads on the ground
were strictly kept secret: We know now that the Soviet
asked for novel studies and novel solutions, just like the
Union played a pioneering role in the UNPP area with
very heavy reactors of the French UNGG model (natu-
the construction and operation of the 3 Zheleznogorsk
ral uranium fuel, graphite moderator, gas heat transfer
plants in Siberia since 1951. The French plant was
fluid), namely stiffer reinforced concrete piles, built
only presented by Pinto, from Swiss Atom research
like short parts of slurry walls; instead of the clas-
centre Würenlingen, as no French representative did
sical steel piles. Of course most of nuclear sites in
mention it (when it was the only commercial one
France called for soil mechanics problems, say dif-
in operation, and its electric output only was over-
ferent settlements along close buildings more or less
passed by secret Soviet ones. Pinto also summarized
heavy at Dampierre; the more at Blayais on estuary
the many studies yet performed in California, Canada,
Garonne and Dordogne rivers: the site was free of any
Scandinavia, Switzerland and Japan. Let’s remem-
occupation as even the lightest shepherd hut could not
ber Scandinavia was yet famous for mastering the
stand: all peat like soil had to be removed and replaced
use of explosives and developing so many uses of
by compacted sand inside slurry walls (a spectacular
underground space.
geotechnical solution).
Later I took interest in works and ideas of French
architect Edouard Utudjian who had launched the
concept of underground city planning (in French,
Urbanisme souterrain). So I proposed to build nuclear 3 THE THREE MODELS OF UNPP, MOUND,
power plants underground, first in a conference in PIT, TRUE ROCK CAVERN
Roma (Fig. 2 and Duffaut, Marin, 1986) and later
in a 2007 ACUUS conference in Athens (Duffaut, – In the mound, or tumulus, model, the plant is built
2007). on the surface and covered under a rubble mound
When I discovered ICEGT, this conference of ISS- (a model well fitted to flat plains of Germany, which
MGE TC 308, I was surprised power plants did not harbor most of industry and cities).
appear on the list of topics, nor any underground – It is interesting to lower the reactor inside a pit, to
limit the height of the mound. This model calls for
pit support and tightness, thanks to slurry walls all
around.
– No German author did discuss the rock cav-
ern model in any detail but cavern projects had
been presented by California, Japan, and the more
Ontario Hydro, Canada (Oberth, Lee, 1979): I took
part the same year in a study tour of the Toronto
Geological conference guided by C. F. Lee. The
most conspicuous feature was the rock, granite, at
400 m depth which was supposed to be the sole
barrier, without any concrete or steel one.
4 GEOTECHNICAL RISKS OF UNPP SITED

IN ROCK CAVERNS
Figure 2. A sound rock hill by the sea could easily host an
underground plant likely to power a big city (any semblance Theory and practice of underground works in rock
with a known site would only be a coincidence!) (Duffaut, may be regarded as parts of Geotechnics. Most of
Marin, 1986). authors have asked for rather high quality rock for
654
locating UNPP caverns (RQD over 75 or 80). How- was the negative sculpture proposed by artist Chillida:
ever in Sweden the general “Rib in Rock” method had close to a cube, 40–50 m side, with plane faces and
been devised in the case of a lower quality rock mass: sharp angles, with two vertical pits “toward sun and
early small galleries are filled with concrete all around moon” (Fig. 4).
the future cavern, acting like the ribs of a skeleton;
then the cavern is dug inside a reinforced rock mass
(Fig. 3, from the Mission to the USA of the Swedish 5 PAPERS PRESENTED
Underground Industry, 1976).
Dealing with rock caverns, the main geotechnical In addition to this introduction paper, two more have
risk is any lack of stability, which may be due to: first been sent:
a wrong outline of cavity inducing excess stresses, – Why do nuclear engineers refuse the concept of
either compressive at sharp angles or tensile in flat underground nuclear power plants? by De Bruyn D.
walls, second to plane discontinuities dipping towards – Underground Nuclear Power Plants: A Realistic
the cavity and permitting downward movement of so Option? by Varun, Fairhurst C, Pierce M., Lagger
called “mega blocks”. Many caverns have experienced
cave-ins along high vertical walls when natural sur- The first one deals with the commission on UNPP
faces of discontinuity dip toward the hole; conversely launched by ISRM on suggestion by Japanese prof.
very few cases of ceiling failure have been reported: Shun Sakurai and myself; the second mainly focuses
arch effect naturally operates. on advantages of an underground location mainly face
Unfortunately few civil engineers have acquired a to seismic hazards.
minimum of “underground culture” and can rely on
the rock mass as a rather benevolent matter: the theory 6 CONCLUSIONS
of the hole highly differs from the classic “strength
of materials” used when designing bridges and monu- Rock engineers cannot go farther. Whatever could be
ments: there are no slender members in underground discussed inside this meeting, nuclear engineers have
engineering and “the task of the engineer is to help to study in detail the underground option of future
the rock mass to form natural arches” (Dodds, in the nuclear reactors, just as it has been decided for nuclear
Swedish Mission, 1976). A counter example of cavity waste; as has been proposed in the Supergrid project
(Fig. 5, Myers, Elkins, 2004). No surface transport
will be needed if the used fuel storage or even treat-
ment is underground at short distance of the nuclear
reactor.
Figure 3. Rib in Rock method for supporting a big Figure 4. Artist view of the negative sculpture designed by
cavern before excavation (Swedish Mission to the USA of Chillida for Tindaya mountain, Canary Island, Fuerteventura,
the Underground Industry, 1976). Spain (two people on the floor give a scale).
655
Figure 5. Sketch of an underground nuclear park, after Myers, Elkins, 2004.
REFERENCES Duffaut P., 2007, To be sure, nuclear power plants must be

underground, ACUUS conf. Athens
Bender F. 1981, Proc. Intern. Conference on Underground Duffaut P., Marin G., 1986, Quels ouvrages souterrains
Siting of Nuclear Power Plants – Unterirdische Bauweise pour le XXI siècle? Cong. Grandi Opere Sotterranee,
von Kernkraftwerken, Hannover (Germany), (partly in Florence; in French Tunnelling & Underground Space
German). Technology. 2 2, pp. 155–164, Gallerie, 25, oct 1987,
Collective, 1976, Mission to the USA of the Swedish 13–22 (in Italian).
Underground Industry (private booklet). Myers W., Elkins N., 2004, Siting Nuclear Power Plants
De Bruyn D., Sakurai S., Myers W., Duffaut P., 2016, Under- Underground: Old Idea, New Circumstances, Nuclear
ground N uclear Power Plants: an old idea that should News, v. 47, no. 3, pp. 33–38.
be reconsidered? ICAPP Conf., San Francisco. Oberth, R.C., Lee C.F., (1979) Underground siting of
Duffaut P., 1981, Compte rendu de la réunion de Hanovre CANDU Power Stations, Journal of the American Under-
sur l’implantation souterraine des centrales nucléaires, ground Space Association, Pergamon Press, vol. 4, no. 1,
BRGM report 81 SGN 476 GEG, July, 15 p. (in French). pp. 17–27.
656
Why do nuclear engineers refuse to consider the concept

of underground power plants?
D. De Bruyn & P. Duffaut

ISRM Commission on Underground Nuclear Power Plants
ABSTRACT: Siting a nuclear power plant below the surface has been studied by many authors, but has been
realized only a few times and not recently. Advantages of underground siting are well recognized not only by
the geotechnical community, but also by the nuclear community as well. Every time a major catastrophic event
occurs (TMI, Chernobyl, Fukushima) the question arises again “why not underground”. Still, classical above-
ground layouts remain the standard when dealing with the siting of a nuclear power plant. The ground- and
rock-mechanics communities are well aware of the numerous advantages of underground siting of large plants,
nuclear or not. The nuclear engineers on the other hand still seem to be afraid of following this new path for them.
The present paper summarizes the major concerns of the nuclear community and concludes with an approach
for geotechnicians to convince their nuclear colleagues.
1 INTRODUCTION had been underground, they would not have sustained

so much damage from the tsunami and would not have
The total or partial underground siting of nuclear developed into such a disaster. In fact, they may not
reactors (UNPP for Underground Nuclear Power have been damaged at all (…).
Plants) has been investigated in the past, although At the same time, underground construction tech-
few examples have been put into practice. In Western niques have progressed significantly since the 1970s,
Europe there are only four such instances: the Agesta information is now available on the real cost of disman-
power plant in Sweden (Abadie-Maumert 1965), the tling existing power plants, and building smaller power
French-Belgian power plant at Chooz in the Ardennes plants is no longer inconceivable. Underground Small
(Erkes & Grau 1965), the Halden research reactor Modular Reactors (SMR) generating up to 300 MW
in Norway and an experimental reactor in Lucens electricity could be sited far closer to the residential
(between Lausanne and Bern, Scott 1975) in Switzer- areas which they need to supply with power, at a per-
land, dismantled in the early 1980s. The Swiss reactor fectly acceptable cost (The New Scientist 2001, The
is of considerable interest today, even if details about Energy Collective 2011, CNN 2012). Some recently
it were not all that widely known at the time. On formulated ideas are presented in Section 3.
21 January 1969, the plant witnessed a partial fusion The International Society for Rock Mechanics
of the reactor core, but this accident did not cause any (ISRM) took the initiative to set up a dedicated
radiation harm to the operational staff and radioactiv- commission to study the feasibility of UNPPs. The
ity was confined within the rock cavity. We also know commission was set up in 2013 (Section 4). A progress
now that the Soviet Union played a pioneering role in report has been presented at the ISRM conference in
the UNPP area with the operation of the Zheleznogorsk 2015, where the commission mandate has been length-
plant in Siberia in the utmost secrecy since 1954. In ened for four more years. Section 5 presents a status
Section 2, a non-exhaustive summary of the studies of our activities so far. Finally, Section 6 deals with
and publications during the 1970s and 1980s is drawn. the concerns of the nuclear engineering world against
It took several catastrophes to put the underground UNPPs and how we should respond.
option firmly back into the picture, in particular the ter-
rorist attack on the WTC in New York in 2001 and the
natural disaster of Fukushima in 2011. Several deci- 2 THE 1970S AND 1980S
sion makers have been asking themselves whether the
savings made with the aboveground construction of a Studies on the feasibility of underground siting of
large power plant are not neutralized by such a disaster. nuclear power plants were undertaken in several coun-
Gunard Nord, Atlas Copco specialist in underground tries for various reasons:
construction, put it this way in Mining & Construction • In Canada (Lee et al 1979, Oberth 1981), the decen-
online (ATLAS COPCO 2011) on 19 July 2011: (…) tralization of power generation was investigated
I am convinced that if the Fukushima plants in Japan with the construction of power stations closer to
657
cities; the Canadian concept consists of reactor cav- before 1981, in particular Allensworth 1977, whereas
erns at a depth of about 400 metres, excavated Lee et al 1979 recall that by decentralizing the pro-
in competent granitic gneiss, with the powerhouse duction of electricity (more ‘small’ production units),
sited in a below-grade, near surface excavation at a the electricity transport cost can be reduced, and
depth of about 50 metres (Figure 1); that such units are easier to dismantle than their
• In Germany (Bachus 1981, Buchhardt 1981, Kröger aboveground counterparts, through the use of in-place
et al 1978), the main concern is a better protec- decommissioning.
tion of the public against the effects of external In conclusion, the specialists at the time considered
factors (aircraft impacts, earthquakes) and internal that although underground siting had its advantages,
causes (leaks, explosions); the concept in loose soil the extra cost was too high. The aboveground power
is close to the surface, partly (Figure 2) or entirely stations were safe enough to convince the competent
buried; authorities that this extra cost was not justified. And
• In Norway (Løken 1979), studies were prompted so the ‘underground’ solution was not repeated after
by concerns of protection against acts of war, with the few examples cited in Section 1.
power stations being required to carry on operating
during conflicts;
• In Japan (Ichikawa 1981), the high population
3 MORE RECENT STUDIES
density (only 30% of the total surface area is
inhabitable) calls for a rational use of surface;
•
It was not until the first decade of the new millennium
In the United States (Allensworth et al 1977,
that there was a resurgence of interest in underground
Finlayson 1981), focus is on an improved protec-
power stations in the USA. The terrorist attack in 2001
tion of the public against accidental releases of
certainly had something to do with it, yet there were
radioactive material.
other contributing factors too (Myers & Elkins 2004).
A comparison between conventional construction Primary was the potential to collocate reactors, stor-
and underground construction was made in several age and disposal facilities at the same underground
articles (Lee et al 1979, Bachus 1981, Kröger et al site, thereby creating an “Underground Nuclear
1978, Løken 1979, Allensworth et al 1977). All authors Park” (UNP).
agree on several points: It became clear that a key factor would be the choice
•
of an appropriate geological site that at the same time
Underground structures offer greater resistance to
would permit:
external factors (aircraft impacts, earthquakes –
seismic being a surface wave, seismic shaking is • (Relatively) easy construction;
lower at depth –, acts of war, terrorist acts) than • Long-term confinement and stability qualities;
conventional structures, • Compatibility with underground storage and
• but at the same time the extra building cost (espe- disposal.
cially due to the longer construction time involved)
may be considerable. Although granite and some sedimentary rock types
could certainly be suitable, the original concept for the
The technical feasibility of an underground solution UNP was introduced (Myers & Elkins 2004) based on
has been called into question in several publications bedded salt as the host rock as shown in Figure 3.
Figure 1. Canadian concept in a hard rock environment (taken from Oberth 1981, page 11) (note that the ground level is
arbitrarily set at +1,000 m).
658
The authors list the following advantages: total electricity production cost. This is a significant
factor. It should also be noted that this study was for
• Reduced decommissioning costs (since part of the
a UNP sited in granite, and was based on prior work
installation remains in place);
by the authors that examined use of a tunnel boring
• Reduced spent fuel transport costs as a result of the
machine for excavating the underground facilities.
combined site (shorter distances);
Three years later, the same authors focused on the
• Reduced excavation costs for salt compared with
lessons learnt from the Fukushima disaster (Kunze
hard rock such as granite;
et al 2012). They conclude that an underground site
• Reduced building costs since the rock serves as
could have attenuated, or even altogether prevented,
confinement structure;
the dramatic consequences of that accident.
• Reduced cost of surveillance and operational safety;
According to the recent studies, we may conclude
• Potential for greater public acceptance.
that the underground siting of a nuclear power plant is
The UNP concept has been studied (Giraud et al not only technically feasible as well as a safer option,
2009) in terms of cost analysis and of safety, with the as this was already known since the 1980s. Provided
explicit title Cost advantages of large underground that various parameters are judiciously chosen, such
nuclear parks.The authors conclude (…) the combined as reactor size and geology, it also has, and this is new,
total of possible cost savings using the underground the potential to not be more expensive, on the contrary,
nuclear park concept can be as much as (…) 10% of the than the surface siting.
Figure 2. German concept in loose soil (adapted from Bachus 1981, page 49).
Figure 3. Concept for underground collocated reactors, storage and waste disposal facilities (Myers & Elkins 2004).
659
4 THE ISRM COMMISSION chamber and reduces its cost. This is valid for both
small UNPPs and large UNPPs. Finally, an area where
If the underground option represents added value for improvement has still to come is the collocation of an
the whole planet, it is now seen as virtually indis- UNPP and its long-term storage and disposal facility:
pensable for Japan since the Fukushima disaster. All not only access shafts and transport tunnels would be
NPPs have been shut down immediately, but renew- in common, but large savings can also be expected in
able energy solely cannot provide enough power, so security and nuclear waste transportation. Decommis-
thermal plants have been put in full operation, with as sioning costs would also be strongly reduced through
consequence not only an increase in cost, but also an in-place entombment.
increase in CO2 emissions.
If new NPPs are to be built, the Japanese authori-
ties need to convince public opinion as a whole that 6 DEALING WITH NUCLEAR ENGINEERS
the safety of these new plants will be guaranteed, even
in the case of highly improbable scenarios (such as a As summarized in the previous sections, underground
combination of severe earthquake and large tsunami) siting has many advantages compared to surface siting.
or events with a difficult-to-quantify degree of proba- So why do we see so few practical realizations? The
bility (such as a terrorist attack) and that the approach answer is not straightforward.
taken in the sizing of those plants will truly have taken First, the additional safety level is well admitted by
into account the lessons learnt from the Fukushima the nuclear community, but considered as not strictly
disaster. compulsory, as surface siting was sufficiently safe.
A request to set up a dedicated commission on Then come the concern of the cost. Many people, and
the underground siting of nuclear reactors was sub- most of the nuclear engineers, believe still today that
mitted to the ISRM, on the personal initiative of underground construction implies additional costs.
Prof. Sakurai, former ISRM vice-president. The combination of those two factors is quite
The purpose of that commission is to discuss the obvious: underground siting being more expensive
feasibility of such sitings and to formulate recommen- for supplemental, but not compulsory safety, is not
dations for their implementation. The rock mechanics interesting for industrial purposes.
community has wide experience in the excavation It is not the purpose of the geotechnical community
of large underground caverns, such as underground to execute detailed nuclear engineering, but we may
hydro-power plants. advance new elements on those topics, as introduced in
The ISRM approved that proposal at its annual the next subsections, in order to restart the discussion.
meeting in April 2013, and the first plenary meeting
was held at Wroclaw in Poland in September 2013
during the European conference of ISRM. Intermedi- 6.1 Additional, but not compulsory safety?
ate results have been gathered in a report presented Most of today’s concepts are still derived from GEN
at the ISRM 2015 conference, where our mandate has II concepts that date from the years 1970s. Safety
been lengthened; a short summary is given in Section 5 requirements were then not so difficult to meet as
below. today, among others in terms of earthquake, aircraft
impacts, acts of war or terrorist attacks. Should a
5 STATUS OF ACTIVITIES SO FAR surface NPP be designed today, its functional and tech-
nical specifications would be much more challenging,
An extended summary on the ISRM commission thus more expensive.
activities on UNPPs has been recently published On the other hand, underground siting, by nature,
(De Bruyn et al 2016) and the main points are given scores very well against all those increased require-
below. All commission members have agreed that the ments.
technical feasibility of UNPP is real, provided that cer- Our message should be: when considering today’s
tain technical criteria are satisfied, like geology, topog- safety requirements, do not exclude directly under-
raphy, location, reactor type and size, configuration ground siting.
and cooling system.
Surface siting is facing new challenges in terms of
6.2 Additional costs, really?
low-probability/high-consequence accidents, attacks
and extreme natural hazards. Interest is growing Again, most of the calculations for the whole costs
in SMRs of advanced designs. SMRs are in fact of a surface siting, including construction, operation
strong candidates for UNPPs. Modern excavation and dismantling, are still derived from those GEN II
techniques allow now to create cavities large enough concepts from the years 1970s. Underground siting is
to host an NPP of GEN III size (Giraud et al 2009), not merely “take the surface design and put it, as such,
provided that the host rock has sufficiently competent just below ground”.
characteristics. About the dismantling part of the total costs, it
It should also be kept in mind that underground should be recalled that it is a rough estimate, in per-
siting provides already a very good containment and centage of the construction costs. Several dismantling
shielding, which facilitates the design of the reactor pilot projects are now available with real costs and not
660
estimates. An update of the calculations would be an Bachus, K.P. 1981. Das Studienprojekt zur unterirdischen
interesting exercise. Bauweise von Kernkraftwerken in der Bundesrepublik
Finally it should be emphasized that underground Deutschland, Proc. Int. Conference on Underground Sit-
siting does not require a priori a full dismantling, as ing of Nuclear Power Plants – Unterirdische Bauweise
von Kernkraftwerken, Hanover (Germany), edited by
already mentioned in Section 5. F. Bender, p. 45–61 (in German).
Our message should be: take into account the costs Buchhardt, F. 1981. Pro and con decision criteria to under-
as estimated today and not forty years ago. ground power plants, Proc. 6th Int. Conference on Struc-
tural Mechanics in Reactor Technology, Paris, Volume J,
6.3 Underground siting is different Paper J 1/6.
CNN 2012. U.S. backs project for small nuclear reactors,
The host rock capabilities have to be taken into http://edition.cnn.com/2012/11/21/us/small-nukes/index.
account, in terms of structural elements, confinement html.
and shielding. In an appropriate host rock, thousands De Bruyn et al. 2016. Underground nuclear power plants:
of cubic meters of reinforced concrete are simply not an old idea that should be reconsidered? Accepted for
publication, Proc. Int. Congress on Advances in Nuclear
required. The possibility to combine power plant and Power Plants (ICAPP16), San Francisco, California, Paper
disposal site, with major savings in transport costs, 16170.
should be seriously investigated. Erkes, P. & Grau, P. 1965. Aspects particuliers de la cen-
Our message should be: a complete new approach trale nucléaire des Ardennes, Energie Nucléaire, 7, 189
is necessary to really fully use the underground’s (in French).
potential. Finlayson, F.C. 1981. The feasibility and effectiveness
of underground nuclear power plants – A review of
the California Energy Commission’s study, Proc. Int.
6.4 Underground siting, but not everywhere Conference on Underground Siting of Nuclear Power
This is another fear for the nuclear community: “if one Plants – Unterirdische Bauweise von Kernkraftwerken,
Hanover (Germany), edited by F. Bender, p. 19–33.
UNPP is built, then the safety authorities will make it Giraud, K. et al. 2009. Cost advantages of large Underground
compulsory every time and everywhere”. Nuclear Parks, Proc. ICONE17 Conference, Paper 75750,
Here our job as geotechnicians is to provide a list Brussels, Belgium.
of locations where host rock characteristics are suf- Ichikawa, Y. 1981. Activities in Japan on underground
ficiently suitable to make the UNPP economically siting of nuclear power plants, Proc. Int. Conference on
acceptable. Underground Siting of Nuclear Power Plants – Unterirdis-
Our message should be: Not necessarily every- che Bauweise von Kernkraftwerken, Hanover (Germany),
where, but well where it is economically acceptable. edited by F. Bender, p. 35–44.
Kröger, W. et al 1978. Beurteilung der unterirdischen
Errichtung von Kernkraftwerken im Boden in einer offe-
7 CONCLUSIONS nen Baugrube. Abschlussbericht über eine Studie für den
Bundesminister des Innern. BMI – Kennzeichen SR 44,
The advantages of underground siting are known and KFA Jülich, 194 p. (in German).
even recognized by the nuclear community. However, Kunze, J.F. et al. 2012. Underground siting of Nuclear
Power Plants – Insights from Fukushima, Proc. ICONE20
practical realization has not been largely performed
Conference, Paper 54900, Anaheim, CA, USA.
due to a persistent feeling of “underground means Lee, C.F. et al. 1979. Some geotechnical and planning con-
more expensive”. siderations in underground siting of nuclear power plant,
As geotechnicians, we should convince the nuclear Proc. Int Conference Rapid Excavation and Tunneling
community that figures from the past should be Conference (RETC), 1, p. 1386–1408.
revised. First for surface siting, which will be more Løken, P.C. 1979. Rock cavity construction of a nuclear power
and more costly due to increasing safety requirements. plant – A case study, Nuclear Technology, 42(2), 155.
Then also for underground siting, which might be Myers, C.W. & Elkins, N. 2004. Siting nuclear power
even economically more favourable, when properly plants underground: Old idea, new circumstances, Nuclear
News, December 2004, p. 33–38.
designed at appropriate locations. Finally, it should be
Oberth, R.C. 1981. The Canadian study of under-
recalled that for some countries, underground siting ground nuclear power plants, Proc. Int. Conference on
might be the only possible alternative for building new Underground Siting of Nuclear Power Plants – Unterirdis-
plants. che Bauweise von Kernkraftwerken, Hanover (Germany),
edited by F. Bender, p. 7–17.
REFERENCES Scott, F.M. 1975. Locating nuclear power plants underground,
Environmental Letters, 9(4), 333.
Abadie-Maumert, F.A. 1965. La centrale thermo-électrique The New Scientist 2011. Nuclear power to go underground,
nucléaire suédoise d’Agesta, Energie Nucléaire, 7, 321 http://www.newscientist.com/blogs/onepercent/2011/03/
(in French). nuclear-power-to-go-undergroun.html.
Allensworth, J.A. et al 1977. Underground siting of nuclear The Energy Collective 2011. Underground Reactor advan-
power plants: potential benefits and penalties, Report of tages, http://theenergycollective.com/charlesbarton/66055/
Sandia Laboratories for NRC, SAND 76-0412, NUREG- underground-reactor-advantages.
0255, 256 p.
ATLAS COPCO Web site 2011. http://viewer.atlascopco.com/
MC_international_2011-2/#page/14.
661
Underground nuclear power plants: A realistic option?
Varun, Charles Fairhurst & Matthew Pierce

Itasca Consulting Group, Minneapolis, USA
Hannes Lagger
ARUP, London, England
Pierre Duffaut
French Committee on Rock Mechanics, France
ABSTRACT: Of the technological options now available to reduce global dependence on fossil fuels, none has
been more widely applied than nuclear energy. Considerable development in nuclear technology has taken place
over the years to develop Inherently Safe Nuclear Reactors, but public opposition to expansion of nuclear power
has hardened in the light of Fukushima. One option that has long been advocated, especially for seismically
active regions, is to construct the nuclear power plant underground – at relatively shallow depth (∼50 −150 m)
in rock. Underground nuclear plants have been operated at various locations around the world. The development
of small, modular nuclear reactors has provided additional flexibility in the design of underground nuclear power
plants. A prominent concern with respect to nuclear power plants, especially since Fukushima, is the vulnerability
to damage when subject to a severe earthquake. There is considerable ‘anecdotal’ and qualitative evidence to
indicate that underground excavations are considerably more ‘robust’ than surface facilities when subject to
severe seismic excitation. However, there are situations where the epicenter is much closer and high frequency
effects may cause significant damage to underground structures, as in the case of the Mw = 6.9 Kobe earthquake
of January 17,1995. Advances in the three-dimensional numerical modeling of dynamic loading of underground
excavations in rock now allow more quantitative explanations of the qualitative observations of the response of
structures in rock, such as referred to above. This modeling provides a very effective tool to (i) demonstrate the
ability of the numerical modeling procedures (codes) to replicate observed behavior; (ii) optimize the design of
underground excavations in rock to house underground nuclear power plants, providing a much more robust and
safer environment for generation of nuclear power than a surface – based plant in the same general location. The
paper presents several examples to illustrate the use of these numerical codes for this purpose.
1 INTRODUCTION public opposition to the continued use of nuclear

energy.
“If over the next decade the world demonstrates that This opposition moderated somewhat with rising
it can do nuclear power well, a global expansion of concern over climate change. The ‘green’ nature of
nuclear power would have to be – indeed, should be – nuclear energy was emphasized, countries progressed
seriously reexamined.” (Socolow and Glaser, 2009) in the development of geological repositories for dis-
Of the technological options now available to reduce posal of the high-level radioactive waste from nuclear
global dependence on fossil fuels, none has been reactors; and the lack of further serious power plant
more widely applied than nuclear energy. Currently, accidents increased confidence in the safety of nuclear
13 countries rely on this source to supply at least one- power.
quarter of their total electricity (WR1 [Web Reference Japan, a country with few fossil fuel resources, gen-
1]). France, once heavily dependent on oil imports, erated 30% of its electricity from its 43 nuclear reactors
now generates 80% of its electricity from nuclear and, even though part of a seismically active region of
energy. the world, planned to increase this to 40%, when, in
Loss of coolant accidents, resulting in core melt- March 2011, a large (Mw = 9.0) offshore earthquake
down, at the TMI-2 reactor, Three Mile Island, produced a major tsunami [15m above sea level at
Pennsylvania, USA in 1979 (WR2), followed by Fukushima (WR4) and almost 40 m elsewhere on the
the more severe event at Chernobyl, Ukraine, in coast (WR5) at Sendai] that inundated the Daiichi 1
April 1986 (WR3),when the explosion resulted nuclear plant and resulted in three core meltdowns. All
in release of radioactive particles into the atmo- 43 nuclear reactors were shut down after Fukushima –
sphere and dispersion internationally, provoked strong two have re-opened recently (WR6).
663
Although considerable development has taken place 2.1 Earthquake resistance of underground nuclear
in nuclear technology over the years, see e.g. Devel- power plants
opment of Inherently Safe Nuclear Reactors, Ragheb
A prominent concern with respect to nuclear power
(2015), public opposition to expansion of nuclear
plants, especially since Fukushima, is the vulnerability
power has hardened in the light of Fukushima.
to damage when subject to a severe earthquake. There
is considerable ‘anecdotal’ and qualitative evidence to
2 UNDERGROUND NUCLEAR POWER indicate that underground excavations are consider-
PLANTS ably more robust than surface facilities when subject
to severe seismic excitation. In the Northridge, Califor-
“Plainly, mankind cannot renounce nuclear power, so nia, Mw = 6.7 earthquake of Jan.17, 1994 for example,
we must find technical means to guarantee its absolute it was reported
safety and exclude the possibility of another Cher- “...the subway tunnels have not suffered any dam-
nobyl. The solution I favor would be to build reactors age; train service was quick to return…. Generally
underground, deep enough so that even a worst case speaking, subways in many other areas have survived
accident would not discharge radioactive substances earthquakes with minimal or no damage— and often
into the atmosphere.” Andrei Sakharov (1990). far less damage than is suffered by buildings and
Construction of nuclear power plants underground roads” (WR7)
at relatively shallow depth [∼ (50 −150) m] in rock, Comments such as these are typical for cases where
has long been advocated as a way to improve safety. the earthquake epicenter is tens or more kilometers
Among the first of these studies was that by Watson from the underground structure. In these cases most
and Kammer (1972). Underground nuclear plants have of the high frequency components of the seismic exci-
been operated at various locations around the world tation have been substantially attenuated before the
(Duffaut and Vaskou, 2014). wave reaches the underground structure.
A rock cover of just 25 m is sufficient to provide There are at least two basic reasons for these
essentially the same containment as an above-ground observations –
containment structure (i.e. 0.6 MPa or 80 psi). Under- • The ‘Seismic Advantage’i.e. reduced kinetic energy
ground plants are also much more resistant to aircraft
- away from a free surface.
impact – and to the threat of terrorist attack. • Underground structures are not free-standing, and
Neretnieks (1980) has proposed a design, illustrated
hence are much less subject to the large inertial
in Figure 1, to prevent release of radioactive gases to
damage often observed in above–ground structures.
the atmosphere even in the event of a core meltdown.
Underground ‘safety tunnels’ are filled with crushed An example of effects when the earthquake motion
rock and are of sufficient total volume to accommodate has significant energy in high frequencies, as in the
the contaminated gas without release of radionuclides Kôbe (Japan) Earthquake, January 17, 1995, will be
to the atmosphere. Figure 1 illustrates the concept. In discussed later in this paper.
this illustration, the safety tunnels are connected to an
above-ground nuclear plant but, obviously, the idea is
readily applied to underground nuclear plants.
The development of small, modular nuclear reac- 3 THE ‘SEISMIC ADVANTAGE’ OF
tors has provided additional flexibility in the design of UNDERGROUND STRUCTURES
underground nuclear power plants (Myers and Mahar,
2011). The energy of an elastic wave traveling through the
interior of a solid is partitioned equally i.e. 50% strain
energy and 50% kinetic energy. This changes as the
wave arrives at the surface. By definition, a free sur-
face cannot sustain strain when subject to an elastic
wave. Thus, at the surface, the strain – together with
the associated stress, and strain energy – must be zero
at all times. Correspondingly, the kinetic energy at the
surface is doubled, i.e. to 100%.
To satisfy the free surface condition, the arriving
elastic wave results in a complementary wave of equal
magnitude and opposite sign that is reflected (i.e. an
arriving compression wave is reflected as a tensile
wave of equal magnitude, and vice-versa) traveling at
the same velocity as the incident wave, back into the
solid medium.
The situation is illustrated in Fig. 2 (Kolsky, 1963)
for a plane elastic wave of arbitrary shape reflected at
a free surface. In Fig. 2(a), the wave is about to arrive
Figure 1. Design principle for safety tunnel. at the free surface. In 2(b) reflection has started to
664
Also, there are situations where the epicenter is
much closer and high frequency effects (10 Hz or
possibly higher) may cause damage to underground
structures, as has been presumed to have occurred in
the case of the MW = 6.9 Kobe earthquake of January
17, 1995 (Sakurai, 2014 and Uenishi, 2012). The con-
sequences of these effects will also be illustrated below
by numerical analysis involving typical earthquake
waveforms.
4 NUMERICAL MODELING OF
EARTHQUAKE LOADING OF
UNDERGROUND STRUCTURES
Advances in numerical modeling of dynamic loading

of underground excavations in rock now allow more
Figure 2. Interaction between incident and reflected waves
at a free surface. [Kolsky (1963) p.89.]
quantitative explanations of the qualitative observa-
tions of the response of structures in rock, such as
referred to above. This numerical modeling provides
occur. The net strain at the free surface due to passage a very effective tool to (i) demonstrate the ability of
of the wave must be identically zero at all time during the modeling procedures (codes) to replicate observed
passage of the wave, and hence can be determined behavior; (ii) optimize the design of underground
by superimposing the portion of the wave that can be excavations in rock to house underground nuclear
imagined to have ‘passed beyond’ the free surface [i.e. power plants, to provide a much more robust and safer
the dotted portions of the wave to the right of the free environment for generation of nuclear power than a
surface in Figs. 2(a), 2(b) and 2(c)] but have reflected surface – based plant in the same general location.
as a wave of equal magnitude, opposite sign, – as seen The following discussion illustrates the use of these
in the lower left quadrant of Figs. 2(a),2(b), 2(c). The numerical codes to explain the observed response of
net strain experienced by the elastic solid is the sum underground excavations to earthquake loading, and
of the incident and reflected waves i.e. the heavy solid demonstrate the use of the codes to design robust, safe
line shown in Figs 2(a) – (e). It is seen that the net strain underground nuclear power plants.
is identically zero at the free surface and relatively low
during the first portion of the wave overlap i.e. 0 to λ/4
[i.e. as illustrated in Figs 2(b), 2(c), 2(d)]. 4.1 Kinematic and inertial response of structures
Seismic slip on a fault produces waveforms com-
The deformation of a structure during shaking is
posed of a spectrum of frequencies, high and low.
affected by interactions between three linked systems:
Higher frequencies attenuate more rapidly during
the structure, the foundation, and the geologic medium
propagation; hence the energy in waveforms imping-
underlying and surrounding the foundation.
ing on a structure from a distant epicenter tends to
The response of the structure can be divided into two
be concentrated in the lower frequency (1 Hz∼5 Hz)
main components, kinematic and inertial, as illustrated
range of the spectrum. Recalling that the wavelength
in Figure 3.
λ = v/f, where v is the wave speed and f is the
wave frequency, then, for a shear wave velocity vs of • Kinematic response: Even if the mass of structure
1000 m/sec and a dominant shear wave frequency of is essentially zero, the presence of structural ele-
5 Hz, the ‘λ/4 overlap’ region corresponds to a depth ments of finite stiffness either on the formation or
of the order of 200 m. embedded in the underlying medium, result in some
One consequence of proximity to the free surface is deviation of the foundation motion with respect to
the fact that dynamic stressing of the rock in the over- the corresponding motion of the ‘free-field’, i.e. the
lap region is less than at greater depths but at the same response of the geologic medium in the absence of
time the kinetic energy is higher. The principal advan- the structure. The prominent mechanisms contribut-
tage of the underground location is associated with ing to such deviations are embedment effects and
the kinetic effects of the wave motion. As noted ear- wave scattering produced by corners and asperities
lier, the kinetic energy is a maximum at the surface and of the foundation. The embedment effects are sig-
declines towards the interior. This attribute is accom- nificant if the foundation is much stiffer than the
panied by another – the ability to limit the motion of geologic medium. In this case the medium cannot
the subsurface structures by ‘attaching’ them to the deform in exactly the same form as the free-field,
rock mass. As discussed below, the net kinetic energy and the free-field motion is ‘filtered’ by the foun-
of the incoming and reflected waves results in inertial dation to a degree that depends on the excitation
and kinetic response of structures either on the surface wavelength.
or underground. Particular attention is therefore paid • Inertial response: This mechanism refers to the
to this aspect of the wave interaction. response of the structure-foundation system to
665
Figure 3. Inertial response (left) and Kinematic response
(right) of structures embedded in rock or soil.
excitation by inertial forces (D’ Alembert forces)

associated with the acceleration transferred by the
foundation to the structural mass from kinematic
interaction.
Nuclear power plants are usually constructed on top
of competent rock rather than on soft soils. Since the
stiffness of the power plant structure is comparable to
that of rock, embedment effects are not significant. For
a shear wave velocity of 1000 m/s – typical for rock –
Figure 4. Incoming ground motion for the two earthquakes
the shortest wavelength for frequencies with reason- considered in the analyses; Mw = 7.01 event (top); Mw = 7.35
able energy in incoming motion (up to 20 Hz) is around event (bottom).
50 m. Assuming a typical dimension of 30 m × 30 m
for the containment structure, wave scattering effects
are not expected to be significant. The ground motion
transferred to the structure is thus expected to be
almost the same as the free-field ground motion at
that level.
The following sections demonstrate that siting of
a reactor underground results in reduced effects, both
kinematic and inertial, on the emplaced structure.
4.1.1 Free surface amplification

We consider two earthquake ground motions for this
case, with incoming motion and frequency content as
shown in Figure 4 and Figure 5. The first ground
motion corresponds to a 7.01 Mw event with PGA
(Peak Ground Acceleration) around 0.5 g and a dom-
inant frequency of 1.5 Hz; the second motion corre-
sponds to a 7.35 Mw event with PGA around 0.4 g and
a broad frequency content from 0.2–8.0 Hz. While the
analysis is presented for just these two ground motions,
the qualitative response is likely to be the same for any
type of earthquake. The site is assumed to be rock with
shear wave velocity Vs = 1000 m/s
Study of the interaction between the Incident and
Reflected Wave for the Mw = 7.01 event is shown in
Figure 6. Results for the Mw 7, 35 event show essen-
tially similar overall behavior. The ‘doubling effect’
can be seen at t = 4.0 sec when the main pulse reaches
the surface. It is this amplified acceleration that is
‘imposed’ on free-standing surface structures in the Figure 5. Frequency content for the two earthquakes con-
vicinity of the earthquake. sidered in the analyses; Mw = 7.01 event (top); Mw = 7.35
The results for the net acceleration (in blue) at depth event (bottom).
indicate local minimum for ground acceleration in
the range of 150 m–225 m, which corresponds to the
same λ/4 criterion (λ = Vs /fD = 667 m in this case). frequencies in the second motion. Nonetheless, the
The profile also shows another minimum closer to motion at the surface is the highest. This high surface
the surface for Mw = 7.35 earthquake at around 50 m acceleration is transmitted to the base of surface struc-
depth. This is due to the presence of multiple dominant tures, setting these into inertial motion - that is usually
666
Figure 6. Ground acceleration in various stages of interac-
tion between incident and reflected wave.
considerably larger than for the structure embedded in

the rock.
As noted earlier, nuclear power plants are gener- Figure 7. Maximum ground acceleration experienced at
ally required to be constructed on top of hard rock. various depths from a free surface during the Mw = 7.01
This is due to the fact that softer soil deposits on top Seismic Event (left); and Mw = 7.35 Event (right).
of stiffer rock tend to trap seismic waves propagating
downwards, causing resonance, and thereby greatly are also included in the numerical simulations for
amplifying the incoming motion at times. Protection the softer deposit. The results for both motions show
against damaging effects of a tsunami require, in prin- amplification of ground motion in the soft deposit,
ciple, locating the plant at an elevation higher than but a significant decrease in maximum acceleration
any potential tsunami could reach. However, these two recorded in the rock.
objectives can often conflict with each other for free The profiles correspond to two cases; (i) rock only
standing and embedded configurations as in the case (blue line); (ii) 50m soil on top of rock (green line).
of the Fukushima I Nuclear Power Plant. The reactor
was built on a bluff that was originally 35 meters high. 4.1.2 Inertial response
The height of the bluff was lowered to 10 m to allow Waves transmitted to free-standing surface structures
placement of the foundation on hard rock to reduce can create high oscillations, leading eventually to col-
the vulnerability to earthquakes and also to reduce lapse. However, excavations are not ‘free-standing’ –
the operating cost for the water pumps. However, this they are in the interior of a 3-dimensional solid. As
increased the vulnerability to tsunamis. discussed before, the dimensions of most excavations
in rock are low compared to the dominant wavelengths
• incident wave approaches free surface, t = 3.5 sec;
(i.e. high energy components) of a large seismic
(peak acc’n – in red circle – 2.5 m/s2 )
event – the incident and reflected waves pass across
• peak acceleration reaches free surface, t = 4.0 sec,
and beyond the excavations with little major change
– acc’n doubles to 5 m/s2 –in red circle;
assuming the excavations are designed to sustain local
• after reflection, t = 4.5 sec; peak acc’n falls again to
stress amplifications around the tunnel periphery. For
2.5 m/s2 .
the next set of simulations, we assume the contain-
Underground configuration on the other hand can ment structure to be 30 m × 30 m and made of 1 m
be protected from the damaging effect of tsunami, thick concrete. The structure is assumed to be square
requiring only the ability to seal the access shafts for simplicity. For the underground case, the structure
and tunnels in the event of a flooding event, or to is placed 150 m below the ground surface. Figure 8
ensure that plant entrances to the sub-surface are at compares the profiles of maximum moment recorded
elevations above the height of any potential tsunami. in the structure during the shaking. The profile is plot-
Furthermore, the presence of softer material such as ted as distance measured from the middle of roof of
soil or rock above the underground structure does not the structure. The peaks correspond to the corners. The
adversely affect the underground design. In fact, in low values (at 0; +/−30 m; +/−60 m) correspond to
most cases, it will lead to even smaller motion in the the centers of each side of the square structure.
rock, since part of the waves reflected from the surface Siting the structure underground results in lower
are trapped by the overlying softer material. moments throughout the structure, especially in the
Figure 7 (green) shows the maximum recorded sidewalls, where the moment is much smaller than for
acceleration profile for the case where a 50 m thick the free-standing case. The moment at the corners (i.e.
softer deposit, with shear wave velocity varying lin- +/−15 m and +/−45 m) can be further reduced for
early with depth from 200 m/s to 500 m/s, overlies the underground case by making the shaft circular.
the rock. Modulus reduction as a function of shear For the response of the reactor itself, we consider
strain and hysteretic damping typical for dense sands a reactor 25 m high and 4 m diameter. The reactor
667
Figure 9. Maximum moment along the height of a 25 m high
(cylindrical) reactor (0 = base) for several configurations –
and two seismic events; Mw = 7.01 (left) and Mw = 7.35
(right).The configurations are (a) free-standing on the ground
surface; (b) free standing in an excavation 150 m below
ground; (c) laterally supported in an excavation 150 m below
ground; (d) embedded in concrete in an excavation 150 m
below ground;
curve) with the maximum moment occurring higher

up from the base. The laterally supported configura-
tion (blue curve) results in about 30% reduction. The
reduction is higher for both cases for second ground
motion.
4.2 Effect of high frequency wave motion on

underground structures
While most literature reports that deeper underground
structures fare much better during earthquakes than
Figure 8. Maximum Moment in a Square Containment above ground or shallow buried structures, especially
Structure - Mw = 7.01 seismic event (left) and Mw = 7.35 cut-and-cover installations, there are some cases where
event (right). (Distance along the structure is measured from damage has been observed in deep underground tun-
the middle of the roof +/−60 m to middle of floor). nels. One such case is the Bantaki tunnel, which was
damaged during the 1995 Kobe earthquake (Mw =
6.9). The Bantaki tunnel is 1766 m long, mainly in
dimensions correspond to a Babcock and Wilson m- granite rock. It was constructed using the NATM. The
power modular nuclear reactor. The external shell of damage was observed only in the central portion where
the reactor is assumed to be 0.20 m thick to calculate the tunnel passes through a fractured zone as shown in
the area and bending moment of inertia. The reactor is Figure 10. The height of overburden is approximately
embedded 5 m in concrete for the free-standing case. 200 m − 250 m at this location. Damage occurred at
Figure compares the bending moment in the reactor two locations along the cross-section as shown in Fig-
for the free-standing case at the surface and at a depth ure 11. The first location was along the spring line of
of 150 m in a 30 m × 30 m shaft. The difference is the tunnel where the concrete lining failed and split,
not that significant as the configuration is essentially and the reinforcing steel bars buckled. The second
the same – but with some reduction attributed to the location was at the base where the base slab rose up
smaller motion at 150 m depth. The maximum moment and detached from the invert (10 cm gap). No damage
is observed at 5 m from the base where the embedment to the invert was observed. Sakurai (2014) and Uenishi
ends. For a more robust seismically resistant design, (2012) attributed this damage to very high frequency
two other options are investigated. In one case the reac- content (50 Hz) in the vertical component of ground
tor is embedded in concrete up to 20 m whereas in the motion. They proposed that P-waves caused shorten-
other case the reactor is attached laterally to the wall ing deformation between the concrete in the tunnel
of the shaft at 20 m from the base. crown and the invert concrete, which led to uplifting
Comparison of the maximum moment in the reactor of the base slab, and failure of the concrete and buck-
for these two options versus the underground case is ling of the steel reinforcing bars at the spring line.
also presented in Figure 9. The distance along the 25 m They supported their hypothesis that the damage was
height of the structure is measured from the base to the caused by seismic waves with the fact that no perma-
top. Results for the Mw 7.01 ground motion indicate nent deformation was observed (at the invert and other
a significant reduction (almost 85%) in the moment locations) that rules out fault movement or instability
in the reactor for the concrete embedment case (green around the portal as the likely cause.
668
Figure 10. Geological and topographic configuration
around the tunnel (Sakurai, 2014).
Figure 12. Acceleration time histories from 1995 Kobe
earthquake recorded at Kobe University. The horizontal com-
ponent is shown in the upper diagram and the vertical
component in the lower diagram.
Figure 11. Cross-section of the Bantaki tunnel showing the

damaged regions (Sakurai, 2014).
Higher frequencies generally attenuate much faster

with distance and are usually not considered when
analyzing the response of structures. The existing seis-
mograms are also not designed to record such high
frequencies (>50 Hz). Based on the observations in
Bantaki tunnel, Sakurai (2014) and Uenishi (2012) rec- Figure 13. Amplitude spectra for the acceleration vs time
ommended that more attention be given to detection histories from the 1995 Kobe earthquake recorded at Kobe
of such very high frequency waves and to account for University. The horizontal component is shown on the top
and the vertical component on the bottom.
them during seismic analyses of structures.
The 1995 Kobe earthquake did have a very strong
vertical component compared to other earthquakes components originated from the main event and trav-
observed previously in the area as shown in Figure 12 elled more than 30 km (the distance of Bantaki tunnel
and Figure 13. Figure 12 shows the horizontal and ver- from epicenter) without significant attenuation.
tical components of acceleration time history recorded The analyses by Uenishi (2012), which relates the
at Kobe University (PEER, 2015). The average shear amplification in hoop stress at the spring line of the
wave velocity for the top 30 m at the site is > 1000 m/s, support (i.e. the dynamic hoop stress at the spring line
so site amplification effects appear to be negligible. normalized by the stress amplitude of the incoming
Figure 13 shows the amplitude spectrum for the hori- wave) as a function of (λ/a), [i.e. wavelength of the
zontal and vertical components, showing high energy incoming wave normalized by tunnel radius] indicates
content up to 10 Hz in vertical component. that P-waves with λ/a = 50 result in a higher stress
While it is possible that very high frequency waves amplification at the spring line than λ/a = 10. For an
(50 Hz) were generated by a smaller event in the frac- inner radius of 4 m and P-wave velocity of 2000 m/s,
tured region in close proximity to the tunnel; triggered λ/a = 50 corresponds to a frequency of 10 Hz, which
by the incoming ground motion from the main event; was indeed observed in the vertical component of the
the authors are skeptical that these very high frequency incoming motion.
669
The higher frequency (50 Hz) P-waves were used to
explain the damage to the base slab only.
Uenishi (2012) used radial particle velocity induced
by the P-waves and compared the difference in peak
velocity at the crown and invert of the tunnel – with
significant difference indicative of base slab damage.
However, this is not the right criterion and contradicts
the evidence that damage was observed only to the base
slab, while both the invert and the crown remained
undamaged. In fact, significant vertical acceleration
(even low frequency) in the downward direction would
detach the slab from the liner (because the liner is
constrained to move with the rock but the slab is not)
and very high frequencies (50 Hz) are not needed to
explain the damage observed.
A simple 2D numerical simulation has been carried
out using FLAC (Itasca, 2011) to evaluate the dynamic
hoop stresses introduced by the vertical component of
the recorded ground motion. The model consists of the
10 m diameter tunnel with a liner thickness of 0.65 m
and an overburden of 200 m. The rock is assumed to
have a P-wave velocity of 2000 m/s, and a stiffness
of 40 GPa is used for the concrete liner. The vertical
component of ground motion is divided by a factor of
two to estimate the incoming motion (upward traveling Figure 14. Numerical model showing the tunnel and liner
wave) which is applied at the base of the model. The (top) and seismic motion induced maximum hoop stresses as
maximum hoop stress resulting from the axial forces a function of location (bottom).
along the liner is monitored and presented as a function
of angle as shown in Figure 14.
The results indicate that compressive stresses, up
to 4.3 MPa, were generated at the spring line due to
the dynamic shaking. These stresses are in addition
to those from the static loading of the liner. Keeping
in mind that the unconfined compressive strength of
concrete is around 25-30 MPa, the seismically induced
stresses due to this vertical component are very high, Figure 15. Numerical model for the Bantaki tunnel site
even without accounting for any energy in the very showing the fractured region, tunnel alignment and location
of the observed failure.
high frequency domain.
The compressive stresses from dynamic loading
calculated above do not account for other amplifica-
fractured region tends to trap the seismic waves. The
tion mechanisms. The most important is the amplifi-
results show that while the horizontal component is
cation of ground motions due to topographic effects
not affected much by the fractured region, high ampli-
and presence of discontinuities and stiffness contrast.
fication of the vertical component is observed in the
A numerical simulation to evaluate this effect is car-
fractured region, with amplification factors as high
ried out as shown in Figure 15. The surrounding rock
as 2.5 observed close to the failure location (marked
mass (granite) is assumed to have a P-wave velocity
on the plots). For comparison, an amplification factor
of 4000 m/s and S-wave velocity of 2000 m/s whereas
of 1.7 was observed at the tunnel location for level
the fractured region has a P-wave velocity of 2000 m/s
ground and no stiffness contrast used in the analy-
and S-wave velocity of 1000 m/s.
ses to calculate the compressive stresses in the liner.
Both horizontal and vertical components of ground
This additional amplification of the vertical compo-
motion are applied at the base of the model and ampli-
nent would lead to even higher dynamically induced
fication factors for peak velocity are calculated along
compressive stresses in the liner and combined with
the entire region for both horizontal (Figure 16) and
higher dynamic stress amplification observed for the
vertical (Figure 17) components. Note that for level
liner in fractured (lower stiffness) rock compared to
ground with uniform properties, an amplification fac-
good quality (high stiffness) rock by Uenishi (2012),
tor of 2 at the surface is expected due to free surface
would explain the failure observed at the spring line
amplification as discussed earlier in this paper. Fac-
only along the portion of the tunnel in the fractured
tors greater than two observed here are a result of (a)
region.
topographic amplification as the hill geometry tends
The above example shows that it is important to
to focus the incoming seismic energy, resulting in
avoid such regions of fractured rock resulting in
amplification of the motion; (b) the presence of the
stiffness contrast especially in hill type geometries
670
groundwater level and/or low permeability rock are
also valuable attributes of a site – to minimize pumping
requirements for the facility.
6 CONCLUSIONS
Underground location has significant safety benefits

compared to surface placement for a variety of com-
Figure 16. Amplification factors for peak ground veloc-
ity (m/s) in the horizontal direction. The Bantaki tunnel mercial and industrial applications. This paper focuses
alignment and the failure location are also shown. on the benefits of underground siting of nuclear
power plants, with emphasis on protection from earth-
quakes.Although the most established source of ‘green
energy’, public opposition to use of nuclear power has
intensified, especially in seismically active regions of
the world, as a result of the very serious accident at
Fukushima, Japan, in 2011 resulting from the major
earthquake (Mw = 6.6) and associated tsunami.
Several underground nuclear plants have been con-
structed and operated, but there is no known record of
such a plant having been subject to a damaging earth-
Figure 17. Amplification factors for peak ground velocity quake. There is evidence that shallow underground
(m/s) in vertical direction. The alignment for Bantaki tunnel structures e.g. subway transit systems, perform sig-
and failure location are also shown. nificantly better than surface facilities during severe
earthquakes.
when seating underground nuclear reactors, as these Numerical modeling techniques are now available
conditions increase the potential for damage during that can provide valuable assessments of the dynamic
a seismic event. However, it should be noted that, response of underground facilities (e.g. mines, nuclear
despite such strong shaking and unfavorable geologi- waste repositories.) and can provide valuable insights
cal and topographic conditions, the damage observed on the influence of local topography and geology in
in the Bantaki tunnel was still much smaller than siting considerations.
the damage observed in the above-ground structures Although the numerical analyses carried out in this
during the 1995 Kobe earthquake. This reinforces paper are highly simplified, they serve to illustrate
the point that underground structures in rock are the benefits of the location and provide an instruc-
safer than aboveground structures during earthquakes. tive qualitative comparison of different underground
Furthermore, the example illustrates how numerical configurations. It is seen that peak ground acceler-
analyses can aid in the selection of site locations for ations are reduced compared to those at the ground
UNPPs by identifying potentially problematic regions surface. Unlike surface facilities, excavations are not
through consideration of the site geology and material free-standing structures, and so experience lower iner-
properties. tial forces. Reactors – especially modular units- can
be laterally supported to the wall of the excavation or
embedded in concrete, reducing forces and bending
moments considerably, resulting in a robust seismic
5 SOME GENERAL CONSIDERATIONS RE design.
SITING OF UNDERGROUND NUCLEAR More detailed and sophisticated numerical model-
POWER PLANTS ing efforts can be conducted to better quantify these
advantages for specific underground locations and
For situations where it is necessary to locate under- designs.
ground nuclear power plants close to the ocean, it is
preferable that the surface access to the plant be at
an elevation of the order of 50m or more, and that REFERENCES
this access be equipped with bulkheads that can be
activated rapidly in the event of a major earthquake. Damjanac. B, Board M., Lin M., Kicker D. & Leem J. 2007.
The underground plant should be located in com- Mechanical degradation of emplacement drifts at Yucca
petent rock. A main advantage of the underground Mountain – A modeling case study. Part II: Lithophysal
location is the ability to reduce inertial loading of the rock. Int. J. Rock Mech. & Mining Sci., 44: 368–399
reactor by lateral support and embedment. This can Duffaut, P. and Vaskou, P. 2014. Geological and Geographical
Criteria for Underground Siting of Nuclear Reactors, 8th
be accomplished at any depth in the rock, indicat-
Asian Rock Mech. Symp. Sapporo, Japan
ing that the reactor can be as shallow as considered Giraud, K. M., Kunze J F., Mahar J. M. & Myers C.W.
desirable, provided there is a minimum cover of 25 m 2009. Cost Advantages of Large Underground Nuclear
below surface-in order to meet containment require- Power Parks, American Nuclear Society, Annual Meeting,
ments comparable to a typical surface design. A low Atlanta, GA, June 14–18, 2009.
671
Itasca Consulting Group, Inc. 2011. FLAC (Fast Lagrangian Sakurai, S. 2014. Case Studies on the Dynamic Behav-
Analysis of Continua), Version 7.0. Minneapolis: Itasca. ior of Tunnels Caused by Hyogoken–Nanbu Earthquake,
Kolsky, H. 1963. Stress Waves in Solids. Dover (New York) Whose Epicenter was Very Close to the Tunnels. Proc.
213p. (See p.89) 8th Asian Rock Mechanics Symposium ARMS8 14–16
Lin. M, Kicker D., Damjanac B. & Karakozian M. 2007. October 2014, Sapporo, Japan.
Mechanical degradation of emplacement drifts at Yucca Socolow, R. & Glaser A. 2009. Balancing risks: nuclear
Mountain – A modeling case study. Part I: Nonlithophysal energy & climate change; Daedalus, Amer. Acad. Arts
rock. Int. J. Rock Mech. & Mining Sci., 44: 351–367. & Sci. Fall 2009, 138(4): 31–44.
Giraud, K. M., Kunze J F., Mahar J. M. & Myers C.W. Uenishi, K. 2012. Elastodynamic Analysis of Underground
2009. Cost Advantages of Large Underground Nuclear Structural Failures Induced by Seismic Body Waves
Power Parks, American Nuclear Society, Annual Meeting, J. Appl. Mech 79(3), 031014 (Apr 05, 2012)
Atlanta, GA, June 14–18, 2009. Watson, M.B., Kammer.W.A. 1972. Underground nuclear
Myers,W. and Elkins N. 2004. Siting Nuclear Power Plants power plant siting. E.Q.L. Report No.6 Environmental
Underground: Old Idea, New Circumstances, Nuclear Quality Laboratory, California Institute of Technology,
News, 47(3): 33–38. Pasadena, CA. 91109, Sept 1972, 150pp.
Myers, C.W. & Mahar J.M. 2011. Underground Siting
of Small Modular Reactors: Rationale, Concepts, and WEB REFERENCES.
Applications. ASME Small Modular Reactor Symposium WR1. http://www.nei.org/Knowledge-Center/Nuclear-
September 28–30, 2011 Washington, DC. USA Statistics/World-Statistics
Neretnieks, I. 1980. Safety Tunnel for Core Melting In WR2. http://www.nrc.gov/reading-rm/doc-collections/
Nuclear Power Plants Underground Space, Pergamon, factsheets/3mile -isle .html Note “the small radioactive
Oxford. 1: 179–180 [See also Neretnieks, I. and Nykvist releases had no detectable health effects on plant workers
B. 1978. Nägra möjligheter till konsekvenslindring vid or the public.”
härdsmältning; Some possibilities to decrease the con- WR3. https://en.wikipedia.org/wiki/Chernobyl_disaster
sequences of a core melt; written in Swedish; Dept. of WR4. https:/ / en.twikipedia.org / wiki / Fukushima _ Daiichi_
Chemical Engineering. Royal Institute of Technology, nucleardisaster “The earthquake triggered a 13-to-15-
Stockholm.] metre (43 to 49 ft) maximum height tsunami that arrived
Pacific Earthquake Engineering Research Center. 2015 approximately 50 minutes later.”
PEER Ground Motion Database http://ngawest2.berkeley. WR5. http://www.livescience.com/39110-japan-2011-earth
edu/ quak e-tsunami-facts.html Note; “The tsunami waves
Ragheb, M. 2015. Inherently Safe Reactor Designs. Univ. reached run-up heights (how far the wave surges inland
Illinois, Urbana–Champaign, USA above sea level) of up to 128 feet (39 meters) at Miyako
Sakharov Andrei. 1990. Memoirs, Random House, USA city and traveled inland as far as 6 miles (10 km) in
773p. Quoted in W. Myers and N. Elkins. 2004. Under- Sendai.”
ground Nuclear Parks and the Continental SuperGrid WR6. http://www.nei.org/News-Media/News/Japan-Nuclear-
(Los Alamos Nat’l Lab.) SuperGrid 2 Conference, Univer- Up date
sity of Illinois at Urbana-Champaign, Oct. 25–27, 2004. WR7. http://thesource.metro.net/2012/08/10/designing-a-
[A similar opinion by Edward Teller is also quoted.] subwa y-to-withstand-an-earthquake/
672
An analytical study of PRACLAY Heater test
G.J. Chen
European Underground Research Infrastructure For Nuclear Waste Disposal In Clay Environment (EURIDICE),
Belgian Nuclear Research Center, Mol, Belgium
L. Yu
Institute for Environment, Health, and Safety (EHS), Belgian Nuclear Research Center, Mol, Belgium
Xiang Ling Li
European Underground Research Infrastructure For Nuclear Waste Disposal In Clay Environment (EURIDICE),
Belgian Nuclear Research Center, Mol, Belgium
ABSTRACT: In the context of a large scale in-situ PRACLAY Heater test which is currently running at the
HADES underground research facility (URF) in Mol, Belgium, an analytical solution is developed to study the
Thermo-Hydro-Mechanical (THM) responses in the Boom Clay, backfilling sand and supporting liner around
the heated PRACLAY gallery. After sufficient verification, the present solution is applied in the PRACLAY
Heater test for a predictive reference calculation and a parametric sensitivity study. The new analytical solution
proves to be a convenient tool for a good understanding of the resulting coupled THM behaviour and is therefore
valuable for the interpretation of measured data in engineering practices and for a rational design of potential
radioactive waste repositories.
1 INTRODUCTION
There is a general consensus that geological disposal is

considered a possible solution for the long-term man-
agement of high-activity, long-lived radioactive waste
(NEA, 2008). As a side effect of radioactive decay,
a significant amount of heat is released even after
several decades of cooling in surface facilities, which
makes the host formation undergo significant temper- Figure 1. Layout of the PRACLAY gallery.
ature gradient, induce hydraulic PWP gradient and a
change in effective stresses. period of time. Therefore, the gallery was backfilled
The large scale PRACLAY Heater test is carried with high-permeability sand which is required to be
out in the HADES URL in Mol, Belgium with the fully saturated before heating, and a target tempera-
purpose to simulate the thermal impact generated by ture of 80◦ C is decided at the interface between liner
heat-emitting high-level waste (HLW) on Boom Clay and host Boom Clay in the test for at least 10 years.
at a repository-representative configuration (Li et al. After the pioneering consolidation theory (Biot
2013, Van Marcke et al. 2013), and to directly gather 1956 & 1957, Rice and Cleary 1976), there exists
knowledge of the THM response of the Boom Clay a substantial and growing suite of analytical solu-
to complex loadings. The PRACLAY Heater test is tions to predict the change in stress and pore pressure
hosted in the PRACLAY gallery (Figure 1), which has induced by a thermal loading in a saturated poroe-
been established in the Boom Clay at 223 m depth. lastic medium (Booker and Savvidou 1984 & 1985,
The PRACLAY gallery is a horizontal drift with a Giraud and Rousset 1995, McTigue 1986, Savvidou
length of about 44 m, internal and external diame- and Booker 1989, Wang and Papamichos 1999), and in
ters of 1.9 and 2.5 m, respectively. A hydraulic seal layered saturated poroelastic media (Small and Booker
was installed to separate the heated part (33 m) of 1986, Giraud et al. 1998).
the gallery from the nonheated part (10 m). Due to In the context of PRACLAY Heater test, an analyt-
the restrictions of an in situ test compared to the real ical solution of the temperature, pore pressure, stress
repository, the PRACLAY Heater test intends to reach and displacement fields around an infinite-long cylin-
the most critical thermal load in terms of maximum drical cavity is obtained with a time-dependent heat
temperature as well as temperature gradient within a source applied in the cavity. The cavity is backfilled
675
with saturated material, supported by casing or liner
and hosted in a low-permeability geological forma-
tion. The analytical solution is formulated in Laplace
domain, and results in real time space are obtained
by numerical inversion. Since this analytical solution
considers the plane strain condition, it is more repre-
sentative for the middle cross-section of PRACLAY
Heater test (Figure 1). The new solution is applied in
PRACLAY Heater test for a predictive reference calcu-
lation, and the results are verified by numerical results
obtained from the commercial finite element software
COMSOL Multiphysics (COMSOL 2008). An exten-
sive parametric study is performed for PRACLAY
Heater test.
Numerical analyses are essential in the whole pro-
cess of PRACLAY Heater test, from test design,
construction, predictions, risk analysis to data inter-
pretations. Although numerical simulations based on
finite element method play an important role, in
parallel, (semi)-analytical solutions are necessary to Figure 2. Problem geometry and materials.
provide confidence in the results, and further con-
tribute to aspects such as code verification, model
conditions. The nature and geometry of the problem
validation, process understanding with respect to cou-
defined in Figure 2 indicate that the problem consid-
pling behaviour, and sensitivity and uncertainty anal-
ered here possesses radial symmetry, and a cylindrical
ysis. The solution provided in this study solves the
co-ordinate system (r, θ) is therefore used. The classic
THM coupled problem in a semi-analytical manner,
sign convention for poroelasticity is used in this study,
hence with a high computational efficiency. Temper-
in which the tensile stresses and strains are considered
ature, pore water pressure, stress and displacement
positive.
fields can be obtained at any time without the need
for time-stepping algorithms which are encountered
in numerical approaches. Therefore, analytical solu- 2.2 Balance equations
tions are of great value in preliminary designs in
which only a limited knowledge of the range of val- The governing equations for non-isothermal consoli-
ues associated with the material parameters is known dation in a linear poroelastic medium are formulated
(Selvadurai 2007). They can be used at this stage for here in terms of three primary variables: temperature
quick order-of-magnitude estimates, for better under- (T), pore water pressure (p) and radial displacement
standing the physical phenomena involved, for scoping (u). For an infinitesimal element in homogeneous and
effects of parameter and boundary condition variabili- isotropic, fully saturated porous medium with a con-
ties. Furthermore, due to the versatility of the solution nected pore structure, Biot formulations of the linear
derived in this paper, its application is not limited to constitutive equations for an axisymmetric problem
nuclear waste disposal, and it can be applied in diverse are expressed using incremental total stresses and pore
areas related to energy geotechnics. water pressure as
2 SOLUTION FORMULATION
2.1 Problem description

Consider a cylindrical cavity excavated or drilled
in a deep geological formation, the geological for- where σr , σθ , p and T denote increments of
mation occupies the region of b ≤ r ≤ ∞, and the the total radial stress, total circumferential stress, pore
cavity is backfilled after the ring-shaped liner/casing water pressure and temperature respectively. They are
is installed in the region of a ≤ r ≤ b (Figure 2). The expressed commonly as σr = σr − σr0i , σθ = σθ −
three materials are assumed to be saturated and behave σθ0i , p = p − p0i , T = T − T0i . εv is the volumetric
as poroelastic media. An initially uniform radial stress strain equal to ∂u∂r
+ ur , in which u denotes the radial
σr0i , pore water pressure p0i and temperature T0i are displacement; αsi is the volumetric thermal expansion
assumed, with i = 1, 2, 3 denoting the backfill material, coefficient of the solid skeleton; Kbi is the bulk drained
liner/casing and surrounding formation respectively in modulus of the porous medium, expressed as 2G i (1+νi )
3(1−2νi )
;
the rest of this paper. A heat flux F(t) is applied at Gi is the shear modulus and νi is the Poisson’s ratio;
the interface between the liner/casing and the backfill ξi is Biot’s coefficient and expressed as ξi = 1 − KKbisi
material. The problem is analyzed under plane strain where Ksi is the bulk modulus of the solid particles.
676
Mass conservation of a compressible fluid phase Similarly, continuity at the interface between the
yields the continuity equation liner and backfill gives the following conditions
(r = a)
where ki is the conventional Darcy hydraulic conduc-

tivity of the porous medium; γl is the unit weight of
2
liquid phase; the operator ∇ 2 is expressed as ∂r∂ 2 + 1r ∂r∂
in polar coordinates for an axisymmetric problem; βli
is the combined compressibility of the pore fluid and
solid phase, also known as storage coefficient, defined
as the volume change of the pore fluid per unit volume
of porous medium as a result of a unit increase of pore
pressure and expressed as βli = Knlii + ξiK−nsi
i
, where ni Initial radial stress, pore water pressure and temper-
is the porosity of the porous medium, Kli is the bulk ature for each domain are
modulus of the liquid phase; βTi is volumetric ther-
mal expansion coefficient of the porous medium and
expressed as βTi = ni αli + (ξi − ni )αsi , where αli is the
volumetric thermal expansion coefficient of the liquid
phase.
The thermal balance equation for a heat conduc- 2.4 Solution formulation
tion problem can be uncoupled from hydraulic and
mechanical problems when involving low gradients of The uncoupled thermal balance equation can be eas-
fluid flow in low-permeability media (Giraud et al. ily solved by applying Laplace transform to Eq. (3).
1998, Wang and Papamichos 1999, among others). The solution of temperature in Laplace domain is
Therefore, the thermal diffusion equation can be expressed as:
written as
where qTi2
= ρλi ci i s, s is Laplace operator. f˜ (s) is the
where ρi , ci and λi are the density, the specific heat Laplace transform of a given function f (t) and is
∞
and the thermal conductivity of the porous medium, expressed as f˜ (s) = 0 f (t) e−st dt. I0 and K0 are
respectively. the modified Bessel functions of order 0. The six
In conclusion, the problem is characterized by the unknowns ATi and BTi can be determined considering
following 15 independent parameters: (1) Two geo- temperature boundary condition (4c), bounded tem-
metrical parameters: a and b; (2) Heat flux: F(t); (3) perature at r = 0 and the four continuity conditions
12 constitutive parameters for each material: ρi , ci , λi , (5d), (5de), (6d) and 6(e).
ni , ki , Gi , νi , αsi , αli , Ksi , Kli ξi . Substituting equilibrium equation, ∂σ ∂r
r
+ σr −σr
θ
=0
into the constitutive equation (1) with a new symbol
2.3 Boundary and initial conditions ηi defined as (1−ν i)
results in
1−2νi
The variation of the pore water pressure, stresses and
temperature induced by the heating is negligibly small
in the far field of the porous medium: A partial differential equation regarding the water
pressure can be built by combining Eq. (9) and the
continuity equation for the liquid phase Eq. (2). After
applying Laplace transform and substituting tempera-
ture using the solution already obtained from Eq. (8),
Continuity of the displacement, pore water pressure, water pressure is obtained.
stress, temperature, liquid flux and heat flux exists at General solutions of radial displacement and radial
the interface between the clay and liner (r = b): stress are subsequently solved by substituting solutions
of temperature and pore water pressure into Eqs. (9)
and (1): unknown can be solved by considering the
continuity and boundary conditions, together with the
bounded values of p and u at r =0.
So far, all the solutions are solved in Laplace space.
The solutions in the real time domain can be obtained
using a numerical inversion scheme by Crump’s algo-
rithm (Crump 1976) which has been successfully
applied in previous studies (Chen and Yu 2015, among
others).
677
Table 1. Main parameters used for the Heater test.
Parameters Sand Liner clay
Int. radius a m 0.95

of liner
Ext. radius b m 1.25
of liner
Heat flux W/m Figure 3
Bulk density ρ kg/m3 2000 2278 2000
Specific heat c J /kg/K 1466 948 1434
Thermal λ W /m/K 2.9 2.86 1.53
conductivity
Porosity n 0.394 0.1 0.39
Intrinsic k m2 2.34 4.5 4.5
permeability ×10−11 ×10−18 ×10−19 Figure 3. Heat flux for PRACLAY Heater test.
Shear modulus G MPa 12 17600 120
Biot coefficient ξ 1.0 0.23 1.0
Poisson’s ratio ν 0.25
Solid skeleton αs 1/◦ C 3.0×10−5
vol. therm.
expan. coeff.
Liquid phase αl 1/◦ C 3.4×10−4
vol. therm.
expan. coeff.
Bulk modulus Kl0 MPa 2222
of liquid phase
Bulk modulus Ks MPa 30000
of solid particles
◦
Initial T0i C 16
temperature
Figure 4. Temperature at liner extrados.
3 A PREDICTION FOR PRACLAY 3.3 Verification

HEATER TEST In order to verify the correctness of the present solu-
tion, a comparison is made between results from the
3.1 Main parameters present analytical solution and those from a com-
The reference parameters for PRACLAY Heater test mercial software, COMSOL Multiphysics (COMSOL
are listed in Table 1. With a length of 30 m of the heater, 2008), which is a powerful and widely used finite
there will be a sufficiently long, homogeneous section element code for solving THM coupled problems in
in the middle part of the gallery, therefore the analytical geomechanics. The domain considered in the finite
solution obtained in this study is representative of the element model is a quarter of the geometry shown in
middle section. Figure 2. The size of the modelled region is 1000 m
The heat flux shown in Figure 3 is used as the tall and 1000 m wide with a geometry meshed by 3360
thermal boundary condition at r = a, with a stepwise triangular quadratic elements. The temperature, pore
increasing power during the first 278 days and decreas- pressure and radial stress at the outer boundary of the
ing for the continuation of the test. The heat flux is domain are assumed not to be affected by the heat-
a piecewise constant or linear function for which the ing. Numerical results calculated by both the analytical
explicit expression of the Laplace transform exists, and solution and by COMSOL are compared in Figure 4∼
the discretization of the temporal evolution of the heat Figure 6, and the excellent agreement confirms the
source term (heat flux) is not necessary. correctness of the developed analytical solution.
3.2 Blind prediction

4 SENSITIVITY ANALYSIS FOR PRACLAY
Figure 4 gives the time evolution of temperature at HEATER TEST
lining extrados for 10 years. Figure 5 and Figure 6
give the radial profiles of the temperature change and There are variabilities and uncertainties in material
pore pressure change after heating for 278 days, 3 years parameters, such as thermal conductivities of the
and 10.5 years respectively. three materials, intrinsic permeability of the liner and
After heating for more than 10 years, the ther- the Boom Clay, thermal expansion coefficient of the
mally disturbed radius reaches more than 30 m (see Boom Clay, etc. Investigating the consequences of
Figure 5), and the hydraulic disturbed radius reaches parameter variations within their plausible range is
50 m (See Figure 6). valuable for the interpretation of the in situ measured
678
Figure 5. Radial profiles of temperature increment. Figure 7. Effect of Boom Clay thermal conductivity on
temperature around PRACLAY gallery.
Figure 6. Radial profiles of PWP increment.

Figure 8. Effect of Boom Clay permeability on PWP change
around PRACLAY gallery.
data. In this section, some parameter uncertainties are
discussed and their effects on the THM responses in values in the order of 10−19 m2 (Chen et al., 2014).
PRACLAY Heater test are studied. In addition to the reference value of 4.5×10−19 m2 ,
two variants within the range of the measured values
4.1 Thermal conductivity of Boom Clay in both laboratory and in situ, i.e. 3.0×10−19 m2 and
The Boom Clay is a marine sediment, characterised 6.0×10−19 m2 , are evaluated in this section to test the
by a banded structure essentially created by the impact of KBC on the system response.
alternation of silty and clayey beds. The equivalent Figure 8 presents the radial profiles of pore pres-
thermal conductivity of the Boom Clay used in the sure change around the gallery. A significant impact
base case, λBC = 1.53 W/(mK), is the geometric mean of the considered range of KBC on pore water pressure
of the thermal conductivity in the bedding plane change around the gallery can be observed. The max-
(λBCH = 1.65 W/(mK)) and in the perpendicular direc- imum pore pressure difference reaches about 0.5 MPa
tion (λBCV = 1.31 W/(mK)). These two values come in the gallery between the three cases.
from a medium-scale in situ heating test -ATLAS III 4.3 Young’s modulus of Boom Clay
test- in which in situ temperature measurements were
well fitted with a numerical anisotropic model (Chen The Young’s modulus of the Boom Clay is considered
et al. 2011). 300 MPa by many researchers (Bernier et al. 2002 &
In this section, three values for λBC varying between 2007). Another medium-scale in situ heating test
1.31 and 1.65 (W/(mK)) are tested. With the same in Boom Clay-ATLAS III (Chen et al. 2011) –
heat flux as presented in Figure 3, the induced tem- indicates that higher Young’s moduli for the Boom
perature and its gradient in the Boom Clay becomes Clay in both horizontal (Eh ) and vertical directions
higher when using a lower value for λBC as shown in (Ev ) (Eh = 1400 MPa and Ev = 700 MPa) are capable
Figure 7. The temperature difference among the three of reproducing well the measured pore water pressures.
cases is more than 10◦ C. A higher value of 1350 MPa was also obtained from
the in situ wave velocities of a seismic test performed
4.2 Intrinsic permeability of Boom Clay in HADES URL (Chen et al. 2011). Therefore, in addi-
tion to the reference value of 300 MPa used in the base
Measurements of intrinsic permeability of the Boom case, anther variant, i.e. 1000 MPa is evaluated in this
Clay, KBC , by various techniques give quite consistent section.
679
The radial profiles of the pore water pressure change
around the gallery after heating for 278 days and
10.5 years respectively are given in Figure 10, and very
limited difference among the three cases is observed.
4.5 Saturation degree of pore liquid

in backfilling sand
After the PRACLAY gallery was backfilled with sand,
water was injected into the sand. Although some
measures were taken to keep venting the air in the
backfilled gallery, there still remains some free air
trapped in the water. The compressibility of the liq-
uid significantly depends on the compressibility of the
Figure 9. Effect of Boom Clay Young’s modulus on PWP trapped free air. The compressibility of a pore fluid
change around PRACLAY gallery. under such circumstances is significantly higher than
that of pure water, even with 1% of free air in the fluid.
The relationship between the pore fluid compress-
ibility and saturation degree can be approximated as
Equation 10 when it is close to full saturation (Fredlund
and Rahardjo 1997):
where Kl0 is the bulk modulus of pure water and equal

to 2222 MPa (Table 1), Sl is the saturation degree of the
pore fluid, Pl0 is the absolute pore liquid pressure, and
h is the volumetric solubility coefficient of air which
varies slightly with temperature. For simplicity, con-
Figure 10. Effect of Boom Clay thermal expansion coeffi- stant values for Pl0 = 1 MPa and h = 0.02 are assumed.
cient on PWP change around PRACLAY gallery. According to Equation 10, the bulk modulus of pore
liquid Kl is calculated to be 14.4 MPa, 33.1 MPa and
The profiles of pore pressure change for the two 46.7 MPa respectively for Sl equal to 95%, 99% and
cases show a considerable impact of the Young’s mod- 99.9%. When the air is completely dissolved into the
ulus of the Boom Clay on pore pressure change water with Sl = 100%, the liquid compressibility drops
(see Figure 9).A higher pore pressure will be generated to the water compressibility abruptly, i.e. Kl = Kl0 .
when the clay becomes more rigid, reaching a differ- The results in Figure 9 clearly show the effects of
ence of 1.3 MPa in the gallery between the two cases. sand saturation degree on the pore pressure response.
Even when the saturation degree of the sand backfill
reaches 99.9% before heating starts, the induced pore
pressure in the gallery is still ∼0.3 MPa lower than the
4.4 Thermal expansion coefficient
case with completely saturated sand backfill. In the
of Boom Clay
case of saturation degree lower than 95%, thermal-
When the temperature in the saturated Boom Clay induced pore pressure increase in the near field of
increases, the pore water pressure increases because clay can be higher than that induced in the gallery
the thermal expansion coefficient of pore water is as indicated in Figure 11. But it should be noticed that
much higher than that of the solid skeleton. The sen- in the above analysis, the saturation degree of sand
sitivity of the thermal expansion coefficient of solid is assumed constant in the test. This is not realistic
skeleton on the pore pressure is studied in this section. because the saturation degree of sand will increase
In the reference case, the volumetric thermal expan- with the increase of temperature and pore pressure in
sion coefficient for solid skeleton of the Boom the gallery. Therefore the results in this section are only
Clay, αs , is taken to be 3×10−5◦ C−1 . Various val- for indicative purpose.
ues of the linear thermal expansion coefficient for
the solid skeleton of the Boom Clay ranging from
1.38×10−5◦ C∼4.3×10−5◦ C can be found in the lit- 5 CONCLUSION
erature (Baldi et al 1988, Cui et al. 2000, Monfared et
al. 2012). Therefore, in addition to the base case, two An analytical solution has been derived for thermo-
variants for αs , i.e. 6.0×10−5◦ C−1 and 1.2×10−4◦ C−1 , hydro-mechanical responses around a cylindrical
are evaluated to test the impact of αs on the pore water cavity drilled or excavated in a low-permeability for-
pressure change around the PRACLAY gallery. mation when subject to a time-dependent heat flux.
680
REFERENCES
Baldi, G. Hueckel, T. & Pellegrini, R. 1988. Thermal volume
changes of the mineral–water system in low-porosity clay
soils. Canadian Geotechnical Journal 25(4):807–825.
Bernier, F. Li, X.L. Verstricht, J. Barnichon, J.D., Labiouse, V.
& Bastiaens, W. et al. 2002. CLIPEX: clay instrumentation
programme for the extension of an Underground Research
Laboratory. Luxembourg: Commission of the European
Communities, EUR20619.
Bernier, F. Li, X.L. and Bastiaens, W. 2007. Twenty-five
years’ geotechnical observation and testing in the Tertiary
Boom Clay formation. Géotechnique 57(2): 229–37.
Biot, M.A. 1956. Theory of propagation of elastic waves in a
fluid saturated porous solid. Journal of Acoust. Soc. Am.,
28(2): 168–191.
Figure 11. Effect of sand saturation degree on PWP change
Biot, M.A. & Willis, D.G. 1957. The elastic coefficients of
around PRACLAY gallery.
theory of consolidation. Journal of Applied Mechanics
24:594–601.
Booker, J.R. Savvidou, C. 1984. Consolidation around a
The cavity is considered to be backfilled with a satu- spherical heat source. International Journal of Solids and
rated material after the cavity is supported by casing or Structures 20: 1079–1090.
liner. The solution has been sufficiently verified before Booker, J.R. Savvidou, C. 1985. Consolidation around a
it is applied in a large scale in situ PRACLAY Heater point heat source. International Journal for Numerical and
test. Analytical Methods in Geomechanics 9(2): 173–184.
The solution is used for a predictive calculation of Chen, G. Sillen, X. Verstricht, J. Li, X. 2011. ATLAS III
PRACLAY Heater test. The results show that the dis- in situ heating test in boom clay: Field data, observa-
turbed zone reaches more than 30 m from the axis of tion and interpretation. Computers and Geotechnics 38(5):
the gallery for temperature and about 50 m for pore 683–696.
Chen, G.J. & Yu, L. 2015. Consolidation around a tunnel
water pressure. A parametric study has been carried in a general poroelastic medium under anisotropic initial
out for the PRACLAY Heater test by varying some stress conditions. Computers and Geotechnics 66: 39–52.
material properties within their plausible range. The Chen, G. Maes, T. Vandervoort, F. Sillen, X. Van Marcke, P.
results reveal the following conclusions: Honty, M. 2014. Thermal Impact on Damaged Boom Clay
With the same heat flux, the induced temperature and Opalinus Clay: Permeameter and Isostatic Tests with
and its gradient in the Boom Clay become higher when µCT Scanning. Rock Mechanics and Rock Engineering
using a lower value for λBC . The Boom Clay temper- 47(1): 87–99.
ature difference is more than 10◦ C when λBC varying COMSOL. Multiphysics software and User’s Guide, version
between 1.31 and 1.65 (W/(mK)). 3.5a. COMSOL AB, Stockholm, Sweden; 2008.
Crump, K.S. 1976. Numerical inversion of Laplace trans-
The impact of the permeability of the Boom form using a Fourier series approximation. Journal of
Clay on the pore pressure change around the Association for Computing Machinery 23: 89–96.
PRACLAY gallery is significant when varying the Cui, Y.J. Sultan, N. & Delage, P. 2000. A thermomechanical
intrinsic permeability of the Boom Clay between model for saturated clays. Canadian Geotechnical Journal
3×10−12 m2 ∼6×10−12 m2 with a corresponding pore 37(3): 607–620.
pressure difference of maximum ∼0.5 MPa in the Fredlund, D.G. & Rahardjo, H. 1997. Soil mechanics for
gallery. unsaturated soils. John Wiley.
A considerable impact of theYoung’s modulus of the Giraud, A. & Rousset, G. 1995. Consolidation around a volu-
Boom Clay is expected. A higher pore water pressure mic spherical decaying heat source. Journal of Thermal
Stresses 18(5): 513–536.
is foreseen in the gallery with a higher stiffness of Giraud, A. Homand, F. & Rousset, G. 1998. Thermoelastic
the clay. and thermoplastic response of a double-layer porous space
Very limited effect of the Boom Clay thermal expan- containing a decaying heat source. International Journal
sion coefficient on the pore water pressure change is for Numerical and Analytical Methods in Geomechanics
found. 22(2): 133–149.
The saturation degree of the sand backfill has an Li X.L. Chen G.J., Verstricht, J. Van Marcke, P. Troullinos,
important impact on the pore pressure response even I. 2013. The large scale in-situ PRACLAY Heating and
when the saturation degree reaches 99.9%. Seal tests in URL HADES, Mol, Belgium . Proceedings
The solution is derived for a backfilled cavity in of the ASME 2013 15th international conference on envi-
ronmental remediation and radioactive waste management
which up to three materials are involved and it can (ICEM2013), Sep 8–12, 2013, Brussels, Belgium.
be easily adapted to problems with fewer materials. McTigue, D.F. 1986. Thermoelastic Response of Fluid-
The transient fields at any time are obtained in a semi- Saturated Porous Rock. Journal of Geophysical Research
analytical manner. The great flexibility in geometric Atmospheres 91(B9): 9533–9542.
dimension and the high computational efficiency facil- Monfared, M. Sulem, J. Delage, P. & Mohajerani, M. 2012.
itate its application in preliminary designs of potential On the THM behaviour of a sheared Boom clay sample:
repositories for radioactive waste or other applications Application to the behaviour and sealing properties of the
with coupled THM effects. EDZ. Engineering Geology 124(4): 47–58.
681
NEA(Nuclear Energy Agency), 2008. Moving forward with Selvadurai, A.P.S. 2007. The analytical method in geome-
geological disposal of radioactive waste: an NEA RWMC chanics. Appl. Mech. Rev. 60: 87–106.
collective statement. NEA/RWM (2008) 5/REV2. 12 June Van Marcke, P. Li, X.L. Bastiaens, W. Verstricht, J. Chen, G.J.,
2008, NEA, Paris, Report No. 6433. Leysen, J. Rypens, J. 2013. The design and installation of
Rice, J.R. & Cleary, M.P. 1976. Some basic stress-diffusion the PRACLAY In-Situ Experiment. EURIDICE Report
solutions for fluid saturated elastic porous media with 13–129, Mol, Belgium.
compressible constituents. Rev. Geophys. Space Phys. 14: Wang, Y. & Papamichos, E. 1999. Thermal effects on fluid
227–241. flow and hydraulic fracturing from wellbores and cavi-
Small, J.C. & Booker, J.R. 1986. The behaviour of layered ties in low-permeability formations. International Journal
soil or rock containing a decaying heat source. Interna- for Numerical and Analytical Methods in Geomechanics
tional Journal for Numerical and Analytical Methods in 23(15): 1819–1834.
Geomechanics 10(5): 501–519. Yu, L. Rogiers, B. Gedeon, M. Marivoet, J. De Craen, M.
Savvidou, C. & Booker, J.R. 1989. Consolidation around a Mallants, D. 2013. A critical review of laboratory and
heat source buried deep in a porous thermoelastic medium in-situ hydraulic conductivity measurements for the Boom
with anisotropic flow properties. International Journal Clay in Belgium. Applied Clay Science 75–76: 1–12.
for Numerical and Analytical Methods in Geomechanics
13(1): 75–90.
682
Experimental study of thermal effects on clay rock for deep

disposal of radioactive waste
Chun-Liang Zhang
Gesellschaft für Anlagen- und Reaktorsicherheit (GRS), Braunschweig, Germany
ABSTRACT: Thermal effects on the Callovo-Oxfordian and Opalinus clay rocks for hosting high-level radioac-
tive waste were comprehensively investigated under repository relevant conditions. Various kinds of laboratory
experiments were performed on normally-sized and large hollow cylindrical samples in respects of thermal
expansion and contraction, thermally-induced porewater pressure, temperature influences on deformation and
strength, thermal impacts on swelling, fracture sealing and permeability. Major findings are presented in
this paper.
1 INTRODUCTION sealing and permeability, while the large hollow cylin-

der tests focused on examining thermal impact on the
Clay rocks are widely investigated for deep geological damaged zone around HLW boreholes. Major findings
disposal of high-level radioactive waste (HLW) due to are overviewed in this paper.
their favorable properties such as large homogeneous
rock mass, stable geological structure, extremely low
hydraulic conductivity, swelling and creep capabili- 2 CHARACTERISTICS OF THE CLAYSTONES
ties, self-sealing potential of fractures, and especially
high sorption capacity for retardation of radionu- The Callovo-Oxfordian (COX) and Opalinus (OPA)
clides. In France and Switzerland, for instance, the clay rocks have been highly consolidated during the
potential repositories will be constructed in the individual geological history over hundreds of mil-
Callovo-Oxfordian and Opalinus argillaceous forma- lions of years, leading to low porosities of 14–18%
tions respectively. They will be located at great depths and extremely low permeabilities of 10−20 –10−21 m2
of more than 500 m below the ground surface. HLW (Bock et al., 2010). The studied COX claystone con-
canisters will be disposed in horizontal steel-cased tains 25–55% clay minerals, 20–38% carbonates and
boreholes designed in the French concept (Andra, 20–30% quartz (Andra, 2005), while the OPA clay-
2005), and/or emplaced in horizontal drifts and back- stone has higher clay contents 58–76%, less carbonates
filled with bentonite in the Swiss concept (Nagra, 6–24% and quartz 5–28% (Bossart & Wermeille,
2002). The clay host rocks will be impacted by ther- 2003). The pore sizes mainly range from nanoscale
mal load from heat-emitting HLW. The most concern in between the parallel platelets of the clay parti-
is whether and how the favorable barrier properties cles to micro- and mesoscale between solid parti-
of the clay rocks will be altered during the thermal cles. The claystone matrix consists of particles with
period over several thousands of years under designed strongly adsorbed interlayer water and strongly to
temperatures below 90◦ C. weakly adsorbed water at the external surfaces. Only a
This important issue has been comprehensively small amount of mobile water exists in large pores.
investigated at the GRS geo-laboratory during the last Obviously, the neighboring clay particles are not
decade (Zhang et al., 2004; 2007; 2010; 2013). Various directly connected, but mainly through the bound pore
kinds of laboratory experiments were performed on water, which plays a dominant role in coupled Thermo-
normally-sized and large hollow cylindrical samples Hydro-Mechanical-Chemical (THMC) processes in
under repository relevant conditions: (1) stresses cov- the clay rocks (Horseman et al., 1996).
ering the range from the initial lithostatic state to redis- For the laboratory experiments, a large number of
tributed levels after excavation, (2) hydraulic drained core samples were extracted from the Underground
and undrained boundaries, and (3) heating from ambi- Research Laboratories (URLs) at Bure in France
ent temperature up to 90–120◦ C and a subsequent and Mont-Terri in Switzerland. Normal samples were
cooling phase. The tests on normally-sized samples prepared to sizes of diameter/length = 50/100 and
aimed at characterizing thermal properties, thermal 100/200 mm. Large hollow cylinders were prepared to
expansion and contraction, thermally-induced pore- a size of 280 mm diameter and ∼500 mm length with
water pressure, temperature influences on deformation a central borehole of 100 mm diameter. The samples
and strength, thermal impacts on swelling, fracture were unavoidably disturbed by coring and preparation,
683
Figure 2. Thermal expansion and contraction of a COX
sample during heating and cooling.
Figure 1. Setup and THM conditions for triaxial thermal

tests on normally-sized samples.
causing desaturation and micro-fissures. In order to

minimize effects of the disturbances on the quality of
test results, the samples were resaturated and recon-
solidated before testing under isostatic loads equal to
and even higher than the in situ overburden pressures
up to 15–20 MPa.
3 THERMAL TESTS ON NORMAL SAMPLES
Thermal tests on normally-sized samples were carried

out in triaxial apparatus to examine thermal effects
on deformation and strength, porewater pressure, and
sealing of fractures. Figure 1 shows a testing setup and
the assembly of a sample.The equipment allows a max- Figure 3. Thermal expansion coefficient of a stressed
imum axial and radial load of 50 MPa and a maximum COX sample as a function of temperature.
temperature of 200◦ C. Heating is accomplished using
an electrical heater positioned in the cell. Axial defor-
mation is recorded by a LVDT deformation transducer yields contraction. The contraction at the high temper-
mounted inside the cell between the upper and lower atures above 56◦ C might be caused by mobilization
loading platens, while a circumferential extensometer of the heated porewater into the unsaturated pores and
chain is installed around the sample at its mid-height also by possible leakage of the boundary against such
to measure lateral deformation. The hydraulic system high temperatures, allowing release of the porewater
permits monitoring pore back-pressure at the sample and thus leading to consolidation.
ends and fluid flow through the sample. Based on the data, thermal expansion coefficient
can be obtained. Figure 3 illustrates the volumetric
3.1 Thermal expansion expansion coefficients obtained during heating.
For a water-saturated porous medium, the thermal
Thermal expansion behavior of the claystones was expansion is governed by the expansion of the solid
determined by heating and cooling the samples under grains and the pore water. The thermal expansion
different isostatic stresses and undrained conditions. coefficient can be approached by
Strain gauges were attached on the sample surface for
measuring local axial and radial strain. Figure 2 shows
the strain evolution measured on a COX sample during
a heating/cooling cycle at confining stress of 5 MPa.
The sample with a high initial saturation degree of 96% where φ is the porosity, αs , αw and αm are the lin-
was heated by stepwise increasing temperature from ear expansion coefficient of the solid grains, the
23◦ C to 68◦ C and then cooled down. The measured volumetric expansion coefficient of the porewater
data show that (a) each temperature increase gener- and the porous medium, respectively. Using αs -
ates expansion in all directions; (b) the expansion does value of 2.0 · 10−6◦ C−1 for clay minerals (Noynaert,
not change much with time at constant temperatures 2000) and αw = 3.4·10−4◦ C−1 for pore water, αm =
below 47◦ C and then turns over to a gradual con- 6.1 · 10−5◦ C−1 is obtained for the claystone with a
traction at higher temperatures; and (c) cooling down porosity of φ = 16.5% at 20◦ C.
684
Figure 4. Anisotropy of thermal expansion and contraction
of an OPA sample during a heating-cooling cycle without
confining load.
Additionally, the data also show an increase of

the coefficient with temperature, which may be
approached by a linear relation
Figure 5. Thermally-induced porewater pressure (a) and

where T and To are the actual and the reference tem- deformation (b) in an OPA sample.
perature (20◦ C), respectively, and ω is a factor. Fitting
the data yields ω = 1.0 · 10−6◦ C−1 .
The sedimentary clay rocks commonly exhibit ther- gradually down as unexpected. Similar evolution of
mal anisotropy with bedding structure. Figure 4 show the measured water back-pressures appeared at further
the thermal expansion anisotropy of an OPA sample elevated temperatures of 100 and 120◦ C.The reduction
without any confinement. The expansion coefficient of the pore pressure might be due to some release of
(α⊥ ) in direction perpendicular to the bedding plane the thermally-mobilized porewater from the imperfect
is about one order of magnitude higher than that sealing of the test boundary against such high temper-
(α// ) parallel to the bedding. The significant thermal atures. As consequence of the dissipation of the pore
anisotropy shall reduce when the claystone is subjected pressure, the effective stress increases and compresses
to external stresses, under which thermal expansion of the pore volume.
the micro-fissures mostly oriented along the bedding
planes will be suppressed.
3.3 Creep at high temperatures
Thermal impact on the long-term deformation of
3.2 Thermally-induced porewater pressure
the clay rocks was examined in triaxial creep tests
The difference in thermal expansion between porewa- under various temperatures and stresses. Figure 6
ter and solid grains in saturated claystone is the driving shows an example of the creep tests on highly water-
force for build-up of over pore pressure in case of saturated COX samples at different temperatures of
heating. Because the very low permeabilities of the 28◦ C to 110◦ C. The strain-time curves obtained at
claystones do not allow the expanding porewater to σ1 /σ3 = 15/0.5 MPa show that: (a) each temperature
disperse adequately, heating will result in an increase increase leads to a short-time radial expansion but
in pore pressure. Figure 5 shows an example observed a slight axial compression; (b) the axial, radial, and
on an OPA sample. volumetric strains (ε1 , ε3 , εv ) increases quite linearly
The sample was firstly resaturated with synthetic with time at each elevated temperature below 90◦ C,
porewater and consolidated at a stress of 15 MPa. suggesting no or negligible thermal transient creep;
Undrained heating from 30◦ C to 60◦ C resulted in an (c) at higher temperatures above 90◦ C, the creep slows
increase in pore pressure from 1 MPa to 8 MPa at inlet down; and (d) cooling down results in a short-term
and 7 MPa at outlet. The increase of the pressures con- radial contraction and small axial strain, but almost no
tinued slowly with time. The different responses of the creep any more at each lowered temperature. The creep
water back-pressures (Pin > Pout ) might be caused by acceleration at elevated temperatures up to 90◦ C prob-
the different reservoirs in volume. The second heat- ably results from the reduction of viscosity and friction
ing up to 80◦ C yielded a short pressure rising to of bound water-films between solid particles. As men-
about 11 MPa. After the peak, the pressures dropped tioned before, the testing system was not absolutely
685
Figure 6. Triaxial creep of a COX sample under different
temperatures.
in undrained condition against high temperatures,

allowing release of the porewater. The amount of the
water release was identified up to 2% after testing.
As a consequence of the porewater release, the pore
structure was consolidated under the applied load,
increasing the friction resistance between particles and
hence decelerating the creep.
The creep of a water-saturated clay rock is mainly
governed by slip and rupture of bound water-films
between solid particles. Based on the long-term creep
experiments on the COX and OPA claystones (Zhang
et al., 2007; 2010; 2013; Zhang, 2015), a creep equa-
tion has been derived by modification of the Mitchell’s
law (1976). The stationary creep rate can be expressed
as function of stress and temperature
Figure 7. Stationary shear creep rate of the water-saturated

where ε̇ is the stationary shear creep rate (s−1 ), COX claystone as function of deviatoric stress (a) and
σd the deviatoric stress (MPa), T the absolute temperature (b).
temperature (K), R the universal gas constant
(8.31433·10−3 kJmol−1 K−1 ), Q the apparent activa- on the water-saturated OPA samples at lateral stresses
tion energy (kJmol−1 ), A a parameter in s−1 , and around 3 MPa, a significant reduction of the strength
α a parameter in MPa−1 . For the COX claystone, was observed from 20 MPa at 20◦ C down to 5 MPa at
the parameters are established: A = 2.2·10−4 s−1 , 116◦ C (Zhang et al. 2007). The different conclusions
α = 0.2 MPa−1 , and Q = 45 kJmol−1 . Figure 7 com- drawn from the COX and OPA claystones have to be
pares the quantified mathematical model with the examined in the future.
measured creep rates as function of deviatoric stress at Interesting is that the maximum strength was
T = 25◦ C (a) and temperature at different stress states achieved after drained heating at high temperatures
(b). A reasonable agreement can be found. of 90◦ C and 150◦ C. This is actually attributed to the
release of the thermally-mobilized porewater, which
enhances the friction resistance between particles
3.4 Temperature influence on strength against shearing. This finding implies that the mechan-
Temperature influence on the strength of a claystone is ical stability of the host rock surrounding the HLW
differing from saturated to unsaturated state and from boreholes and drifts will be enhanced during heating
drained to undrained conditions. Figure 8 compares phase.
the peak strength values obtained on the COX samples
during undrained compression at ambient temperature
3.5 Swelling after heating
of 20–30◦ C (black points) and at elevated temperatures
of 40–100◦ C (collar points) as well as the strength The clay rock close to HLW will be heated up to the
after drained heating (upper part). It can be seen that designed maximum temperature of 90◦ C and dried in
the strength curves at the different temperatures are some degrees. The mineralogical components, partic-
close to each other, indicating that the temperature ularly the expansive clay minerals, may be altered. This
influence is insignificant. However, in the other tests may affect the swelling capability of the claystone and
686
Figure 9. Development of swelling pressure in the COX
samples pre-heated and dried at 100–120◦ C.
Figure 8. Comparison of drained and undrained thermal
strengths of the COX claystone.
thus in turn the sealing of fractures in the damaged

zone. This issue was studied by measuring swelling
strain and pressure during wetting samples that had
been pre-heated and dried at 100◦ C and 120◦ C. The
swelling deformation was measured during wetting
the samples in unconstraint conditions. The swelling
pressure was obtained by measuring the reaction of
axial stress under axially-fixed and laterally-free con-
ditions. Details of this test method are given in (Zhang
et al., 2010; 2013). Figure 9 shows the evolution of
swelling pressure obtained on two COX samples pre-
heated at 100 and 120◦ C. It can be found out that
wetting by increasing the humidity to 100% increased
the swelling pressure from the pre-loads of 1–2.5 MPa
up to 4–5 MPa. The free swelling tests showed the
volume expansion of the preheated claystone up to
10–12%. Obviously, the heated claystone still pos-
sesses significant swelling potential, but somewhat
lower than that of the unheated claystone. For instance,
the unheated claystones exhibit higher swelling Figure 10. Water permeability of fractured COX samples
pressures up to 10–12 MPa at COX samples and preheated at 50, 100 and 150◦ C.
4.5–5.5 MPa at OPA samples (Zhang et al., 2013;
Zhang, 2015).
gas dropped immediately by five to seven orders of
magnitude down to 10−17 –10−19 m2 , depending on the
3.6 Thermal impact on fracture sealing
fracture intensity of each sample. This drastic drop in
As shown before, the swelling capacity of the claystone permeability is mainly attributed to the water-induced
becomes relatively lower after heating. Logically the swelling, slaking, and clogging of the fractures. At
self-sealing potential of the heated claystone, which is each load level, the permeability decreased gradu-
largely determined by the swelling, shall be altered. ally with time and tended to constant. The influence
This issue was examined by measuring water per- of the confining stress on the permeability variation
meability of fractured claystones in various thermal was not significant in the testing range. Interesting
conditions (Zhang, 2011; 2013). is that the pre-heating up to 150◦ C did not hinder
Figure 10 shows the water permeability measured the sealing process of the fractures. The final perme-
on three cracked and pre-heated COX samples at 50, ability values determined after 3 years are very low
100 and 150◦ C respectively. They were compressed at at 3·10−20 –7·10−21 m2 , being the same as that of the
relatively low confining stresses of 2 to 3.5 MPa and intact clay rock.
flowed with synthetic porewater through over more Figure 11 shows the similar results obtained on
than 3 years. As soon as the water was supplied, the cracked COX and OPA samples during heating and
high initial permeability of 3·10−12 m2 measured with cooling between 20◦ C and 90◦ C for more than 3 years.
687
Figure 13. Principle of borehole simulation tests with large
hollow cylindrical samples.
Figure 11. Water permeability of fractured COX and OPA
samples during heating and cooling.
decrease more or less with increasing temperature dur-
ing heating and decrease during cooling. Generally
speaking, the self-sealing of fractures in the claystones
is not affected by the applied thermal loads.
4 LARGE HOLLOW CYLINDER TESTS
In order to investigate thermal impact on the EDZ

around HLW boreholes, a series of large-scale heating
tests were conducted on big hollow claystone cylinders
in a triaxial apparatus, which permits a maximum
lateral stress of 50 MPa, axial stress up to 75 MPa,
temperature up to 150◦ C, and fluid pressure up to
15 MPa. The big hollow samples were prepared to
an outer diameter of 280 mm and lengths between
460 mm and 530 mm with axially-drilled central bore-
holes of 100 mm diameter. Figure 13 shows the test
layout. The heater-packer in the borehole allows sim-
Figure 12. Water permeability of fractured claystones as a ulating the heat output from HLW up to a maximum
function of temperature. temperature of 150◦ C and a maximum back pressure of
15 MPa. Thermal load can also be applied by means of
an outer heater mounted around the cell. Axial strain is
The water-enhanced sealing of fractures is more signif- measured by a LVDT deformation transducer mounted
icant during the first stage at 20◦ C. The permeability outside the cell, while external radial strain is recorded
decreased with time from 1·10−15 to 3·10−18 m2 at using two circumferential extensometers installed at
OPA sample and from 1·10−17 to 5·10−19 m2 at COX the middle and the lower position of ¼ sample length.
sample, respectively. The permeability reduction rate Borehole convergence is measured by the oil volume
was less affected by the temperature increase up to in the inner packer. A hydraulic system is connected
60◦ C. Further heating up to 90◦ C and also cooling with the inlet and outlet for measurement of water or
down back to 60◦ C had no or only little effect on the gas flow through the sample.
permeability reached before. Further cooling down The large hollow cylinder tests were performed by
to 20◦ C, however, caused a further reduction of the simulating the borehole conditions such as excavation,
permeability down to 3·10−19 at OPA sample and backfill support, fluid flow, heating and cooling. Fig.
1·10−19 m2 at COX sample respectively. 14 presents results of a borehole simulation test in
The water permeability values obtained after sta- terms of applied stress and temperature (a), resulting
bilization of each temperature stage are depicted in convergence (b), permeability change (c), and water
Fig. 12 as a function of temperature. It is obvious flow (d) during a heating/cooling cycle.
that the permeabilities of the fractured claystones The test started from an isotropic stress of
are slightly influenced by heating and cooling. They 15 MPa. The borehole excavation (II) was simulated
688
observation implies that in the much longer disposal
boreholes of tens of meters, fractures in the EDZ may
not be connected to form hydraulic pathways along
the boreholes. An intensive damage was generated
by increasing the external stress up to 20 to 24 MPa
(III), which yielded a drastic increase in gas perme-
ability from 10−21 to 10−15 m2 (Fig. 14c). However,
the following injection of synthetic porewater (IV) led
to sealing of the fractures. The measured water per-
meability of 1·10−18 m2 is three orders of magnitude
lower than the gas permeability measured before. The
subsequent heating (V) from 30◦ C to 75◦ C resulted in
a significant rock deformation towards the borehole
and thus a large convergence. At the increased temper-
ature, the convergent deformation continued with time
and the water permeability increased only slightly to
3·10−18 m2 (Fig. 14d). The cooling phase (VI) caused
a divergence of the borehole but a small displacement
of the outer surface towards the borehole due to the
thermal contraction of the material. At the lowered
temperature, the borehole tended to close up with time
due to the rock creep and the permeability decreased
slightly down to 2·10−18 m2 . The small variation of
the permeability during heating and cooling is consis-
tent with the observations made on the small samples
(cf. Figs. 11–12).
5 SUMMARY AND CONCLUSIONS
Thermal impact on the barrier properties and the

integrity of a clay host rock is the key concern for deep
geological disposal of HLW. This issue was compre-
hensively investigated with the thermal experiments
on the COX and OPA claystones under the prevail-
ing repository conditions. Significant responses of the
claystones to thermal loading were observed, includ-
ing thermal expansion and contraction, thermally-
induced porewater pressure, temperature influences
on deformation and strength, thermal impacts on
swelling, fracture sealing and permeability. These are
strongly determined by the inherent properties (min-
erals, porosity, water saturation, anisotropy etc.) of
the material itself and the external conditions (stress,
drainage and temperature).
In the water-saturated claystones, the thermal
expansion is predominantly controlled by the porewa-
ter because of its much higher expansion coefficient
compared to that of the solid grains. Heating yields
expansion and mobilization of the porewater. Under
undrained conditions, the water expansion results
in high porewater pressures decreasing the effective
stress. Because of the thermally-induced reduction of
Figure 14. Results of a borehole simulation test with initial
state (I), excavation (II), EDZ generation (III), water flow
the inherent cohesion and friction resistance between
(IV), heating (V) and cooling (VI). particles, the claystone becomes more ductile at
elevated temperatures and the creep is accelerated.
In contrast, drained heating drives the thermally-
by reducing the borehole pressure from 15 MPa down mobilized porewater out of the pore space. Under
to 1 MPa. As unexpected, no increase in permeability external load, the release of the bound porewater
was observed (Fig. 14c), indicating no build-up of any leads to collapse of the pore structure and thus con-
flow pathway through the sample of half a meter. This solidation of the porous medium. Even under high
689
deviatoric stresses, the consolidation is dominating Heitzmann, P. & Tripet, J.-P. (ed.): Mont Terri Project –
during the pre-failure stage. With heating and drying, Geology, Paeohydrology and Stress Field of the Mont
the contacts between particles become increasingly of Terri Region – Reports of Federal Office for Water and
the solid-to-solid type, so that the friction resistance Geology (FOWG), Geology Series 4, 65–92.
Horseman S.T., Higgo J.J.W., Alexander J., Harrington J.F.
between particles increases, enhancing the stiffness 1996: Water, Gas and Solute Movement through Argilla-
and strength. ceous Media. Nuclear Energy Agency Report CC-96/1,
Another important finding is that the high seal- OECD, Paris.
ing capability of the claystones is not affected by the Mitchell, J.K. 1976: Fundamentals of Soil Behavior,
applied thermal loads. The fractures in the claystones University of California, Berkeley, USA.
pre-heated up to 100–150◦ C and under low stresses of Nagra 2002: Project Opalinus Clay, Models, Codes and
2–3.5 MPa can be resealed to very low water perme- Data for Safety Assessment – Demonstration of disposal
abilities of 10−19 to 10−21 m2 within months to years. feasibility for spent fuel, vitrified high-level waste and
It is evident that the favorable barrier properties of long-lived intermediate-level waste. 2002.
Noynaert, L. (Ed.) 2000: Heat and radiation effects on the
the clay rocks will not be altered during the thermal near field of a HLW or spent fuel repository in a clay for-
loading from HLW. mation (CERBERUS Project). EUR 19125EN, Contract
No F14W-CT95-0008.
Zhang, C.L., Rothfuchs, T., Moog, H., Dittrich, J., Müller,
ACKNOWLEDGEMENTS J. 2004: Thermo-Hydro-Mechanical and Geochemical
Behaviour of the Callovo-Oxfordian Argillite and the
The author gratefully acknowledges the funding by Opalinus Clay. Project Report, GRS-202.
the German Federal Ministry of Economics and Zhang, C.L., Rothfuchs, T., Jockwer, N., Wieczorek, K.,
Technology (BMWi) and by the European Com- Dittrich, J., Müller, J., Hartwig, L., Komischke, M. 2007:
mission (EC) in the framework of several R&D Thermal Effects on the Opalinus Clay – A Joint Heating
Experiment of ANDRA and GRS at the Mont Terri URL
projects (HE-D, THM-TON, TIMODAZ). The sup-
(HE-D Project), Final Report, GRS-224.
port from Andra for providing the core samples Zhang, C.L., Czaikowski O., Rothfuchs, T. 2010: Thermo-
and for the fruitful discussions is also gratefully Hydro-Mechanical Behaviour of the Callovo-Oxfordian
acknowledged. Clay Rock – Final report of the TIMODAZ/BURE project,
GRS-266.
Zhang C.L. 2011: Experimental Evidence for Self-sealing
REFERENCES of Fractures in Claystone, Physics and Chemistry of the
Earth, Vol. 36, 2011, 1972–1980.
Andra 2005: DOSSIER 2005, Synthesis – Evaluation of the Zhang C.L. 2013: Sealing of Fractures in Claystone, Jour-
feasibility of a geological repository in an argillaceous nal of Rock Mechanics and Geotechnical Engineering 5,
formation. 2013, 214–220.
Bock, H., Dehandschutter, B., Martin, C.D., Mazurek, M., Zhang C.L., Czaikowski O., Rothfuchs T., Wieczorek,
Haller, A.D., Skoczylas, F., Davy, C. 2010: Self-Sealing K. 2013: Thermo-Hydro-Mechanical Processes in the
of Fractures in Argillaceous Formations in the Context of Nearfield around a HLW Repository in Argillaceous
Geological Disposal of Radioactive Waste – Review and Formations, Project Report THM-TON, GRS-312, 2013.
Synthesis. Clay Club Report, OECD, NEA No. 6184. Zhang, C.L. 2015: Deformation of Clay Rock under THM
Bossart, P. and Wermeille, S. 2003: The Stress Field Conditions, Geomechanics and Tunnelling 8 (2015), No.
in the Mont Terri Region – Data Compilation. In: 5, 426–435.
690
Temperature impact on the creep behaviour of compacted illitic clay
N. Jarad, O. Cuisinier & F. Masrouri

LEMTA – UMR 7563 CNRS/Université de Lorraine, Vandœuvre-lés-Nancy, France
ABSTRACT: In some engineering applications such as geothermal piles, nuclear waste storages, etc clayey
soils could be exposed to thermal cycles. These temperature changes could affect dramatically the mechanical
behaviour of these soils. Especially in the long term the creep deformation could be completely modified. Creep
deformation which contributes to the volume changes of clay is related to the viscous property of the clay
skeleton. In this context, the paper aims to investigate the temperature effect on the creep behavior of saturated
compacted illitic clay. Temperature controlled oedometric cells have been developed to perform Constant Rate
of Strain (CRS) consolidation tests under different strain rates (0.002%/min to 0.02%/min) within a temperature
range of 20 to 70◦ C. Results indicated that the compression and swelling indices could be considered independent
of temperature and strain rate. The preconsolidation pressure, and creep index, decreased, and increased with
temperature increase respectively. The hydraulic conductivity increased with temperature increase, while the
intrinsic permeability remained independent of temperature.
1 INTRODUCTION relationship (e.g. Boudali et al. 1994, and Marques

et al. 2004). Boudali et al. (1994) carried out differ-
The issue of temperature impact on the hydraulic ent constant rate of strain (CRS) consolidation tests
and mechanical behavior of clayey soils including at different temperatures to investigate the viscous
creep deformation is a major issue in geotechnical behavior of natural clay. They found that the strain
engineering. rate-preconsolidation pressure relationship is indepen-
A better understanding of this phenomenon will dent of temperature. A similar behavior of natural
help to reduce the undesired effects of temperature clay was also observed by Marques et al. (2004).
changes and to a better selection of design meth- Boudali et al. (1994) and Marques et al. (2004) found
ods (Sultan et al. 2002, Cekerevac & Laloui 2004, approximately parallel slopes of logarithmic strain
Watabe et al. 2012, Kholghifard et al. 2014). For rate-preconsolidation pressure relationship at differ-
example, compacted clayey soils are used as engi- ent temperatures. These studies were however focused
neered barriers for the disposal of high and long-life on the natural and structured clays only.
radioactive wastes at great depth, and therefore they In this context, the main objective of this study
are usually subjected to cyclic changes of temperature. was to investigate the influence of temperature on the
Among the mechanical properties of the compacted hydraulic and mechanical behavior of a compacted
clay, creep may contribute to the settlement changes clay soil including the long term consolidation behav-
of the structure in the long term. ior (creep). To characterize the behavior in terms of
Although temperature variations could highly affect strain rate effect, constant rate of strain consolida-
the creep behavior, there are still very few studies tion tests with different strain rates were performed
on this subject. Gupta (1960) performed creep tests within temperature range between 20 and 70◦ C using
at different temperatures and stated that the creep temperature controlled oedometric cells. A saturated
index increases with temperature increase. Towhata compacted illitc soil was used to assess the thermo-
(1993) raised the room temperature up to 90◦ C dur- hydro-mechanical-behavior of compacted soils.
ing secondary consolidation phase, and observed a
volume contraction of the sample. Burghignoli et al.
(1992) indicated that the thermal loading accelerates
the mechanical creep phenomena. Green (1969) noted 2 MATERIALS AND METHODS
a higher value of creep index at higher temperature
values, but the influence of temperature on the creep 2.1 Material
was dependent on the effective stress level. In other Saturated compacted illitic clay was studied. It con-
words, the creep was more affected by temperature at sists of 77% illite, 10% kaolinite, 12% calcite and
lower effective stress levels. traces of quartz and feldspar. Illite is a non-expanding
A few researchers studied the influence of tem- clay material which is common in many argillaceous
perature on the strain rate-preconsolidation pressure sediments (Gaudette et al. 1964, Srodon et al. 1986,
691
Lynch 1997, Adamis et al. 2005). Atterberg limits was installed into the load frame after assembly. An
(AFNOR 1993) and Proctor curve (AFNOR 1999) external load was applied to the specimen using the
for studied clay was determined (Table 1). The maxi- load frame to generate a load with a predefined con-
mum dry density of illitic clay was 1.43 Mg/m3 at the stant rate of strain on the specimen, and the vertical
optimum water content of 31.3%. A laser diffraction deformation of the specimen was measured using a
particle size analyser (Malvern Mastersizer 2000®) displacement sensor.
was used to determine the particle size distributions The temperature device can operate in a tempera-
of the soil. Studied clay contains about 85% of clay ture range from −40◦ to 80◦ C. It supplies a thermal
particles (smaller than 0.002 mm) and 15% of silt par- liquid to the oedometric cell through insulated tubes
ticles (smaller than 0.02 mm).The classification of that for heating or cooling the sample. The thermal liquid
clay according to the French standard for soil classifi- circulates through these tubes from the device to the
cation (GTR 2000) is A3, and classified as a fat clay, cell and vice versa. In addition, the thermal liquid cir-
CH, in accordance with the Unified Soil Classification culates around the sample through a spiral tube for
System (Standard ASTM 2006). purpose of heating or cooling.
Water injector was used as a water supply to saturate
the sample by applying pore water pressure through
2.2 Experimental device the sample base. During testing, all sensors data were
stored in the data logger, and then were displayed on
A schematic diagram for the modified temperature
the PC.
controlled oedometric cell is shown in the Figure 1.
The whole system includes load frame, oedometric
cell, temperature device, and water injector. 2.3 Sample preparation
The oedometric cell was designed to hold high tem-
The powder of illitic clay soil was mixed with distilled
peratures up to 90◦ C. The sample was inserted in the
water at room temperature to adjust the water content
pressure chamber between two ceramic porous stones
corresponding to the value of the optimum water con-
with using filter papers on the top and bottom of the
tent (at maximum dry density). The prepared slurries
sample. The cell base contains drainage system for sat-
were then kept into closed bags for 24 hours at least to
uration, and the pore water pressure can be measured
obtain the homogeneity. Samples were then statically
through a cell pressure transducer. The oedometric cell
compacted in a vertical dimension into a rigid mold in
one layer to form specimens with 71.4 mm diameter
Table 1. Characteristics of illitic clay.
and 20 mm height. The weights of the compacted spec-
Setting Value
imens are representative to the maximum dry density
at optimum water content. The average compaction
Liquid limit (%) 65 stress for the material is 467 kPa. At this condition,
Plastic limit (%) 34 the initial void ratio of the compacted samples is 0.85.
Plasticity index (%) 31
Optimum moisture content (%) 31.3
Maximum dry density (Mg/m3 ) 1.43
2.4 Experimental procedures
Specific gravity 2.65 Before saturation starting, the sample was loaded ver-
tically under constant vertical stress of 10 kPa. Both
Figure 1. a. Schematic diagram of the apparatus. b. oedometric cell detail.
692
Table 2. Experimental program. Where Ht = sample height at specific time, t; Ho =
Initial sample height, and ε̇t = the strain rate at specific
Strain rate time, t, and can be calculated as following (ASTM
Temperature 2008):
◦
Setting C %/min
Value 20 0.002 0.01 0.02

47.6 0.002 0.01 0.02
69.2 0.002 0.01 0.02
porous stones at the bottom and top were saturated, The effect of temperature on the hydraulic conductiv-
and the air was flushed from the base and top of the ity of clays can be understood better in terms of the
cell. One of the base drainage was closed off for pore intrinsic permeability, K, as following:
water pressure reading, while the top drainage was per-
mitted. A pore water pressure of 10 kPa was applied
through the porous element of the base to saturate the
sample. After fully saturation with deaired water, the
sample was heated to the desired temperature, and
when the stabilization was reached after 48 hrs, the Where µ = pore water viscosity at the tested temper-
constant rate of strain (CRS) consolidation test was ature, T.
carried out on the sample according to ASTM (2008) The value of free pure water viscosity change with
recommendations. temperature (T) as the following equation:
In CRS test, the sample is loaded at a constant rate
of vertical deformation. Drainage is permitted only at
the top surface of the specimen.
During testing, the excess pore water pressure is
generated through the sample with a variation from
the maximum at the base until zero at the top sur- 3 RESULTS
face. The reaction force, axial deformation, and pore
water pressure at the base have been measured dur- 3.1 Temperature and strain rate influence on the
ing the testing at certain time interval. The sample was stress strain curve
loaded from 10 kPa until reaching, the maximum stress Figure 2 shows the variation of strain with vertical
of 5120 kPa. effective stress under different temperatures (20, 47.6,
69.2◦ C) for CRS tests at .01%/min strain rate. At least,
2.5 Experimental program there is no large difference in stress strain behavior
Table 2 shows the experimental program to study the between temperatures 20◦ C and 50◦ C. However, at
influence of temperature on the illitic soil. specific effective stress, higher values of strain were
observed under temperature of 69◦ C compared with
2.6 Experimental theory other temperatures.
From another perspective, figures 3-5 show the
Constant rate of strain oedometric test (CRS) was
variation of strain with vertical effective stress under
used in this study. By considering a parabolic distribu-
different strain rates at a certain temperature. For
tion of pore water pressure through the sample during
example at 20◦ C, the values of strain at strain rate
CRS test, the effective stress can be evaluated as the
of 0.02%/min are lower than others at specific effec-
following (Wissa et al. 1971):
tive stress at this temperature due to the pore pressure
increase with strain rate increase.
Results showed that the strain-log σ’ curves slightly
changed with temperature. A similar behavior for clay
was reported by Kholghifard et al. (2014). They men-
Where σ = measured total stress during the test, and tioned that for dense samples, the soil response is
u = measured pore water pressure at the base of the controlled by the compression caused by the applied
specimen during the test. load. In contrast, lower values of void ratio were
The measured parameters during isothermal CRS obtained at higher temperatures using oedometer tests
consolidation tests were employed to determine the by Towhata et al. (1993), Boudali et al. (1994), &
hydraulic conductivity kt of the studied clay soil at Marques et al. (2004). However, Burghignoli et al.
specific temperature and strain rate as indirect method (2000) indicated that the void ratio changes due to
of measurement according to the following equation rearrangement of particles, partially from the temper-
(ASTM 2008): ature variation, and also influenced by other factors
like stress history, thermal history, over consolidation
ratio (OCR), time duration between the application of
the last mechanical load and the start of heating, and
the duration of the heating phases.
693
Figure 3. Influence of strain rate on the effective
stress-strain curve and pore water pressure at 20◦ C.
Figure 2. Influence of temperature on the effective
stress-strain curve and pore water pressure.
Values of compression index which represent the

slope of the compression line converge to the same
value at different temperatures. The values for all tests
range between 0.191 and 0.217 at temperatures of
20, 47.6, and 69.2◦ C. Also, the compression lines are
almost parallel at different strain rates of 0.002, 0.01,
and 0.02%/min at a given temperature. The values of
swelling index which were obtained from the slopes
of the first part of the consolidation curves at different
temperatures varied between 0.0051 and 0.016. Fig-
ure 6 shows the variation of compression and swelling
indexes with temperature and strain rate for 9 tests on
the tested clay.
Accordingly, compression index for studied clay
could be considered independent of temperature and
strain rate. Also, the extent of change in the val-
ues for all tests indicates that the swelling index
remains not strongly dependent on temperature and
strain rate.
3.2 Temperature and strain rate effect on the pore

water pressure
Figures 2-5 show the variation of excess pore pres-
sure with different temperatures and strain rates. For Figure 4. Influence of strain rate on the effective
example at 0.01%/min strain rate, the maximum excess stress-strain curve and pore water pressure at 47.6◦ C.
pore pressure generated at 47.6◦ C is 1.53 times smaller
than at room temperature, while the pore pressure gen- At a certain temperature, results showed that the
erated at 69.2◦ C is 2.5 times smaller than at room excess pore pressure decreases as strain rate decreases
temperature. where it is smallest at low strain rate of 0.002%/min.
694
Figure 7. Influence of temperature and strain rate on the
preconsolidation pressure obtained from 9 CRS tests.
pressure which was measured at the base of the spec-

imen showed a strong dependency on the temperature
and strain rate for the tested illitic clay. The generated
excess pore pressure relative to the total stress is less
than 15% which is consistent with the ASTM (2008)
recommendations.
3.3 Temperature and strain rate influence on the

Figure 5. Influence of strain rate on the effective preconsolidation pressure
stress-strain curve and pore water pressure at 69.2◦ C.
The preconsolidation pressure relationship with tem-
perature was experimentally identified by running
different isothermal CRS tests at different tem pera-
tures. The determined preconsolidation pressure for
illitic clay for all tests at different temperatures (20,
47.6, and 69.2◦ C) was decreased with temperature
increase as a linear representation. In addition, pre-
consolidation pressure was found higher at high strain
rate (0.02%/min). Figure 7 shows the variation range
of preconsolidation pressure with temperature at dif-
ferent strain rates which was determined from 9 CRS
tests on illitic clay.
The main reason for the the elastic domain decrease
with temperature increase is the reduced viscosity
at high temperatures. This reduction in viscosity
increases the number of mineral-to-mineral contacts
Figure 6. Influence of temperature and strain rate on the instead of mineral-to-water-to-mineral contacts, and
compression and swelling indices. this causes the plastic deformation (Shariatmadari &
Saeidijam, 2012). In fact, water acts as an elastic mate-
rial between two clay particles. A similar result of
In general, at a certain strain rate, the lower the tem- negative temperature dependency was found by Sultan
perature, the higher the generated pore pressure at et al. (2002), and Moritz (1995). Conversely, Mon et
the base of the specimen and for high temperatures, the al. (2013) concluded that heating causes increasing in
generated excess pore pressure decreased. Indeed, the the preconsolidation pressure.
occurrence of high permeability at high temperatures
beside the drainage from one side limits the buildup
of pore pressure in the specimen. At specific tempera-
3.4 Temperature and strain rate influence on the
ture, the higher the strain rate, the higher the generated
permeability
pore pressure at the base of the specimen. Therefore,
at low strain rates, the generated pore water pressure is Table 3 shows the variation of hydraulic conductiv-
negligeable, and cannot be measured at very low strain ity and the intrinsic permeability with temperature
rates. Furthermore, it is noted that at combined high for tested clay at void ratio of 0.65 as an example.
temperature and low strain rate, the generated excess The hydraulic conductivity at 69.2◦ C was larger than
pore pressure is smaller. The generated excess pore the hydraulic conductivity at 20◦ C by a factor of 3.
695
Table 3. Influence of temperature on the permeability (k)
and intrinsic permeability (K) for e = 0.65.
Tempe- Strain rate

rature %/min
0,002 0,01 0,02
k K k K k K
m/s m2 m/s m2 m/s m2
◦
C E-11 E-19 E-11 E-19 E-11 E-19
20 1,31 1,33 1,59 1,61 1,62 1,64

47.6 2,08 1,27 1,44 87,9 4,29 2,62
69.2 3,76 1,64 5,93 2,58 6,31 2,75
Figure 9. Influence of temperature on the creep index of
studied clay.
20, 47.6, and 69.2◦ C, respectively. The compression

index, cc , was obtained from the slope of the void
ratio versus logarithm of effective stress curve (e-log)
at the end of primary consolidation (EOP) at a spe-
cific temperature. According to the values of cα /cc
and cc , creep index, cα , increased with temperature
increase. Figure 9 shows the variation range of creep
index with temperature. The estimated values of creep
index were 0.0165, 0.012, and 0.0223 corresponding
to temperatures of 20, 47.6, and 69.2◦ C, respectively.
The amount or rate of creep increases with temper-
ature increase because of the reduction in the apparent
viscosity of the contacts between particles at higher
temperatures (Gupta 2013). The results showed that
Figure 8. Variation of preconsolidation pressure with strain the value of cα /cc of illitic clay varied with temper-
rate at different temperatures.
ature between 0.039 and 0.125. According to Mesri
et al. (1999), the highest value of cα /cc for geotechni-
However, at a specific temperature, the hydraulic con- cal materials rarely exceeds 0.07, but it is significantly
ductivity increases as strain rate increases for tested less than 0.21. However, Boudali et al. (1994) & Mar-
clay. ques et al. (2004) found a parallel slopes at different
According to the obtained results of the studied temperatures.
material, the hydraulic conductivity is temperature
dependent, while the variation of the intrinsic per-
meability with temperature is negligible. There is no
important change in the unit weight of water with 4 CONCLUSIONS
temperature compared with the viscosity.
The results of the experimental study performed on
saturated compacted illitic clay using a temperature
3.5 Temperature influence on the creep controlled oedometric cell to investigate the tem-
per ature influence on the hydraulic and mechanical
The cα /cc concept developed by Mesri & Godlewski
behavior of clay including the creep deformation can
(1977) was used in this study to determine the creep
be concluded as following:
index value, cα . At a specific temperature, the value
of cα /cc together with the compression index, cc , were • Compression index (cc ) and swelling index (cs )
used to determine the value of creep index.The value of variation with temperature and strain rate can be
cα /cc was obtained from the slope (α) of the preconsol- considered to be negligeable.
idation pressure–strain rate logarithmic curve at differ- • The preconsolidation pressure (pc ) is temperature
ent strain rates (0.002%/min, 0.01%/min, 0.02%/min) and strain rate dependent. It decreases with tem-
at a specific temperature. Figure 8 shows the variation perature increase, while it increases with strain rate
range of preconsolidation pressure with strain rate at increase.
different temperatures for the studied clay. The pre- • The creep index (cα ) increases with temperature
consolidation pressure values were estimated among increase.
the variation to determine the cα /cc values. The val- • The hydraulic conductivity (k) increases with heat-
ues of cα /cc varied with temperature. These estimated ing and strain rate, while the intrinsic permeability
values were 0.081, 0.056, and 0.106 at temperatures (K) can be considered independent of temperature.
696
The hydraulic conductivity at 69.2◦ C was larger than Kholghifard, M., Ahmad, K., Ali, N., Kassim, A., Kalate-
the hydraulic conductivity at 20◦ C by a factor of 3. hiari, R. & Babakanpour, F. 2014. Temperature effect on
• Pore water pressure (u) decreases with temperature compression and collapsibility of residual granitic soil.
increase and strain rate decrease during CRS tests. GRAÐEVINAR 66(3): 1–10.
Lynch, F.L. 1997. Frio shale mineralogy and the stoichiom-
• Stress strain behavior of dense samples changes
etry of the smectite-to-illite reaction: the most important
slightly with temperature, while it is strain rate reaction in clastic sedimentary diagenesis. Clays and Clay
dependent. Strain value increases with strain rate Minerals 45(5):618-631.
decrease at specific effective stress. Mon, E.E., Hamamoto, K., Kawamoto, S., Komatsu, T. &
Møldrup, P. 2013. Temperature effects on geotechnical
properties of kaolin clay: simultaneous measurements of
REFERENCES consolidation characteristics, shear stiffness, and perme-
ability using a modified oedometer. GSTF International
Adamis, Z. & Richard, W. 2005. Bentonite, kaolin, and Journal of Geological Sciences (JGS) 1(1).
selected clay minerals. Geneva: International programme Marques, M. E., Leroueil, S. & Almeida, M. S. 2004. Viscous
on chemical safety. behaviour of St Roch-de-l’Achigan clay, Quebec, Can.
AFNOR. 1993. NF P94-051 Sols: reconnaissance et essais; Geotech. J. 41: 25–38.
Détermination des limites d’Atterberg – Limite de liquid- Mesri, G. & Godlewski, P.M. 1977. Time- and stress-
ité à la coupelle- Limite de plasticité au rouleau [Soil: compressibility interrelationship. Journal of the Geotech-
Inverstigation and testing. Determination of Atterberg’s li nical Engineering, ASCE 103: 417–430.
its. Liquid limit test using cassagrande apparatus. Plas- Mesri, G., Feng, T.W. & Shahien, M. 1999. coefficient of
tic limit test on rolled thread] (p. 15). Paris: Association consolidation by inflection point method. Journal of the
Française de Normalisation. Geotechnical and geoenvironmental Engineering: 716–
AFNOR. 1999. NF P 94-093 Sols: Reconnaissance et 718.
essais Détermination des références de compactage d Moritz, L. 1995. Geotechnical properties of clay at ele-
‘un matériau. Essai Proctor Normal-Essai Proctor Mod- vated temperatures. Swedish Geotechnical Institute (SGI).
ifié [Soils: Investigation and testing. Determination of Linköping, Sweden.
the compaction characteristics of a soil. Standard Proc- Sultan, N., Delage, P. & Cui,Y.J. 2002. Temperature effects on
tor test. Modified Proctor test] (p. 18). Paris: Association the volume change behaviour of Boom clay, Engineering
Française de Normalisation. Geology 64(2-3): 135–145.
Boudali, M., Leroueil, S. & Srinivasa Murthy, B.R. 1994. Srodon, J., Morgan, D.J., Eslinger, E.V., Eberl, D. D. &
Viscous behaviour of natural clays. In: Proceedings of Karlinger , M.A. 1986. Chemistry of illite/smectite and
the 13th International Conference on Soil Mechanics and end-member illite. Clays and Clay Minerals 34: 368–378.
Foundation Engineering1: 411–416. New Delhi, India. Standard ASTM. 2006. D2487 Standard practice for clas-
Burghignoli, A., Desideri, A. & Miliziano, S. 1992. Deforma- sification of soils for engineering purposes (Unified Soil
bility of clays under non isothermal conditions. Rivista Classification System) (p. 12). West Conshohocken, PA:
italiana di Geotecnica 92(4): 227–236. ASTM International.
Burghignoli, A., Desideri, A. & Milizia no, S. 2000. A laborat Standard ASTM. 2008. D4186-06 Standard test method for
ry study on the thermomechanical behaviour of clayey one-dimensional consolidation properties of soils using
soils. Can. Geotech. J. 37: 764–780. controlled-strain loading (American Society of Testing
Crawford, C. B. 1967. Interpretation of the consolidation test. Materials). West Conshohocken, PA: ASTM Interna-
J. Soil Mech. Found. Div., ASCE: 90: 87–102. tional.
Cekerevac, C. & Laloui, L. 2004. Experimental study of ther- Shariatmadari, N. & Saeidijam, S. 2012. The effect of thermal
mal effects on the mechanical behaviour of a clay. Inter- history on thermo-mechanical behavior of bentonite-sand
national Journal for Numerical and Analytical Methods mixture. International Journal of Civil Engineering 10(2):
in Geomechanics 28(3): 209–228. 162–167.
Gaudette, H.E., Eades, J.L. & Grim, R.E. 1966. The nature Towhata, I., Kuntiwattanakul, P. & Seko, I. 1993. Volume
of illite. Clays and Clay Minerals 13:33–48. change of clays induced by heating as observed in consol-
GTR. 2000. Réalisation des remblais et des couches de idation tests. Soils and Foundations 33(4): 170–183.
forme (p. 102). Paris: Laboratoire Central des Ponts et Wissa,A.E.Z., Christian, J.T., Davis, E.H. & Heiberg, S. 1971.
Chaussées. (In French). Consolidation at constant rate of strain. Journal of Soil
Green, W.J. 1969. The influence of several factors on the Mechanics and Foundations Division, ASCE 97(SM10):
rate of secondary compression of soil. Master thesis, The 1393–1413.
Missouri University of Science and Technology, Rolla, Watabe, Y., Udaka, K., Nakatani, Y. & Leroueil, S. 2012.
Missouri, USA. Long term consolidation behavior interpreted with iso-
Gupta, B. 1960. Creep of saturated soil at different tempera- tache concept for worldwide clays. Soils & Foundations
tures. Master Thesis, The University of British Columbia, 52(3): 449–464.
Canada.
Gupta, V. K. 2013. An Experimental Study on Secondary
Consolidation of Organic Clay. Master Thesis, Jadavpur
University, Kolkata, India.
697
Impact of temperature variation on pressuremeter test parameters

in compacted soils
A. Boukelia & H. Eslami

LEMTA (CNRS, UMR 7563), Université de Lorraine, Vandœuvre-lès-Nancy-France
ESITC de Metz, Metz-France
S. Rosin-Paumier & F. Masrouri

LEMTA (CNRS, UMR 7563), Université de Lorraine, Vandœuvre-lès-Nancy-France
ABSTRACT: In geotechnical engineering, the proper design of thermo-active geostructures (piles, foundations,
etc.) and deep waste storage disposals requires a better understanding of the thermo-hydro-mechanical behaviour
of natural and compacted soils. For design purposes, important mechanical parameters of the soils such as the
pressuremeter modulus (EP ), limit pressure (Pl ) and creep pressure (Pf ) are usually obtained by pressuremeter
in-situ tests. In the present study, pressuremeter tests were conducted in laboratory, using a mini-pressuremeter,
to characterise compacted soils. The objective was to examine and quantify the influence of temperature changes
on pressuremeter parameters of two different compacted soils: a clay and a loam. These soils were compacted
at their optimal water content and 90% of their maximal dry density (standard Proctor) in a thermo-regulated
metric scale container of 800 mm in height and 600 mm in diameter. The compacted soils were then subjected
to a range of temperatures from 20 to 50◦ C.
Only six tests were performed in each container to prevent edge effects and interaction between the tests.
The thermal cycles were applied to the soil massif as following: A heating-cooling cycle (20–40–20◦ C) for the
clay; three heating-cooling cycles (20–50◦ C) for the loam. Pressuremeter tests were conducted at the end of
several temperature steps. The obtained results showed a decrease in Pl and Pf with increasing temperature
for both tested soils, while the variation of EP was less significant. Through the temperature range tested, a
quasi-reversibility of the effect of a heating cycle is obtained.
1 INTRODUCTION et al., 1997; Graham et al., 2001; Cekerevac and Laloui

2004; Marques, 2004; Uchaipichat and Khalili 2009).
In geotechnical engineering the effect of tempera- In this study, tests on larger samples of contrasted
ture has to be studied particularly for applications materials are performed.
such as the storage of nuclear waste (Rutqvist et al. The pressuremeter test is currently used to calculate
2002), the burying of high-voltage cables (De Lieto the bearing capacity of deep foundations (AFNOR,
Vollaro et al. 2011) and the thermo-active geostruc- 2000; Standard ASTM, 1999). This test consists in
tures (Pahud 2002). These thermo-active geostructures introducing in the soil a cylindrical probe with a flexi-
are used to provide sustainable heating and cool- ble cover membrane which can expand radially in a
ing by thermal exchange between buildings and the hole. As the soil has a pseudo-elastic reaction ver-
ground through liquids that flow through closed-loop sus the probe pressure during a part of the test, it
circulation systems integrated into the geostructures. is possible to calculate a pressuremeter modulus EP .
Thermo-active geostructures may be piles, diaphragm After that, at higher pressures, large displacements take
walls, tunnel linings, basement slabs or walls (Brandl, place and the soil becomes plastic. The creep pres-
2006; Fromentin et al., 1999; Laloui et al., 2003). sure, Pf is defined as the yielding pressure between
The classical way in which they are used results in the pseudo-elastic and plastic behavior. This parame-
temperature cyclic changes in the surrounding ground ter is directly correlated to the yield locus. As a result it
temperature. Consequently, there are important ques- should be possible to quantify the potential contraction
tions about the effect of temperature variations on of the elastic domain using pressuremeter tests. Thus,
hydro-mechanical soil properties which may affect we developed an experimental method to carry out
long-term structure performance. In particular, the mini-pressuremeter tests on homogeneous soils com-
contraction of the elastic domain (i.e., the yield locus) pacted at a controlled density and water content and
with increasing temperature has been demonstrated for subjected to monotonic and cyclic thermal variations
various clay materials on centimetric samples (Tanaka (Eslami, 2014).
699
Figure 1. Granulometric curves of the studied materials.
Table 1. Basic characterization of studied materials.
PL LL PI MBV wOPN ρdmax Figure 2. Compaction curves of the studied materials.

Material % % g/100 g % Mg/m3
Table 2. Characteristics of the prepared samples.
Illitic soil 34 65 31 5.41 31 1.43
Plaisir loam 21 27 6 1.85 16 1.81
Tmin Tmax Nb
Material w (%) ρd (Mg/m3 ) (◦ C) (◦ C) cycles
Illitic soil 30.5 ± 0.5 1.31 ± 0.01 20 40 1

In the following sections, the material and the devel- Plaisir loam 16.1 ± 0.2 1.72 ± 0.02 20 50 3
oped experimental device are first described. Then, the
results are detailed and the impact of thermal variations
on the pressuremeter parameters is discussed.
The optimum water contents (wOPN ) and maximum
dry densities (ρdmax ) were obtained from the standard
2 MATERIAL AND METHOD Proctor curves performed for each material (AFNOR,
1999b) (Figure 2). For the same compaction energy,
In this part, first the studied material is presented, then the maximum dry density reached with the Plaisir
experimental device is described and finally the mini- loam is significantly higher than with the illitic soil
pressuremeter tests are detailed. in accordance with its coarser granulometry and its
mineralogical content.
The illitic soil is classified as A3 in the French stan-
dard for soil classification (GTR, 2000) and as a fat
2.1 Basic characterization of materials clay, MH, according to the Unified Soil Classification
The tested materials were an illitic soil and a loam System (Standard ASTM, 2006). The Plaisir loam is
taken in the Paris region (Plaisir). The soils were first classified as A1 in the French standard for soil classi-
dried, pulverized and sieved through 2 mm sieve before fication and as a sandy lean clay, CL, according to the
quartering and then used for various experimentations. Unified Soil Classification System.
The illitic material contains 77% illite, 10% kaolin-
ite, 12% calcite and traces of quartz and feldspar. The 2.2 Experimental device and methods
Plaisir loam contains 81% quartz, 7% dolomite, 5%
calcite, 5% clay materials and 3% felspar. To perform mini-pressuremeter tests at controlled tem-
The particle-size distribution of the illitic soil peratures, an homogeneous massif with a large volume
was determined using a laser Malvern Mastersizer is required as well as an experimental device to impose
2000®device (AFNOR, 2009) whereas the particle thermal variations to the massif. In this part, the devel-
size distribution of the Plaisir loam was determined oped methodology to obtain a meter-scale sample with
using sieving and sedimentometric method. Almost homogeneous properties (temperature, water content
85% of the particles of the illitic material are smaller and dry densities) is described.
than 0.002 mm (clay particles). The Plaisir loam is First, the required amount of water was added to
coarser as only 20% of the particles are smaller than the material in a large capacity fixed-speed (38 rpm)
0.002 mm. mixer (MIX120®) to reach the target water content
Other basic parameters like liquid limit (LL), plastic (Table 2).
limit (PL), plasticity index (PI) (AFNOR, 1993) and The wet material was stored in 8 plastic drums for
specific surfaces (AFNOR, 1999a) were compiled in a minimum of 5 days to ensure good homogeniza-
Table 1. tion. A pneumatic compactor was used to compact the
700
Figure 5. The 3 steps of a pressuremeter curve.
Figure 3. Thermo-regulated metric scale container.
Finally, the entire device was placed in a box con-
structed of 40-mm-thick extruded polystyrene plates
to reinforce the insulation.
Seven thermal sensors PT100 were positioned
within the massif at various depths and various dis-
tances from the wall of the container. The sensors were
plugged to a data logger to monitor the temperature
variation inside the compacted soil.
2.3 Mini-pressuremeter tests

The principle of a pressuremeter test is to introduce
a cylindrical probe with a flexible cover membrane
which can expand radially in a hole (AFNOR, 2000;
ASTM, 1999). The pressuremeter curve consists in
3 steps: (i) the probe inflates to obtain the contact with
the wall of the hole, (ii) the volume increases linearly
with the increasing pressure allowing the calculation of
the pressuremeter modulus EP (the soil pseudo-elastic
reaction against the probe pressure), and (iii) the large
Figure 4. Position of the pressuremeter cell. displacements take place and the soil becomes plas-
tic (Figure 5). The creep pressure, Pf is the boundary
between the second and the third steps of the test. The
material in a cylindrical stainless steel container of limit pressure Pl is the measured pressure when the
800 mm in height and 600 mm in diameter (Figure 3). injected volume reaches twice the original volume of
The compactor applied dynamic forces on a metallic the cavity.
plate of 4 mm in thick and 600 mm in diameter put Only six mini-pressuremeter tests were performed
on the top of the material to facilitate a homogeneous in each massif to prevent edge effects and cross-
compaction of the soil. To ensure a homogeneous den- influence between the tests. The test points were posi-
sity of the massif, the compaction was performed in tioned on a concentric circle with a diameter half than
eleven 70-mm-thick layers (Figure 4). that of the container. All the test points were located
A stainless steel tube welded to the outside at the same radial position for a correct comparison of
of the container was connected to a Vulcatherm® the results.
thermoregulator to facilitate the circulation of an ethy- The pressuremeter tests were conducted with an
lene glycol-water solution at the target temperature APAGEO®mini-pressuremeter probe of 380 mm in
(maximum range: −20 to 90◦ C) in the tube (Figure 3). height and 28 mm in diameter. Before each pres-
Thus, the soil compacted in the container was heated suremeter test, a core with a diameter equal to
or cooled by the container’s outer lateral surface to that of the mini-pressuremeter probe was extracted
a maximal range of temperature from 20 to 50◦ C using a core sampler. The 630-mm-length core was
(Table 2). Insulating sleeves were placed around the divided into small pieces to measure the water
tube to reduce the heat exchange with the surround- content and density of the material as a function
ing atmosphere. The top of the massif was insulated of depth. The center of the probe was positioned
with a plastic film to preserve the initial water content. halfway along the height of the compacted soil
701
(385 mm). The probe was connected to a GDS®
pressure-volume controller. The pressure controlled
test consisted in applying increasing pressure with
equal increments of 25 kPa for at least one minute per
step. The equilibrium volume was measured for each
increment and the volume was plotted as a function
of pressure. The test was stopped when the injected
volume (i.e., the volume variation limit of the probe)
reached 140,000 mm3 . Immediately after the pres-
suremeter test at a given temperature, the borehole was
filled with the same material, at the same water con-
tent, to avoid influencing the later tests. The effect of
temperature variations on the resistance of the mini-
pressuremeter membrane was measured by placing the
probe in a climatic chamber at a given temperature Figure 6. Temperature variation in the illitic compacted soil
during the calibration test of the membrane resistance. at 150 mm from the edge of the container and the chronology
of the tests: heating-cooling cycle (20–40–20◦ C), time of the
pressuremeter tests (I20a, b, c, d, I40a, b).
3 RESULTS
Table 3. Pressuremeter parameters for the test performed
The results of the pressuremeter tests performed on on illitic samples: heating cycle (20–40–20◦ C).
the illitic soil and the Plaisir loam are successively
presented in this part. T w ρd Pf Pl Ep
Test (◦ C) Cycle (%) (Mg·m−3 ) (kPa) (kPa) (MPa)
3.1 Illitic soil I20b 20 0 31.0 1.31 194 355 3.54

I40a 40 1 30.1 1.31 166 321 4.16
For the first test series, starting from 20◦ C, the soil I40b 40 1 30.4 1.31 165 288 3.25
was heated up to 40◦ C and then cooled down to 20◦ C. I20c 20 1 31.1 1.31 197 379 4.54
Figure 6 shows the temperature variation as recorded I20d 20 1 31.1 1.32 178 352 3.89
by a thermal sensor placed at 150 mm from the edge of
the massif and at the middle of its height. The thermal
equilibrium was reached after 65 h within the limit
of ±1◦ C.
The water content (w) was maintained unchanged
all test long. To verify this point, the average water
content of the soil close to the test area (from 200 to
580 mm in depth) was measured (Table 3). At 40◦ C,
(w) was slightly lower than at 20◦ C. This might be due
to the evaporation of water during the core drilling and
the sample weighting as the (w) values measured at the
end of the test (I20c,d) were identical to the (w) at the
beginning of the test (I20b). The average dry densities
were almost identical, confirming the homogeneity of
the massif.
Two mini-pressuremeter tests were performed at
each temperature step: I20a, I20b, then I40a, I40b and
finally I20c, I20d. The test I20a failed due to a mem-
brane leakage. The results of the five other tests are
Figure 7. Pressuremeter curves for the tests on illitic mate-
presented in Figure 7.
rial: I20b, I40a–b and I20c–d.
The evaluation of the pressuremeter parameters
(Ep , Pl and Pf ), allowed the quantitative comparison of
the reversibility of the observed effects on the creep
the tests at different temperatures (Table 3, Figure 11).
pressure and the pressure limit were also noticed on
The values of Pl and Pf decreased significantly when
the pressuremeter modulus (Ep) to a lower extent.
the massif was heated up to 40◦ C. Thus, the soil was
yielded under lower pressures at 40◦ C than at 20◦ C. In
other words, the results showed a softening of the soil
3.2 Loam tests
with heating.
The pressuremeter parameters for I20b and For the second test series, the test began 6 days after
I20c–d tests were close that showed a good reversibil- the compaction of the massif to ensure a good homo-
ity of the effect of the temperature variation in the geneity of the moisture in the entire sample. The initial
range of 20–40◦ C. The softening due to the heating and temperature of the loam compacted soil was 20◦ C, the
702
Figure 8. Impact of temperature variation on pressuremeter
test parameters in illitic compacted soil. Figure 10. Pressuremeter curves for the tests on loam
material: heating-cooling cycle (20–50◦ C).
Table 4. Pressuremeter parameters for the test performed

on loam samples: 3 heating cycles (20–50–20◦ C).
T w ρd Pf Pl Ep
Test (◦ C) Cycle (%) (Mg·m−3 ) (kPa) (kPa) (MPa)
L20a 20 0 16.2 1.71 278 657 5.39

L20b 20 0 16.2 1.72 280 591 5.18
L50a 50 1 15.9 1.74 220 517 4.90
L50b 50 1 16.0 n.c. 218 480 5.46
L20c 20 3 16.3 1.73 270 648 5.63
L20d 20 3 16.0 1.71 250 535 5.86
test after the application of the 3rd cycle. The results

of these six tests are presented in Figure 10.
Figure 9. Temperature variation in the loam at 150 mm The evaluation of the pressuremeter parameters
from the edge of the container and the chronology of the (Ep , Pl and Pf ), allowed the quantitative compari-
tests: 3 heating-cooling cycles (20–50–20◦ C) and time of the son of the tests at different temperatures (Table 4,
pressuremeter tests.
Figure 11). The values of the three pressuremeter
parameters decreased significantly when the massif
soil was submitted to three heating cycles up to 50◦ C was heated up to 50◦ C. As observed for the illitic com-
and then cooling down to 20◦ C. Figure 9 shows the pacted soil, the results showed a softening of the loam
imposed temperature on the outer lateral surface of the soil with heating. At the end of the cycles, at 20◦ C, the
container and the temperature variation as recorded pressuremeter parameters returned back to their initial
by a sensor placed at 150 mm from the edge of the values: results of L20c-d tests were close to those of
massif and at the middle of its height. The thermal L20a-b. A reversibility of the effect of the temperature
equilibrium was reached after 30 h within the limits variation in the range of 20–50◦ C was noticed for this
of ±1◦ C. Each step lasted 30 to 60 h. An electric fail- material as for the illitic soil.
ure caused a detectable decreasing of the temperature
during the first heating step. The I20a–b tests were
performed after the achievement of the previously set 4 DISCUSSION
temperature.
The average water content of the soil close to the In this part, the results obtained on both materials are
test area (from 200 to 580 mm in depth) was measured compared. The pressuremeter parameters measured on
(Table 4). At 50◦ C, as presented for the 1st test series, the illitic soil are lower than for the Plaisir loam:
(w) was slightly lower than at 20◦ C. The average dry – Pf illitic soil = 70% of Pf Plaisir loam;
densities were identical, confirming the homogeneity – Pl illitic soil = 57% of Pl Plaisir loam;
of this parameter over the massif. – Ep illitic soil = 67% of Ep Plaisir loam.
Two mini-pressuremeter tests were performed at
several temperature steps: L20a and L20b before the To compare the first and the second series of tests, the
first cycle; L50a and L50b at the end of the first heat- average values obtained at the end of the first step at
ing cycle and finally L20c and L20d at the end of the 20◦ C, Pf (20), Pl (20) and Ep (20) were considered as
703
Figure 11. Impact of temperature variation on pressureme-
ter test parameters in compacted loam.
Table 5. Normalized pressuremeter parameters for both

series.
Material T (◦ C) Cycle Pn.f Pn.l En.p
Illite 20 0 1.00 1.00 1.00

Illite 40 1 0.85 0.86 1.05
Illite 20 1 0.97 1.03 1.19
Loam 20 0 1.00 1.00 1.00
Loam 50 1 0.78 0.80 0.98
Loam 20 3 0.93 0.95 1.09
reference values, and all other values were normalized Figure 12. Evolution of the normalized parameters accord-
with respect to these reference values (see equations 1, ing to the test temperature for two materials.
2 and 3). Normalized parameters were presented in
Table 5. the compacted loam from 20 to 50◦ C decreased Pn.f
by 22% and Pn.l by 20%, while cooling down to 20◦ C
increased these parameters up to 15%. The level of
Pn.l and Pn.f measured in the compacted loam at the
end of the three thermal cycles was a little lower
than the initial values. This trend has to be followed
in a future study with more cycles to test a possible
cumulative effect.
The comparison between these two materials
shows that in spite of their difference in mineralogy,
initial water content and dry densities, normalized
Figure 12 shows the evolution of the normalized parameters can be considered as linear. A simple lin-
parameters at each step of each test series. ear regression (equations 4 and 5) may be used to
For the studied materials compacted at 90% of their determine Pf and Pl values according to the sample
maximal dry density, the creep pressure and the limit temperature (T):
pressure decreased with increasing temperature that
evidenced the thermal softening of both materials.
These results are consistent with the literature results
obtained with oedometric and triaxial tests (Marques
et al., 2004, Cekerevac and Laloui, 2004, Graham
et al., 2001, Tanaka et al., 1997, Hueckel & Baldi,
1990).
In this study, after heating-cooling cycles, for both 5 CONCLUSIONS
soils all pressuremeter parameters recovered approx-
imately their initial values (reversibility of thermal The objective of this paper was to quantify the
softening). Heating the illitic sample from 20 to 40◦ C effect of both monotonic and cyclic temperature
decreased Pn.f by 15% and Pn.l by 14%, while return- variations on different soil pressuremeter parameters.
ing back to 20◦ C increased these parameters respec- A specific device was developed to measure the mini-
tively up to 12% and 17%. As a similar way, heating pressuremeter parameters in laboratory conditions and
704
at different temperatures. The evolution of the pres- Cekerevac, C. & Laloui, L. 2004. Experimental study of ther-
suremeter modulus (Ep), the creep pressure (Pf ) and mal effects on the mechanical behaviour of a clay. Int. J.
the limit pressure (Pl ) was measured under imposed Numer. Anal. Methods Geomech. 28: 209–228.
thermal conditions on a clayey compacted soil and a De Lieto Vollaro, R., Fontana, L. & Vallati, A., 2011. Ther-
mal analysis of underground electrical power cables buried
compacted loam. in non-homogeneous soils. Applied Thermal Engineering,
The results of the mini-pressuremeter tests on both 31(5), pp. 772–778.
materials showed a decrease in the creep pressure and Eslami, H. 2014. Comportement Thermo-hydromécanique
the limit pressure with temperature increase (thermal des sols au voisinage des géo-structures énergétiques. PhD
softening), while the variation of the pressuremeter Thesis, Université de lorraine, France, 214p.
modulus was less significant, this could be due to the Eslami, H, Rosin-Paumier, S, Abdallah, A, Masrouri,
well-known higher variability of this parameter. The F, 2014. Impact of temperature variation on pene-
results tended to show the reversibility of the temper- tration test parameters in compacted soils. European
ature effects on the measured parameters for one and Journal of Environmental and Civil Engineering DOI:
10.1080/19648189.2014.960952.
three heating cycle both for the illitic soil in the tested Fromentin, A., Pahud, D., Laloui, L., Moreni, M.,
temperature range (20–40◦ C) and the compacted loam 1999. Pieux échangeurs: conception et règles de pré-
in the tested temperature range (20–50◦ C). dimensionnement. Rev. Fra. GC. 3, 387–421.
Graham, J.,Tanaka, N., Crilly,T. &Alfaro, M. 2001. Modified
Cam-Clay modelling of temperature effects in clays. Can.
ACKNOWLEDGEMENTS Geotech. J. 38: 608–621.
GTR, 2000. Réalisation des remblais et des couches de forme.
The authors acknowledge C. Fontaine from IC2MP Laboratoire Central des Ponts et Chaussées, Paris, p. 102.
laboratory (Poitiers, France) for the mineralogical Hueckel, T., Baldi, G., 1990. Thermoplasticity of saturated
clays: experimental constitutive study. J. Geotech. Eng.
analyses and the Region Lorraine for the financial 116, 1778–1796.
support. Laloui, L., Moreni, M. & Vulliet, L. 2003. Comportement
d’un pieu bi-fonction, fondation et échangeur de chaleur.
Can. Geotech. J. 40: 388–402.
REFERENCES Marques, M. 2004. Viscous behaviour of St-Roch-de-
l’Achigan clay, Quebec. Can. Geotech. J. 38: 25–38.
AFNOR, 1993. NF P94-051 – Sols: reconnaissance et essais; Pahud, D., 2002. Geothermal energy and heat storage.
Détermination des limites d’Atterberg – Limite de liq- SUPSI – DCT – LEEE Laboratorio di Energia, Ecologia
uidité à la coupelle- Limite de plasticité au rouleau. ed Economia, pp. 1–133.
Association Française de Normalisation, Paris, France, Rutqvist, J. et al., 2002. A modeling approach for analy-
p. 15. sis of coupled multiphase fluid flow heat transfer, and
AFNOR, 1999a. NF EN 933-9 – Tests for geometrical deformation in fractured porous rock. International Jour-
properties of aggregates – Part 9: Assessment of fines nal of Rock Mechanics and Mining Sciences, 39(4),
– Methylene blue test. Association Française de Normal- pp. 429–442.
isation, Paris, France, p. 12. Standard ASTM, 1999. D4719-00 Standard test method for
AFNOR, 1999b. NF P94-093 – Sols: Reconnaissance et Prebored Pressuremeter Testing in Soils. ASTM Interna-
essais Détermination des références de compactage d ’un tional, West Conshohocken, PA. ASTM. p. 9.
matériau. Essai Proctor Normal-Essai Proctor Modifié. Standard ASTM, 2006. D2487 Standard practice for clas-
Association Française de Normalisation, Paris, France, sification of soils for engineering purposes (Unified
p. 18. Soil Classification System). ASTM International, West
AFNOR, 2000. NF P94-110-1 – Sols: reconnaissance Conshohocken, PA www.ASTM.org.
et essais Essai pressiométrique Ménard. Association Tanaka, N., Graham, J. & Crilly, T. 1997. Stress-strain
Française de Normalisation, Paris, France, p. 44. behaviour of reconstituted illitic clay at different temper-
AFNOR, 2009. ISO 13320: Particle size analysis – Laser atures. Eng. Geol. 47: 339–350.
diffraction methods. Association française de normaliza- Uchaipichat, A. & Khalili, N. 2009. Experimental inves-
tion, 60p. tigation of thermo-hydro-mechanical behaviour of an
Brandl, H. 2006. Energy foundations and other thermo-active unsaturated silt. Géotechnique 59: 339–353.
ground structures. Géotechnique 56: 81–122.
705
Clay-water interactions in swelling claystones: The case of the

Callovo-Oxfordian claystone
H. Menaceur, P. Delage & A.M. Tang

Ecole des Ponts ParisTech, Navier/CERMES, Marne la Vallée, France
J. Talandier
Andra, Châtenay-Malabry, France
ABSTRACT: An investigation of the microstructure changes of the Callovo-Oxfordian (COx) claystone

(a possible host rock for geological radioactive waste disposal in France) due to suction changes was con-
ducted by means of mercury intrusion porosimetry tests. A unimodal pore population around a mean diameter
of 32 nm was observed at initial state, with pore size distribution (PSD) curves slightly moved towards smaller
mean diameters (27–28 nm) under suctions of 150 and 331 MPa, respectively, with diameter reducing to 20 nm
after oven-drying. Wetting from the initial state (suction 34 MPa, degree of saturation of 77.6%) to 9 MPa suction
led to water saturation with no significant change of the PSD curve, whereas wetting at zero suction gave rise to
the appearance of cracks of several micrometers width, together with an enlargement of the initial pore popula-
tion. The step hydration mechanisms of smectites by successive placement of distinct layers of water molecules
within the clay platelets along the smectite faces with respect to the suction applied helped understanding these
microstructure changes. An average number of 22 layers in a platelet was estimated from the dry PSD curve,
based on a brick model, and it was shown that hydration at suction of various hundreds of MPa was due to the
adsorption of one water layer whereas that under suction larger than 9 MPa (various tens of MPa) were charac-
terised by the adsorption of 2 water layers. Significant microstructure changes below 9 MPa corresponded to the
third adsorbed layer. This data evidence the significant difference in status of the strongly adsorbed intra-platelet
water (estimated to 25–30% of the total water) compared to the inter-platelets water (70–75%) that behaves like
free water and that is involved in water transfer and hydro-mechanical couplings.
1 INTRODUCTION 2 MATERIALS AND METHODS
The Callovo-Oxfordian claystone (COx), an indurated 2.1 The Callovo-Oxfordian claystone

clay rock 155 million years old (limit upper-middle
The COx claystone specimens investigated here come
Jurassic), is considered as a potential host rock for
from the Underground Research Laboratory excavated
radioactive waste disposal at great depth in France.
at a depth of 490 m close to the village of Bure (eastern
Due to the excavation and ventilation of the galleries,
France) by Andra, the French Agency for the man-
the rock mass around galleries and disposal cells is
agement of radioactive wastes. The COx claystone is
prone to desaturation, that will be followed by pro-
155 millions years old (limit upper-middle Jurassic)
gressive resaturation at the end of the operating stage
and the thickness of layer is about 150 m.
once the galleries are closed. In this regard, the water
At this depth, the COx claystone is made up
retention properties of the COx claystone have been
(Gaucher et al., 2004) of a clay matrix (45–50%)
investigated in details by Wan et al. (2014), follow-
containing detritic grains of carbonate (20%), quartz
ing previous investigations by Pham et al. (2007) and
(22%) and other minerals (feldspars, pyrite, dolomite
Boulin et al. (2008) who related it to its pore size
and siderite, 9%). The clay matrix is mainly made up of
distribution (PSD) investigated by mercury intrusion
interstratified minerals of illite–smectites (Yven et al.
porosimetry (MIP).
2007).
In this paper, the changes in microstructure with
The specimens were drilled from a 80 mm diameter
respect to changes in suction are investigating based
and 300 mm long core (EST44584) parallel to bedding.
on MIP experiments that were carried out on freeze-
The water content was determined after drying in the
dried specimens put at various suctions along both the
oven at 105◦ C for 48 hours. The porosity and degree
drying and wetting paths starting from the initial state
of saturation were calculated from carefully measuring
of the specimen in the laboratory.
707
Table 1. Saline solutions used. carried out in a Micromeritics-AutoPore IV 9500
porosimeter from a low initial pressure of 3.4 kPa up to
Solution Relative humidity (%) Suction (MPa) 227.5 MPa, corresponding to entrance pore diameters
of 363.6 µm and 5.5 nm respectively.
KOH 9 331 The intruded mercury porosity (nHg ) was defined
MgCl2 33 150
as the ratio VHg /V of mercury intrusion volume VHg
KNO3 93.7 9
Pure water 100 0 to specimen total volume V . The pore entrance diam-
eter (D) was determined from the intrusion pressure
PHg by assuming a cylindrical pore shape according
to the Laplace-Young equation (D = 4 σ cos θ/PHg )
the sample volume by using hydrostatic weighing. The where σ is the mercury-solid interfacial tension and θ
initial total suction was determined by using a dew is the mercury-solid contact angle (σ = 0.484 N/m and
point tensiometer (WP4, Decagon). The degree of sat- θ = 141.3◦ according to Diamond 1970).
uration was around 77.6% corresponding to a suction
of 34 MPa for a porosity of 17.0%. The dry density and
2.3 Test program
grain density values are 2.16 and 2.7 Mg/m3 , respec-
tively. Partial saturation resulted from the combined A first series of samples were submitted to drying-
actions of coring, stress release, transport, storage and wetting path from the initial water content (equal to
specimen preparation. 6.12% with a 34 MPa suction and a degree of satura-
tion of 77.6%). Specimens A1 and B1 were used to
investigate their microstructure at 150 and 331 MPa
2.2 Experimental techniques
respectively. Once dried at 331 MPa, specimens A2
All tests were run on specimens of 38 mm in diameter and B2 were afterwards wetted at 9 MPa and zero suc-
and between 8–10 mm in height. To do so, the core was tions, so as to determine the main wetting path. Along
firstly sliced by using a diamond wire saw at low speed the wetting path, both specimens were periodically
in dry condition to obtain a cylinder sample of 80 mm taken out from the desiccator for weighing so as to
in thickness that was then placed in a special metal determine their water content. Their suction was also
confining mould and cored to the desired diameter measured by means of the WP4 dew-point tensiometer.
(38 mm) by using a diamond barrel in dry condition. A second series of tests corresponded to wetting-
Finally, small specimens of 8–10 mm in thickness were drying paths. Starting from the initial water content
cut by using a diamond saw. (6.12%), three specimens (C, D1 and D2) were firstly
Suctions were imposed by the vapour control tech- wetted under decreasing suctions (9 MPa for specimen
nique using saturated saline solutions as shown in C and zero suction for specimens D1, D2). Specimen
Table 1. D2 was then used to determine the main drying path
Specimens were left in the desiccators until reach- up to suctions of 150 MPa and 331 MPa.
ing mass stabilization (checked by periodic precision
weighing). Once equilibrated, a suction measurement
was made by using the WP4 dew point potentiometer. 3 WATER RETENTION CURVE
Afterwards, the specimens were immediately waxed
using slush wax (at lowest possible temperature before The water retention curve expressed in terms of
solidification, see Wan et al. (2013) for more detail). changes in water content versus suction is presented
Careful weighing was carried out prior and after wax- in Figure 1. The points at zero suction are arbitrar-
ing, giving a good determination of the wax weight ily plotted at a suction of 0.01 MPa. Starting from the
and volume. The specimen volume was obtained by initial state (w = 6.12%, suction 34 MPa), two points
subtracting the wax volume from that of the waxed were obtained from two distinct specimens (A1 and
specimen. The wax technique was also used to obtain B1) with a direct single step drying at controlled suc-
the specimen volume at initial state (34 MPa of suction, tions of 150 and 331 MPa, respectively. Similarly, two
degree of saturation of 77.6%). Finally, the specimens other points were obtained from specimens C and D1
were cut into small pieces to measure their water con- with a direct single step wetting at suctions of 9 and
tent by oven drying. The void ratio and degree of 0 MPa.
saturation were determined by measuring the volume Data along the main wetting and drying paths were
by means of hydrostatic weighing. also obtained by suction measurements and water con-
Mercury intrusion porosimetry (MIP) tests were tent determination along the wetting path from 331
conducted on various specimens at different suc- to 0 MPa for specimen B2 and along the drying path
tions along both drying or wetting paths. Dehydration from zero suction to 331 MPa for specimen D2. Good
was made by freeze-drying small pieces of claystone correspondence is observed along the main wetting
(1–3 g in weight) that were previously quickly frozen path between the data of specimens B2 and A2 that
by immersion in slush nitrogen obtained by previously was wetted between 331 and 9 MPa, providing some
applying vacuum (reducing nitrogen temperature from confidence in the quality of the data.
boiling temperature at −196◦ C to freezing tempera- Examination of the main drying and wetting paths
ture at −210◦ C (Delage et al. 2006). MIP tests were gives evidence of a hysteresis. The data also show that
708
Figure 1. Water retention curve of the COx claystone.
the initial state resulting from desaturation from the

in-situ saturated state is located, as expected, on the
main drying path.
The wetting path starting from the initial state shows
that the point at 9 MPa of suction (specimen C) is
located below the main drying path, which is typical of
scanning curves (there is unfortunately no other point
between 9 MPa and zero suction along this path). The Figure 2. Changes in degree of saturation and volume with
point at zero suction obtained along the main wetting respect to suction.
path starting from the dry state at 331 MPa provides a
water content (11.2%) higher than that obtained after
wetting the specimen from the initial water content infra-porosity found here confirms previous findings
(10.4%). on the COx claystone by Yven et al. (2007), Boulin
The changes in degree of saturation and volume et al. (2008) and Delage et al. (2014) who estimated
with respect to suction are presented in Figure 2. the porosity not intruded by mercury at 25% of the
The Figure shows that the specimen becomes quasi- total porosity, in accordance with the 23.5% value
saturated at 9 MPa (Sr = 96%, w = 7.65%) along the with diameter smaller than 5.5 nm found in the present
wetting path with however little volume changes, research. The density function curve exhibits at ini-
whereas most swelling is observed for suctions tial state a typical monomodal curve with quite a
between 9 and 0 MPa with 6.1% swelling at zero well defined pore population identified by an inflec-
suction. tion point at 32 nm. As shown by Sammartino et al.
Along the drying path from initial state, a linear (2003),Yven et al. (2007) and Boulin et al. (2008), this
relationship is observed between the degree of satura- pore population is related to the average entrance pore
tion and the water content with an average decrease of diameter within the clay matrix.
13% in degree of saturation for a decrease of 1% in Figure 4 presents an enlargement between 0.001 and
water content. The main wetting curve obtained once 0.1 µm of the pore size distribution curves of speci-
the specimen has been dried to the highest suction mens A1 and B2 submitted to a suction increase along
(331 MPa) is located below the curves of drying- the drying path up to 150 and 331 MPa respectively
wetting from initial state, confirming the hysteresis from the initial state, together with the PSD data of
observed by Wan et al. (2013). the specimen at initial state and that of a specimen
oven-dried at 105◦ C for 48 h.
The total porosities obtained from volume measure-
4 MERCURY INTRUSION POROSIMETRY ments are also given. The intruded mercury porosi-
ties nHg of the specimens dried up to 150 (A1)
The pore size distribution curve of the specimen at ini- and 331 MPa (B1) are equal to 12.8 and 12.2%,
tial state (Sr = 77.6%, porosity n = 17.0%) is plotted respectively, compared to total porosities of 15.9 and
in cumulative and density function curves in Figure 3. 15.2%, respectively, showing that the infra-porosities
The total porosity n of the specimen is also plotted not intruded by mercury (D < 5.5 nm) are again
in the cumulative curve, showing that the mercury lower than the total ones. The infra-porosities are
intruded porosity nHg is smaller than the total one, smaller than that at initial state (n − nHg = 3.0% for
with an infra-porosity (n − nHg ) of 4%. both specimens dried up to 150 and 331 MPa, com-
Pores smaller than the lower limit of 5.5 nm (cor- pared to 4% at initial state). Conversely, the mer-
responding to the maximum mercury pressure of cury intruded porosity of the oven-dried specimen
227.5 MPa) cannot be intruded. The non negligible (12%) appears to be quite close to the total one
709
Figure 3. Pore size distribution curves of specimen
EST44584 at initial state. Figure 5. Pore size distribution curves along the drying
path.
at initial state with a only slight increase in total

porosity at 9 MPa (0.2%), showing that the infra-
porosity is also similar in both cases.This is compatible
with the negligible volume change observed in the
volume change/suction curve of Figure 2 when pass-
ing from initial state (s = 34 MPa) to a suction of
9 MPa. The specimen saturation from Sr = 78% to
96% hence occurs by complete filling of the mean pore
population with very little microstructure changes
both in the porosity intruded and not intruded by
mercury.
The reduction in suction from 9 MPa to 0 MPa
occurs at quasi-saturated state with a significant 6.1%
swelling giving a final water content of 10.5% and
degree of saturation of 99% (Figure 2). Indeed, signif-
icant changes are also observed at the microstructure
level, as shown by the change in shape of the PSD
curve of specimen D1 at zero suction that becomes
bimodal. Changes affect three levels of porosity:
i) The infra porosity n − nHg that increases from 4%
Figure 4. Pore size distribution curves along the drying
path.
at 9 MPa to 6.1% at zero suction with an intruded
porosity of 15.6% and a total one of 21.7%;
ii) The large porosity with appearance of a new
(13%) giving an infra-porosity n − nHg = 1% signif-
pore population corresponding to a proportion of
icantly smaller than at suctions of 150 and 331 MPa
10.5% of the total porosity in the range of diame-
(3%).
ters between 7 and 100 µm with a mean diameter
In Figure 5, the pore size distribution curves of
around 12 µm;
the two specimens wetted from initial state at low
iii) An enlargement of the previous pore population
suctions (9 MPa and zero) and higher degree of sat-
that moves from a narrow range between 12 and
uration (Sr = 96% and 99% for s = 9 and 0 MPa,
50 nm (see density function curve in Figure 5) to
respectively) are presented together with the data at
a wider range between 12 and 500 nm with a new
initial state (Sr = 77,6%, s = 34 MPa). Also plotted in
plateau between 60 and 200 nm.
the Figure are the total porosities of the specimens.
The PSD cumulative and density function curves at Note however that no change is observed in the
9 MPa suction (Sr = 96%) are quite similar to that density function curves between 5 and 20 nm.
710
relative humidities), one water layer is adsorbed along
the smectite surface. Two layers are adsorbed at various
tens of MPa whereas a third layer is adsorbed below
around 10 MPa. The limits in suction change with the
origin of the smectites and nature of the cations, but
these average values provide a reasonable order of
magnitude of the correspondence between the number
of water layers and the suction values.
Observation of the PSD curves of Figure 4 along
the drying path (initial state, 150, 330 MPa suctions
and oven-dried specimen) indicates that drying is char-
acterized by a reduction of the mean diameter of
the single well defined pore population from 32 nm
(intact) to 28 (150 MPa), 27 (331 MPa) and 21 nm
(oven-dried). Based on the brick model, this corre-
sponds to a reduction in thickness of the platelets.
This reduction is only due to changes in the interlayer
spacing of smectite, given that illite minerals are not
Figure 6. Schematic model of COx microstructure (after sensitive to changes in water content with a constant
Yven et al. 2007). thickness of 9.6 Å. Based on the suction values given
above, it can be reasonably considered that all spec-
imens at suction higher than 150 MPa have only one
5 DISCUSSION layer of adsorbed water molecules along the smec-
tite minerals within the clay platelets. At initial state
The interpretation of the microstructure changes with and a suction of 34 MPa, the data of Sayiouri et al.
respect to suction changes is interpreted based on the (2004) would indicate the possibility of having two
conceptual model of the COx claystone microstructure layers adsorbed, in reasonable compatibility with the
of Yven et al. (2007) presented in Figure 6. The model data of Ferrage et al. (2005). The reduction of the aver-
schematically shows how individual calcite or quartz age diameter from 32 to 28 nm can then be linked to
detritic grains are embedded into a clay matrix that the transition from two to one adsorbed layer of water
represents 45–50% of total constituents at the depth of molecules.
490 m considered here. Oven drying resulted in having no more water layer
The well defined single pore population defined by adsorbed (Ferrage et al. 2005). An estimation of the
an average value of 32 nm observed in the PSD curve at average number of clay layers in one platelet of mixed-
initial state can be interpreted by assimilating the clay layer illite-smectite can then be obtained by using the
matrix to an assembly of bricks made up of platelets mean diameter of 21 nm measured on the oven-dried
of comparable thickness, as can be seen in Figure 6. specimen, that corresponds to the average platelet
As seen on the scheme, the diameter of the circular thickness. With an interlayer spacing of 9.6 Å, this
pore located within the bricks provides an estimate of provides a number of 21–22 layers per platelet. Con-
the brick thickness. The mean diameter detected in the sidering the proportion of 50–70% smectite minerals
PSD curve hence provides an estimation of the average provided by Yven et al. (2007) at the level considered
platelet thickness (32 nm). in the COx layer, one can conclude that the introduc-
Further understanding on the changes in microstruc- tion of one water layer along the smectite layers would
ture observed can be gained by considering the mech- result in the placement of between 11 and 15 layers
anisms of hydration of smectites. Many studies have of 3 Å thick layers of water molecules, resulting in an
been devoted to the interactions between smectite min- increase in the platelet thickness between 3.3 and of
erals and water and to the hydration mechanisms of 4.6 nm, from 21 to 24.3–25.6 nm. This is not far from
smectites. Based on X-ray diffractometry techniques the 28 nm value measured by MIP under suctions of
and through the observation of the changes during 150 and 331 MPa. Adding another water layer would
hydration of the d001 interbasal spacing, it has been then provide a thickness between 27.6 and 30.2 nm,
shown that hydration from a dry state occurred through reasonably comparable to the 32 nm measured by MIP
the ordered placement, step by step, of one, then two, under 34 and 9 MPa suctions.
then three and finally four layers of water molecules This indicates that the analysis based on the step
along the smectite surface (e.g. Mooney et al. 1952, hydration process evidenced in pure and compacted
Norris 1954). Interestingly, this mechanism is depend- smectites is reasonably compatible with the analysis
ing of the suction of the pore water. Convergent data of the MIP data with respect to the changes in platelets
independently obtained by various authors on vari- thickness in the COx clay matrix with suction. Better
ous smectites (including Saiyouri et al. 2000, 2004, fitting is actually obtained with the highest proportion
Ferrage et al. 2007) indicated that at high suctions of smectite of 70% in the clay fraction. The mecha-
of various hundreds of MPa (imposed by controlled nism of hydration of smectites hence appears to be
711
of significant interest to interpret the water retention the hydro-mechanical couplings characterised by the
properties and related microstructure changes of the water pore pressure as well as water transfers within
COx claystone. the claystone are governed by this free inter-platelet
As observed in Figure 5, little change was observed water and do not concern the remaining two water
between the sample at initial state (34 MPa) and that at molecules that are adsorbed along the smectite faces
9 MPa both in average mean pore diameter and infra- at initial state. This provides an insight into the status
porosity, indicating that the two water layers are stable of water in the claystone.
under this suction range, with the third water layer only
adsorbed at suction smaller than 9 MPa.
The significant swelling due to wetting at zero 6 CONCLUSION
suction was analysed through three mechanisms in
section 4. The mechanism i) that concerns the infra The use of mercury intrusion porosimetry on freeze-
porosity can be related to the adsorption of a third dried specimens provided better understanding of the
or even a fourth layer of water molecules within the changes in microstructure that occur along the wetting
infra-porosity that increased from 4% (initial state and drying paths of the water retention curve of the
at 34 MPa) to 6.1% at zero suction. Based on the COx claystone. The concepts developed to describe the
hydration model used above, adding a new water hydration mechanisms of smectites and their depen-
layer would result in an increase of 7 nm in the dency with respect to suction changes appeared to be
platelet thickness, giving an average value of 39 nm. applicable to the hydration of the mixed-layer illite-
Adding a fourth water layer would lead to an aver- smectite minerals that is responsible of the change in
age thickness of 46 nm. These increases in thickness water content and the swelling-shrinkage behaviour of
are compatible with mechanism ii) with an enlarge- the COx claystone.
ment of the diameter of the mean pore population The initial state of the COx specimen considered
towards larger value. It does not however explain in this work is located on the main drying path of
the appearance of pores as large as 200 or 300 nm the water retention curve and characterized by a suc-
that can be linked to a higher degree of disorder, tion of 34 MPa that corresponds to the adsorption
as observed in compacted bentonites (Sayiouri et al. of two layers of water molecules along the smectite
2004) in which the suction reduction below 0.1 MPa faces. Releasing the suction to 9 MPa results in the
(appearance of the fourth layer) also involved a reduc- quasi-saturation of the claystone but it does not sig-
tion in thickness of the platelets due to exfoliation of nificantly affects the COx microstructure with quite
clay layers together with an increase in the number of little swelling and no change in the pore size distribu-
platelets. Further investigation is necessary here in the tion curve, in accordance with the stability of the two
case of the COx claystone to better understand this layers of water molecules in this suction range. Con-
phenomenon. versely, passing from 9 MPa to zero suction allows
The appearance of the large pore population with the placement of a third, or even a fourth layer of
diameters between 7 and 100 µm is related to the water molecules that results in significant changes in
appearance of cracks that can be observed visually and the inter-platelets porosity. Simultaneously, a network
have also been observed by Wang et al. (2014) by using of saturated cracks appears and the global swelling
Digital Image Correlation at microscopic level. These observed at zero suction comes from the combined
cracks, that are saturated as observed in Figure 2, were action of the changes in the inter-platelet porosity and
suspected in the analysis of the water retention prop- the generation of cracks.
erties of the COx claystone provided by Wan et al. Imposing suctions of 150 and 331 MPa results in a
(2013). reduction from 32 to 28–27 nm of the mean diameter
The investigation carried out here also evidenced of the pore size distribution curve that keeps the same
the roles of two distinct natures of water. Given that shape, in agreement with the placement of one layer of
they correspond to suctions larger than 7–9 MPa, the water molecules in this suction range, which explain
layers of water molecules adsorbed within the mixed- the small differences in microstructure observed at
layers illite-smectite platelets at initial state, along these two suctions. Conversely, oven-drying at 105◦ C
the drying path and even at 9 MPa suction are heav- results in a further decrease to 20 nm in the mean diam-
ily bonded to the smectite surfaces. Contrarily, the eter of the inter-platelet porosity. These features are
water molecules contained in the inter-platelets poros- compatible with the state of smectite minerals at dry
ity with mean diameter of 32 nm can be considered as state, with no water layer adsorbed and an interlayer
free water. An estimation of the relative proportion of space of 9.6 Å.
adsorbed and free water can be obtained from the pore The validity of the concepts of hydration of smec-
size distribution curve of Figure 3. According to the tites indeed provides deeper insight in the under-
shape of the density function curve, the lower limit of standing of changes in water contents and swelling-
the main pore population can be estimated at a diam- shrinkage behaviour of the COx claystone. They con-
eter of 0.01 µm. With a total porosity of 0.170, the firm, as also observed when comparing the water
corresponding proportion of free water of the order of retention behaviour of powder and compacted smec-
68%. In other words, MIP provides a porosity that is tites (Delage et al. 2006), that microstructure effects
not far from that occupied by free water. It is likely that are mainly governed by physico-chemical interactions,
712
with little effects of the initial fabric and structure of Diamond, S (1970). Pore size distribution in clays. Clays Clay
the claystone. In other words, the intensity of these Minerals 18:7–23.
interactions is strong enough to surpass and break the Ferrage E, Lanson B, Sakharov BA, Drits VA (2005). Inves-
inter-particle bonding at initial state, even in the case tigation of smectite hydration properties by modeling
experimental X-ray diffraction patterns: Part I. Mont-
of the strong inter-particle bonds within the clay matrix morillonite hydration properties. American Mineralogis.
of the claystone. These findings should help with any 90:1358–1374.
problem linked to swelling in shales, like in petroleum Gaucher G, Robelin C, Matray JM, Négrel G, Gros Y, Heitz
industry (stability of boreholes in shales) or tunnelling JF, VinsotA, Rebours H, Cassagnabère, BouchetA (2004).
(interaction between shales and the support). ANDRA underground research laboratory: interpretation
Finally, the mechanisms of water adsorption along of the mineralogical and geochemical data acquired in the
smectites allows to better define the status of water Callovian-Oxfordian formation by investigative drilling.
in claystones and shales containing some smectites, Physics and Chemistry of the Earth 29:55–77.
with a distinction between the adsorbed water (located Mooney RW, Keenan AC, Wood LA (1952). Adsorption of
water vapor by montmorillonite. II. Effect of exchange-
within the platelets and made up of two layers strongly able ions and lattice swelling as measured from X-ray
bonded along the smectite faces with suctions larger diffraction. Journal of the American Chemical Society
than 9 MPa) and the free water (located in the inter- 74:1371–1374.
platelets porosity, defining the pore pressure and Norrish K (1954). The swelling of montmorillonite. Discus-
submitted to water transfers). The MIP curve at ini- sions of the Faraday Society, 18, 120–133.
tial state provides an idea of the proportions of the two Pham QT, Vales F, Malinsky L, Nguyen Minh D, Gharbi H
types of water, with around 68% of free inter-platelets (2007). Effects of desaturation-resaturation on mudstone.
water and 32% of adsorbed intra-platelets water. Physics and Chemistry of the Earth 32:646–655
Saiyouri N, Tessier D, Hicher PY (2004). Experimental study
of swelling in unsaturated compacted clays. Clay Minerals
39:469–479.
ACKNOWLEDGEMENTS Sammartino S, Bouchet A, Prêt D, Parneix JC, Tevissen
E (2003). Spatial distribution of porosity and miner-
The authors are indebted to Andra for supporting the als in clay rocks from the Callovo–Oxfordian formation
work and for providing the samples. (Meuse/Haute-Marne, Eastern France)—implications on
ionics species diffusion and rock sorption capability.
Applied Clay Science 23(1–4):157–166.
REFERENCES Wan M, Delage P, Tang AM, Talandier J (2013). Water
retention properties of the Callovo-Oxfordian claystone.
Boulin PF, Angulo-Jaramillo R, Daian JF, Talandier J, International Journal of Rock Mechanics and Mining
Berne P (2008). Pore gas connectivity analysis in Callovo– Sciences 64: 96–104.
Oxfordian argillite. Applied Clay Science 42(1–2): Wang LL, Bornert M, Héripré E, Yang DS, Chanchole S
276–283. (2014). Irreversible deformation and damage in argilla-
Delage P, Marcial D, Cui YJ, Ruiz X (2006). Ageing effects ceous rocks induced by wetting/drying. Journal of Applied
in a compacted bentonite: a microstructure approach. Geophysics 107:108–118.
Géotechnique 56(5):291–304. Yven, B., Sammartino, S., Géroud,Y., Homand, F. & Villiéras,
Delage P, Menaceur H, Tang AM, Talandier J (2014). Suction F. (2007). Mineralogy, texture and porosity of Callovo-
effects in deep Callovo-Oxfordian claystone specimen. Oxfordian claystones of the Meuse/Haute-Marne region
Géotechnique Letters 3(2), 84–88. (eastern Paris Basin). Mém. Soc. Géol. France 178, 73–90.
713
Air injection tests in two argillaceous rock formations:

Experimental results and modelling
L. Gonzalez-Blanco & E. Romero

Department of Civil and Environmental Engineering, Universitat Politècnica de Catalunya, Barcelona, Spain
X.L. Li
EIG EURIDICE, Mol, Belgium
X. Sillen
ONDRAF/NIRAS, Brussels, Belgium
P. Marschall
NAGRA, National Cooperative for Disposal of Radioactive Waste, Wettingen, Switzerland
C. Jommi
Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands
ABSTRACT: Air flow through two argillaceous rock formations is investigated on the basis of laboratory
work and its modelling to help in the interpretation of the results. Priority in the experimental program has
been given to study the volume change response of these initially water saturated materials along relatively fast
and controlled volume rate air injections. These high rates intend to give preference to single-phase air flow
mechanisms associated with the opening of stress-dependent pathways. Particular attention has been focused on
the changes in the pore network to detect opening of fissures after air injection tests using mercury intrusion
porosimetry. Selected experimental results have been simulated using a fully coupled hydro-mechanical finite
element code, which incorporates an embedded fracture permeability model to account for the simulation of the
gas flow along preferential pathways.
1 INTRODUCTION To improve the knowledge on the response to gas

migration on argillaceous host formations, in particu-
Argillaceous sedimentary formations have been stud- lar the transport of free gas phase through preferential
ied during the last decades as potential host for- pathways, a comprehensive series of fast air injection
mations for the geological disposal of long-living tests under oedometer and isotropic conditions has
and heat-emitting radioactive waste in Belgium and been performed. Two potential host rocks have been
Switzerland. A significant issue in the long-term per- studied to provide the basis for a comparative study:
formance of these potential host rocks concerns the a plastic clay formation from Belgium (Boom Clay);
generation and migration of gases. Actually, in the and an indurated and deeper shale formation from
post-closure phase of a disposal system, gases can Switzerland (Brown Dogger). Air injection and dis-
be produced as a result of the anaerobic corrosion sipation tests have been performed at two orientations
of metal canisters, radiolysis, and microbial degra- (orthogonal and parallel to bedding planes) and at two
dation of organic waste (ONDRAF/NIRAS 2013). air injection rates. Sample volume changes have been
The pressure resulting from the gas generation in an measured during air injection and dissipation. Particu-
almost impermeable geological medium in the near- lar attention has been given to the changes in the pore
field of a repository will increase. Under high gas network to detect the opening of fissures or disconti-
pressures, the mechanical and hydraulic properties nuities. Simulation aided techniques have been used
of the host rock are expected to change signifi- to better exploit the information provided by measure-
cantly. Preferential gas pathways may develop taking ments and to get a better comprehension of local and
advantage of the material heterogeneity, anisotropy, coupled processes affecting the response of the mate-
rock discontinuities or interfaces between the dif- rial. To this aim, a coupled hydro-mechanical analysis
ferent components of the repository, which even- with an embedded fracture permeability model, which
tually may lead to the release of the produced is implemented in the finite element code Code_Bright
gases. (Olivella et al. 1996), has been used.
715
2 MATERIALS AND EQUIPMENT Table 1. Geotechnical properties of the samples.
2.1 Boom clay Main properties/ BC BD

Initial conditions Symbol (223 m) (782 m)
Boom Clay (BC) (Northwest European Tertiary Basin,
Rupelian stage) was retrieved at a depth of 223 m in Plastic limit (%) wP 29 24
the URL (Underground Research Laboratory) facility Liquid limit (%) wL 67 17
of EURIDICE at Mol, Belgium. Table 1 summarises Dominant entrance pore 70 22
the main geotechnical properties of this clay (20– size from MIP (nm)
30% kaolinite, 20–30% illite, 10–20% smectite, 25% Air-entry value from 4.8 13
dominant entrance
quartz and feldspar), including the initial suction. BC
pore size (MPa)
specimens displayed a relatively high initial suction, Density of solids (Mg/m3 ) ρs 2.67 2.69
despite being saturated, due to stress relief upon sam- Dry density (Mg/m3 ) ρd 1.66–1.69 2.48
pling. The value measured with dew point mirror Water content (%) w 22.6–24.0 4.78
psychrometer was around 2.5 MPa, in good agreement Degree of saturation Sr close to 1 0.95–1
with data reported by Lima et al. (2012) and Dao Void ratio e0 0.57–0.61 0.09–0.11
et al. (2014). Data obtained from Mercury Intrusion Total suction (MPa) s 2.5 37
Porosimetry (MIP) allowed detecting an initial mono-
modal pore size distribution, as well as estimating an
air-entry value of about 4.8 MPa that corresponded to
the dominant entrance pore size of 70 nm (Fig. 1).
Water retention data of the intact material on dry-
ing are plotted in Figure 2. The data were obtained
by stepwise drying of the specimens using a dew
point psychrometer, starting from the initial suction.
The data were fitted to the van Genuchten’s equation
indicated in the figure.
2.2 Brown dogger

Core samples of Brown Dogger (BD) were recovered
from the geothermal well Schlattingen-1 (SLA-1) at a
depth of around 782 m. SLA-1 is located in the Can-
ton Thurgau (Switzerland) in the North-eastern part Figure 1. Pore size density function of BC.
of the Swiss Molasse Basin. BD corresponds to a
shaly sequence and consists of silty to clay-rich marls
with clay content of 25–45%, 30–50% carbonate and
20–25% quartz (Ferrari et al. 2014). The main geotech-
nical properties of the samples are summarised in
Table 1. The information is complemented with the
initial suction and the dominant entrance pore size
determined by MIP. Figure 3 shows the pore size
distribution of BD, which presents a mono-modal dis-
tribution with a small dominant entrance pore size of
22 nm. Water retention curves on drying are presented
in Figure 4.
2.3 Experimental equipment

An instrumented high-pressure and high-stiffness
oedometer cell was used to determine the
compressibility parameters of the two materials and
to perform the air injection tests on BC.
Figure 2. Drying branch of the water retention curve of BC.
The oedometer samples, 20 mm thick and 50 mm
in diameter, were placed between top and bottom caps
made of concentric stainless steel rings that work stress, one for air injection (upstream boundary-
as coarse porous stones allowing the injection and bottom of the sample-) and two for water (injection
recovery of water and air. Vertical displacements were at upstream and recovery at downstream bound-
measured with an external Linear Variable Differ- aries). The cell and auxiliary devices are presented in
ential Transformer (LVDT). The experimental set-up Figure 5.
included four automatic Pressure/Volume Controllers An equivalent experimental setup was used for the
(PVC): one for hydraulically applying the vertical air injection tests on BD using an isotropic cell and
716
Figure 3. Pore size density function of BD.
Figure 5. Scheme of the experimental oedometer set-up. 1)

Sample; 2) coarse porous concentric rings; 3) axial loading
piston; 4) PVC for vertical stress; 5) PVC for air injection;
6) and 7) PVC for water; 8) LVDT.
injection phase, a fast drainage of the bottom line was

performed to replace water by air. An initial air pres-
sure of 0.5 MPa for BC and 3 MPa for BD was applied
at the upstream boundary.
Figure 4. Drying branch of the water retention curve of BD. Air injection tests at two different injection rates
(2 mL/min and 100 mL/min) were performed to anal-
yse their influence on the coupled hydro-mechanical
low-height specimens (25 mm height, 50 mm in diam-
response. The selected rates were relatively fast (air
eter) confined by several neoprene membranes and
pulse tests) to minimise air diffusion mechanisms
aluminium foils. Axial displacements were registered
through the matrix and to enhance single-phase air
with an external LVDT. Each cap of the isotropic cell
flow mechanisms through discontinuities (Marschall
had inlet and outlet lines prepared for gas and liq-
et al. 2005). The air injection piston was stopped (shut-
uid connections and connected to PVCs, as shown in
off) when air pressure reached a maximum of 4 MPa
Figure 5 for the oedometer cell.
in BC tests and 14 MPa in BD tests. These maximum
values are close to the air-entry values reported in
Table 1. At these stress states, the air pressure was
2.4 Test protocols
let decaying at constant air volume of the inlet line.
Protocols for air injection test on both materials are The recovery lines were initially full of water and the
analogous. A pre-conditioning path – water undrained controller was kept at a constant pressure of 0.5 MPa
loading path – was followed to approximately restore in both set-ups. However, the PVC of the recovery
the in situ effective stress of the samples. BC sam- system was not able to keep this constant pressure
ples were vertically loaded to 3 MPa in the oedometer condition when the air flow was very high, and an
cell, while BD samples were isotropically loaded to increase in the downstream pressure occurred (up to a
10 MPa. Afterwards, samples were put in contact with maximum pressure of 1.8 MPa controlled by a pressure
their synthetic water – prepared according to De Craen release valve). Finally, samples were unloaded under
et al. (2004) and Mäder (2011), respectively – at atmo- undrained conditions.
spheric pressure. Then, the water permeability of the After each injection test, MIP tests were performed
samples was determined under different pressure gra- to study the changes in the pore size distribution
dients and at constant total vertical/isotropic stress. of the tested samples. Details of the test protocols
Afterwards, a drained loading stage was followed to can be found in Romero et al. (2012), Romero &
a total vertical stress of 6 MPa in BC and to a total Gomez (2013), Romero & Gonzalez-Blanco (2015)
isotropic stress of 15 MPa for BD. Just before the air and Gonzalez-Blanco et al. (2016).
717
Figure 7. Time evolution of pressures at the boundaries
Figure 6. Time evolution of pressures at the boundaries (injection pressure in continuous line, outflow pressure in
(injection pressure in continuous line; outflow pressure in dashed line), axial strain and outflow volume for two injec-
dashed line), axial strain and outflow volume for two injection tion tests on BD samples at two different injection rates: 2 and
tests on BC samples at a rate of 2 mL/min and two different 100 mL/min. Test at the fastest rate from Romero & Gomez
orientations of bedding planes (normal and parallel to air (2013).
flow).
Figure 7 presents the comparison of the behaviour

3 EXPERIMENTAL RESULTS along air injection and dissipation stages for BD sam-
ples tested at two different controlled-volume rates
Selected results of air injection and dissipation stages (2 mL/min and 100 mL/min) and at constant isotropic
at constant vertical / isotropic stress are presented in stress (15 MPa). Both samples were tested with the
Figure 6 for BC and Figure 7 for BD. The figures bedding planes orthogonal to air flow. Regarding their
show the time evolution of the air injection pressure volume change response, they displayed some small
at the upstream boundary, the outflow pressure and expansion during the early air injection stage (A to B
volume at the downstream boundary, jointly with the in Fig. 7). Expansion continued in both samples after
axial strain, calculated based on the recorded axial shut-off (point B in the figure), as the air pressure front
displacement. propagated into the sample, causing an increase in the
Figure 6 corresponds to two tests performed on fluid pressure and a decrease of the constitutive stress.
BC with bedding planes orientated parallel (BC// ) The time at which first air outflow occurred depended
and orthogonal (BC⊥ ) to air flow, at the same injec- on the injection rate. In sample tested at 100 mL/min,
tion rate, 2 mL/min. The air pressure at the upstream the first outflow was detected during the dissipation
boundary increased from 0.5 to 4 MPa (A to B in the stage, whereas in sample tested at 2 mL/min, it already
figure), followed by shutoff (point B) and dissipation at occurred during the last phase of the injection stage.
closed air injection line (B to C). The increase in injec- Air injection experiments on these two formations
tion pressure was accompanied by expansion (negative displayed common behavioural features. The air pres-
axial strains), followed by compression strains along surisation process acted as an unloading stage at
the dissipation stage. The sample BC⊥ underwent constant vertical or isotropic stress, inducing some
larger axial strains consistent with the anisotropic expansion of the samples. For BD samples, the expan-
behaviour observed on loading experiments. Sample sion of the samples presented some delay while the
BC// underwent breakthrough (the outflow volume air pressure front propagated inside the samples at the
rapidly increased) just after attaining the maximum two selected rates. Results on BC showed that all the
injection pressure. On the contrary, sample BC⊥ was expansion occurred during the injection, since the pore
able to sustain the maximum pressure for a longer pressures inside the samples were nearly equilibrated.
period after shut-off (point B). After breakthrough, the The difference in the time at which expansion occurred
injection pressure started to decrease along the dissi- could be explained by the difference in porosity /
pation stage towards the value indicated by point C. permeability of both samples (refer to Table 1).
During this dissipation stage, the constitutive stress To better understand the consequences of air injec-
(total vertical stress minus air pressure) increased, tion, MIP tests were performed on freeze-dried sam-
leading to progressive compression of the material. ples to compare the pore size distributions before
718
Figure 8. Pore size distribution changes before and after air
injection tests in BC.
(intact state) and after the air tests. Figure 8 presents

the pore size density functions obtained for the intact
Boom Clay and after the air injection tests. Special
care was taken during the unloading stage (after the
air tests) that was carried out under undrained con-
ditions to prevent further expansion of the material.
A new family of large pores, which was not detected
on intact samples, was systematically observed after
the air injection tests. These new pores at entrance Figure 9. Computed versus measured in the ZFD: a) injec-
sizes larger than 2 µm (dominant pore modes around tion and recovery pressures (gauge pressures); b) average
10 µm and representing around 10% of the total pore axial strains; c) outflow volumes. BC with bedding planes
volume) were associated with fissure opening that parallel to air flow (2 mL/min).
acted as preferential air pathways. The figure also
includes the pore size distribution of a sample, which (undisturbed clay) and a vertical central Zone of Frac-
followed the same loading path but without the air ture Development (ZFD) with a thickness of 2 mm.
injection/dissipation stages, therefore subjected just to To properly simulate the test, it was necessary to
loading and fast undrained unloading. The large pores include both the injection and recovery systems into
were not detected on this sample, which suggested that the model, which corresponded to the drainage lines
they were actually related to the air injection/migration of the experimental set-up.
process. Similar patterns were also found in BD The mechanical model adopted for both zones is
samples after the tests (Romero & Gonzalez-Blanco the elastoplastic model BBM (Alonso et al. 1990). For
2015). the retention properties of the matrix material the van
Genuchten’s model was adopted. The diffusion coef-
ficient was selected according to Jacops et al. (2015)
4 NUMERICAL MODELLING for samples with bedding planes parallel to the flow
direction. Kozeny’s model was used for the intrinsic
To help with the interpretation of the results in terms permeability changes of the matrix as a function of
of injection and outflow pressures, as well as outflow void ratio. The required parameters for the embedded
volumes and axial strain, numerical simulations were fracture permeability and water retention model were
performed by implementing the geometry of BC in the fitted by using experimental data. An initial aperture
oedometer cell and their corresponding boundary con- of 100 nm – slightly higher than the dominant pore
ditions on the injection and outflow sides. The coupled mode of the matrix – was selected. The set of mechan-
mechanical and two-phase flow equations were dis- ical and hydraulic parameters adopted is detailed in
cretised and solved using the finite element program Gonzalez-Blanco et al. (2016).
Code_Bright (Olivella et al. 1996). An embedded frac- The results of the computed injection and outflow
ture permeability model (Olivella & Alonso 2008) in a pressure response in the ZFD as a function of time
fully coupled hydro-mechanical approach was adopted together with the outflow volume and the average axial
to simulate the gas injection tests. This model takes strain along the sample height are shown in Figure 9
into account the variation of the intrinsic permeability and compared with experimental results for BC tested
and the air entry pressure with fracture aperture, which with bedding planes parallel to air flow at an injection
depends on strain. rate of 2 mL/min. The figure shows that the air pressure
A 2D axisymmetric representation of the sample decay at the bottom of the sample during the dissipa-
geometry was selected presenting two zones hav- tion stage is acceptably well fitted. Good agreement is
ing very different hydraulic properties: the matrix also found on the fluid pressure at the top, computed
719
Figure 10. Simulated distribution of absolute gas pressure (left, in MPa), porosity (middle) and liquid degree of saturation
(right) during the air injection (t = 150 min), at shut-off (t = 245 min) and during the dissipation (t = 600 min).
as the maximum between air and water pressures. The

computed sample volume change represented by the
average axial strain is also reasonably well reproduced.
The sample undergoes expansion during the injection
stage followed by compression during air dissipation.
The computed results show the same expansion, while
a slightly larger compression than the measured one
is predicted. Moreover, the time in which the out-
flow takes place compares well with the measured
one. The outflow volume, computed as the sum of
water and air volumes, fits well with the experimental
measurements.
The local sample response is depicted in Figure 10,
with contour plots to better understand the influence
of the embedded fracture response. Absolute gas pres-
sure, porosity and liquid degree of saturation are shown
at three different time steps: 150 minutes after the start
of gas injection; at the end of the gas injection (shut-
off); and during the dissipation stage. On the one hand,
only when the air pressure increased enough, the frac-
ture opened and desaturated allowing the air to flow.
On the other hand, as the air pressure decreased due
to the outflow, the fracture gradually closed over time.
The matrix of the clay remained fully saturated after
the air passage. Indeed, within the matrix, the dominant
transport mechanism was the diffusion of dissolved air.
Figure 11 shows the contribution of the diffusive and
advective flows in the matrix and in the ZFD at the
same elapsed times previously indicated. The domi-
nant advective flow in the ZFD is clearly observed in
the figure.
5 CONCLUSIONS Figure 11. Diffusive and advective fluxes along the

core height in the ZFD and in the matrix at air injec-
Experimental work is important for understanding the tion (t = 150 min), shut-off (t = 245 min) and dissipation
process of gas flow through saturated argillaceous (t = 600 min) stages.
720
rock formations (Boom Clay and Brown Dogger). The Dao, L.Q., Cui, Y. J., Tang, A. M., Pereira, J.M., Li, X.L. &
tests presented corresponded to a series of gas pulse Sillen, X. 2014. Investigating the anisotropy of the shear
tests (relatively fast air injection tests) designed to modulus of natural Boom Clay. Géotechnique Letters 4:
investigate the mechanisms of gas flow at different 98–101 (doi: 10.1680/geolett.14.00015).
De Craen, M., Wang, L.,Van Geet, M. & Moors, H. 2004.
orientations of the bedding planes and at different vol- Geochemistry of Boom Clay pore water at the Mol site.
umetric air injection rates (2 and 100 mL/min). The Scientifict. Report. SCK·CEN-BLG-990.
main focus was given to the coupled hydro-mechanical Ferrari, A., Favero, V., Marschall, P. & Laloui, L. 2014.
response by measuring displacements during the air Experimental analysis of the water retention behaviour
injection and dissipation stages. of shales. International Journal of Rock Mechanics &
The air injection tests were performed in oedome- Mining Science 72: 61–70.
ter/isotropic cells at constant vertical/isotropic stress, Gonzalez-Blanco, L., Romero, E., Jommi, C., Li, X. & Sillen,
on samples with pre-defined orientation of bedding X. 2016. Gas migration in a Cenozoic clay: experimental
planes. The deformation response during the process results and modelling. Geomechanics for Energy and the
Environment. In press (doi: 10.1016/j.gete.2016.04.002).
was fully coupled with the hydraulic process. The sam- Jacops, E., Wouters,K., Volckaert, G., Moors, H., Maes, N.,
ples underwent expansion during the air injection, and Bruggeman, C., Swennen, R. & Littke, R. 2015. Mea-
compression when the air pressure decayed. Large suring the effective diffusion coefficient of dissolved
amounts of fluid volume were measured at the down- hydrogen in saturated Boom Clay Applied Geochemistry
stream reservoir which indicated the break-through 61: 175–184 (doi: 10.1016/j.apgeochem.2015.05.022).
point was exceeded. Mercury Intrusion Porosimetry Lima, A., Romero, E., Piña, Y., Gens, A. & Li X. 2012.
(MIP) tests allowed detecting a new family of large Water retention properties of two deep Belgian clay forma-
pores with entrance sizes larger than 2 µm after the tions. In Unsaturated Soils: Research and Applications:
air injection tests, which suggested the development of 179–184 (doi: 10.1007/978-3-642-31116-1_24).
Mäder, U. 2011. Recipe and preparation of a simplified arti-
preferential paths during the air injection experiments. ficial pore water for Opalinus Clay and Brown Dogger.
Selected experimental results were simulated using AN 11-159 Nagra, Wettingen, Switzerland.
a coupled hydro-mechanical finite element code Marschall, P., Horseman, S., & Gimmi,T. 2005. Characteri-
(Code_Bright), which incorporated elements with an sation of Gas Transport Properties of the Opalinus Clay, a
embedded fracture permeability model to account for Potential Host Rock Formation for Radioactive Waste Dis-
the simulation of the dominant single-phase (air) flow posal. Oil & Gas Science and Technology 60(1): 121–139
along preferential pathways. Rock intrinsic permeabil- (doi:10.2516/ogst:2005008).
ity and its retention curve were made dependent on Olivella, S. & Alonso, E.E. 2008. Gas flow through
fracture aperture changes based on experimental data clay barriers. Géotechnique 58(3): 157–176 (doi:
10.1680/geot.2008.58.3.157).
(an initial aperture of 100 nm was selected based on Olivella, S., Gens, A., Carrera, J. and Alonso, E.E. 1996.
MIP). A central zone of fracture development was con- Numerical formulation for a simulator (CODE_BRIGHT)
sidered to allow for the initiation of air flow pathways. for the coupled analysis of saline media. Enginnering.
Results of the simulation showed encouraging agree- Computations 13(7): 87–112 (doi: 10.1108/02644409610
ment not only with recorded upstream/downstream 151575).
pressures and outflow volumes, but also in the vol- ONDRAF/NIRAS 2013. Research, Development and
ume change response of the material.The experimental Demonstration (RD&D) Plan for the geological disposal
results, combined with the numerical simulations, proof high-level and/or long-lived radioactive waste includ-
vided good insight into the role of the volumetric ing irradiated fuel of considered as waste, State-of-the-art
report as of December 2012. ONDRAF/NIRAS, report
response and the hydraulic changes on the air transport NIROND-TR 2013-12 E.
properties of the samples. Romero, E., Senger, R., Marschall, P. & Gómez, R. 2012. Air
tests on low-permeability claystone formations. Experi-
mental techniques, results and simulations. In Interna-
ACKNOWLEDGEMENTS tionalWorkshopAdvances in MultiphysicalTesting of Soils
and Shales: 69–83 (doi: 10.1007/978-3-642-32492-5_6).
The authors are grateful to the Belgian and Romero E. & Gómez R. 2013. Water and air permeability tests
Swiss agencies for radioactive waste management – on deep core samples from Schlattingen SLA-1 borehole.
ONDRAF/NIRAS and NAGRA, respectively – for NAB 13–51, Nagra, Wettingen, Switzerland.
Romero, E. & Gonzalez-Blanco, L. 2015. Complemen-
their financial support. Thanks are also expressed tary water and air permeability tests on core samples
to Prof. Sebastià Olivella for his valuable comments from Schlattingen SLA-1 borehole. NAB 15–06 Nagra,
regarding the numerical modelling. Wettingen, Switzerland.
REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive
model for partially saturated soils. Géotechnique. 40(3):
405–430.
721

Frank Wuttke, Sebastian Bauer Marcelo Sanchez - Energy Geotechnics-Crc Press (2016) PDF

Uploaded by

Copyright:

Available Formats

Frank Wuttke, Sebastian Bauer Marcelo Sanchez - Energy Geotechnics-Crc Press (2016) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Frank Wuttke, Sebastian Bauer Marcelo Sanchez - Energy Geotechnics-Crc Press (2016) PDF

Uploaded by

Copyright:

Available Formats

COPYRIGHT © 2016 TAYLOR and FRANCIS GROUP, LONDON, UK CRC PRESS / BALKEMA - PROCEEDINGS AND MONOGRAPHS IN ENGINEERING, WATER

TER AND EARTH SCIENCES

Editors: Frank Wuttke, Sebastian Bauer and Marcelo Sánchez

INTERNATIONAL ADVISORY BOARD

Tony Amis (UK) Fernando Pardo (Spain)

NATIONAL TECHNICAL PROGRAM COMMITTEE

Frank Wuttke, Kiel University, Germany

Henok Hailemariam, Hem Motra and Frank Wuttke

Malek Bouazza Enrique Romero

Kiel University, Germany

ISSMGE TC 308 on Energy Geotechnics

APS Antriebs-, Prüf- und Steuertechnik

Potential of district-scale geothermal energy in urban cities

K. Soga, Y. Zhang & R. Choudhary

1 INTRODUCTION utilise the heat convection mechanism of groundwa-

Parameter Unit Value Justification

Coefficient of / 3.3 Kavanaugh and

Figure 2. Borehole allocation map.

out of a square cylinder with sides equal to 6 metres

red colour proportion, which stands for the C/D ratio

Multiphysical phenomena and mechanisms involved with energy piles

Lyesse Laloui & Alessandro F. Rotta Loria

1 INTRODUCTION obtained through laboratory experiments dealing with

Figure 4. Evolution of vertical strains in the energy pile

piles (labelled P1-16 in Figure 7 (b)) below a heav-

than the expansion that this pile would have undergone

3.1.2 Observed behaviour: monotonic temperature

4 ANALYSIS AND DESIGN OF ENERGY PILES

4.1 Geotechnical and structural performance

from the configuration of the pipes. However, a length-

Geomechanical and numerical modeling of gas hydrate sediments

M. Sánchez & X. Gai

1 INTRODUCTION and geomechanical properties of hydrate sediments.

2.3 Main equations

Figure 4. Yield surfaces considered in the model.

Test Test Test Test

M 1.30 1.30 1.30 1.30

where a and γ are model constants; n is the parameter

structure is huge during dissociation. The main rea-

Coupled chemo-mechanics: A comprehensive process modeling

1 INTRODUCTION three decades of experience in thermo-mechanics

requires that a reversible part of mass change, µel

Or through a Legendre transform one can employ

As a result the chemo-elasticity equations can be

4 CHEMICALLY INDUCED SEDIMENT

In this section we shall discuss modeling of chemically

solid grain mineral. Stress and strain are microstress

In particular, at constant stress, an irreversible strain

where XH4 SiO4 is molar fraction of aqueous silica in the

5 CHEMO-MECHANICALLY INDUCED its variable part of ua as a function of time depending

Table 1. Characteristics dimensions of the flow conduit.

Name Symbol Value

As a result of linearization of eq. (x), the coating

The two variables shown in Fig. 5 control the overall

where µ is dynamic viscosity of the pore fluid, and

Figure 8. Stress–strain curve from Uniaxial Compression

the increase of stiffness is concerned (Hu and Hueckel,

Thermomechanical properties of a new small-scale reinforced

A. Minto, A.K. Leung, D. Vitali & J.A. Knappett