Proceedings of TC209 Workshop - Www-Version

ICSMGE Workshop Proceedings_Cover_crop_marks.
pdf 1 2017-09-11 15:45:04
ISSMGE Technical Committee TC 209

Offshore Geotechnics
Proceedings of TC 209 workshop

19th ICSMGE - Seoul, 20 September 2017
Foundation Design of Offshore Wind Structures
CM
MY
CY
CMY
Edited by: Yunsup Shin

ICSMGE Workshop Proceedings_Cover_crop_marks.pdf 3 2017-09-11 15:45:07
ISSMGE Technical Committee TC 209

Offshore Geotechnics
Proceedings of TC 209 workshop

19th ICSMGE - Seoul, 20 September 2017
Foundation Design of
Offshore Wind Structures
Edited by: Yunsup Shin

C
CM
MY
CY
CMY
K

Advert Korean Magazines A4-Print version_NEW.pdf 1 2017-09-07 14:26:38
CM
MY
CY
CMY
K

Foreword
The International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE)
and its Technical Committee 209 (TC209) on Offshore Geotechnics are proud to support
this workshop and the publication of this volume.
The theme for this workshop is “Foundation design of offshore wind structures”.
The offshore wind industry is expanding rapidly throughout Asia. Projects are currently
being developed in China, Taiwan and South Korea, with many other countries
aggressively pursuing opportunities in this market sector. Cost reductions driven by a
mature European market and local pricing, combined with the social benefits of
generating power offshore and away from densely populated areas, makes offshore wind
an attractive economic prospect for meeting national renewable energy targets.
It is with the market situation in mind that TC209 decided to make offshore wind the
focus of this workshop. The workshop comprises a series of carefully selected papers,
each representing different subjects – and all addressing aspects of foundation design.
First is a review of the offshore wind market in Korea, honouring our hosts for this
workshop, and providing an interesting backdrop to past and future developments. The
second paper provides an overview of the need for high quality site characterisation in
support of efficient foundation design. The third and fourth papers present technical
summaries of the key design challenges for monopile and suction caisson foundations,
when used to support wind turbines. Finally, results from an instrumented jacket
supported on suction buckets are presented in detail – adding valuable observations of
actual performance, to enhance future design.
I trust that you will find the workshop both interesting and of strong technical merit.
ISSMGE TC209 is grateful to Norwegian Geotechnical Institute (NGI) for funding the
publication of this volume; to all the authors for sharing their passion, knowledge and
experience; and to the team who coordinated the preparation of this workshop, especially
Dr Yunsup Shin.
Phil Watson
Incoming Chair, ISSMGE TC-209 on Offshore Geotechnics

Editorial address
This Technical Committee 209 workshop at 19th International Conference on Soil Mechanics and
Geotechnical Engineering (ICSMGE) follows the success of the previous workshop in Paris 2013.
Offshore wind farms are constructed in many parts of the world, and many more are being
planned, inlcuding in Korea. Foundation design is an essential part of the design of the offshore
wind structures, and this TC209 workshop has placed focus on several of the important foundation
design aspects, i.e. marine site characterization, foundation design and installation considerations
for various foundation types (such as monopiles and suction caisson foundations), and case
histories of already installed structures. The offshore wind plan and strategy in Korea is also
presented.
NGI is proud to sponsor this edition of the TC 209 Workshop Proceedings which includes:
– "Geotechnical perspective on offshore wind plan, strategy, projects and research in Korea"
prepared by Bae K.T. (DAEWOO E&C), Choo Y.W. (Kongju National University), Youn
H.J. (Hongik University), Kim J.Y., Choi C.H. (KICT), and Kwon O.S. (KIOT) on behalf of
Energy Plant Technical Committee in Korea Geotechnical Society;
– "Marine site characterisation and its role in wind turbine geotechnical engineering" prepared
by Rattley M., Salisbury R., Carrington T., Erbrich C., and Li G. (Fugro);
– "Design aspects for monopile foundations" prepared by Burd H. J., Byrne B. W., McAdan R.
A., Houlsby G. T., Marten C. M., Beuckelaers W. J.A.P. (Oxford University), Zdravković L.,
Taborda D.M.G., Potts D. M., Jardine R. J. (Imperial College), Gavin K., Doherty P., Igoe D.
(Formerly University College Dublin), Gretlund J. S., Andrade M. P., and Wood A. M.
(DONG Energy) on behalf of PISA team;
– "Design Aspects of Suction Caissons for Offshore Wind Turbine Foundations" prepared by
Sturm H. (NGI);
– "Suction bucket jackets for offshore wind turbines: applications from in situ observations"
prepared by Shonberg A., Harte M., Aghakouchak A., Brown C. S. D., Andrade M. P., and
Liingaard M. A. (DONG Energy)
We believe that the papers collected in a single publication will provide the offshore geotechnical
engineers with unique and useful information and recommendations for designing offshore wind
foundations.
We thank Marit Støvne (NGI) for her editorial assistance, Maren Kristine Johnsen (NGI) and
Kjell Hauge (NGI) with graphical design and webpage development. We are also grateful for the
helpful advice from Philippe Jeanjean (BP), Phil Watson (Fugro), Knut Andersen (NGI), Thomas
Langford (NGI), and Hendrik Sturm (NGI) with respect to planning of the workshop.
Yunsup Shin
Norwegian Geotechnical Institute
September 2017

Table of contents
Foreword 7
Editorial address 9
Geotechnical perspective on offshore wind plan, strategy, projects and research in 13

Korea
K.T. Bae, Y.W. Choo, H.J. Youn, J.Y. Kim, C.H. Choi, O.S.Kwon
Marine site characterisation and its role in wind turbine geotechnical engineering 21
M. Rattley, R. Salisbury, T. Carrington, C. Erbrich, G. Li
Design aspects for monopile foundations 35

H.J. Burd, B.W. Byrne, R.A. McAdan, G.T. Houlsby, C.M. Marten,
W. J.A.P. Beuckelaers, L. Zdravković, D.M.G. Taborda, D.M. Potts, R.J. Jardine,
K. Gavin, P. Doherty, D. Igoe, J.S. Gretlund, M.P. Andrade, A.M. Wood
Design Aspects of Suction Caissons for Offshore Wind Turbine Foundations 45

H. Sturm
Suction bucket jackets for offshore wind turbines: applications from in situ observations 65
A. Shonberg, M. Harte, A. Aghakouchak, C.S.D. Brown, M.P. Andrade,
M.A. Liingaard

Geotechnical perspective on offshore wind plan, strategy, projects, and research in
Korea
Le plan, la stratégie, les projets, et la recherche des éoliennes en mer en Corée, d'un point de vue
géotechnique
Kyung-Tae Bae
Chair of Energy Plant Technical Committee, Korean Geotechnical Society
Daewoo Institute of Construction Technology, DAEWOO E&C, Republic of Korea
Yun Wook Choo

Secretary of Energy Plant Technical Committee, Korean Geotechnical Society
Dept. Civil & Env. Eng., Kongju National University, Republic of Korea, [email protected]
Heejung Youn
School of Urban and Civil Eng., Hongik University, Republic of Korea
Jin Young Kim & Changho Choi

Korea Institute of Construction Technology, Geotechnical Engineering Research Institute, Republic of Korea
Osoon Kwon
Coastal Disaster Prevention Research Center, Korea Institute of Ocean and Technology, Republic of Korea
ABSTRACT: This paper introduces the results and current status of the Korean offshore wind technology development from the
viewpoint of geotechnical engineering. First, Korea's offshore wind resources and geographical/geotechnical conditions are
introduced, followed by construction records for offshore wind farms development, and the research and development projects of the
substructure for offshore wind turbine installation.
RÉSUMÉ : Cet article présente les résultats et l'état actuel du développement de la technologie éolienne en mer coréenne du point de vue
de l'ingénierie géotechnique. En premier lieu, les ressources éoliennes en mer en Corée et les conditions géographiques/géotechniques sont
présentées, suivies des archives de construction pour le développement de parcs éoliens en mer et des projets de recherche et
développement de la sous-structure pour l'installation d'éoliennes en mer.
KEYWORDS: Korea, wind energy resource, geological conditions, construction records, research and development.
1 INTRODUCTION country, interest in offshore wind power has been increasing,

and the relevant research, development, and projects have been
Along with the greenhouse gas mandatory reduction goals set carried out.
forth in the Kyoto Protocol in 1997, Korean government has In this paper, the past records, progress and current status of
established a new and renewable energy policy and, in addition, the offshore wind technology development in Korea is
conducted research and development projects as well as pilot presented from the geotechnical engineering perspective. The
projects led by the public sector. structure of this paper is as follows: Section 2 introduces
Among various renewable energy sources, wind power is Korea's offshore wind resources and geographical/geotechnical
known as an energy source that meets the geographical/ conditions. Section 3 introduces the construction performance
geotechnical conditions of the Korean Peninsula (MOTI and of the offshore wind farms development. Finally, Section 4
KEA, 2016). From a geographical point of view, onshore wind introduces the research and development projects of sub-
power has a very great advantage in the mountainous region structure technologies for offshore wind turbine installation.
where the northwest wind develops, and offshore wind power
has a great advantage in terms of the effective transportation
and utilization of production power, since the three sides of the 2 WIND ENERGY RESOURCE AND GEOLOGICAL
Korean Peninsula are surrounded by the sea and the mega-cities CONDITIONS
(Seoul, Busan, Incheon etc.), which are major energy con-
sumers, are located at the coastal line of Korea. 2.1 Wind energy resources
Heavy Industries already have great technological capa-
bilities and infrastructure for the design, manufacture, and Domestic offshore wind potential capacity is estimated to be
installation of wind power generation facilities and support 426 GW and technically feasible capacity is estimated to be
structures. It is attributed to the abundant technological 33.2 GW (MOTI and KEA, 2016). This corresponds to 17.8%
capabilities of the shipbuilding/marine industry and the of the total power generation in Korea. Although Korea is a
established supply chains for production of the wind power peninsula surrounded by sea with abundant offshore wind
plant. However, because the environmental damage caused by resources, all resources cannot be utilized due to economic and
installing onshore wind power generator is rising and there are technological limitations; furthermore, the offshore wind farms
local limitations on generator installations in such a small
13
Proceedings of TC 209 Workshop - 19th ICSMGE, Seoul 20 September 2017
Foundation design of offshore wind structures
are constructed considering wind speed, wind density and water
depth. Figure 3 shows the locations of the offshore wind farms in
Figure 1 shows the domestic weather map provided by the the west coast, HeMOSU-1 Met mast, and the ground
meteorological resources map (www.greenmap.go.kr). The investigation locations. HeMOSU-1 has measured the weather
wind speed measured at 80 m above the ground is about 7.0-7.5 information for three years, showing the monthly average wind
m/s near the Western coast, mostly 7.5 m/s or above at the speed of 7.0 m/s from winter to early spring and less than 5.0
coasts of Jeju Island, and 8.5 m/s or above in other areas. m/s (which is the effective limit wind speed) in May and June,
Since the offshore wind farm becomes economically inferior and then increases again in July (Figure 4). The average annual
at 30 m sea level or above, the East Sea, where the water depth wind speed is 6.97 m/s at 97 m from sea level and 6.71 m/s at
drastically deepens as it gets farther from the coast, is 76 m and does not drastically change with height.
considered to be inappropriate despite its satisfactory wind The mean wind speed distribution per year for the past 30
speed, and the Southwest Sea and Jeju Island coastal areas with years, back-estimated from the observations of HeMOSU 1, is
moderate changes in water depth are evaluated as suitable sites shown in Figure 5, and it is expected that the average wind
for offshore wind farms. Typical water depths for the distribution per month is 6.5–7.0 m/s per month according to
developing or soon-to-be-developed offshore wind farms are this estimation.
10-20 m.
As shown in Figure 2, the main wind direction in the area
considered as potential sites for offshore wind farm power, is
the northwest. This major northwest wind is mixed with the Buan-gun
northeast wind in the waters near Jeju Island and southwest
coast. The wind turbine power generation efficiency is
Wi-do
improved when the wind speed of 5 m/s or more is maintained, Verification
site
and the number of days with wind speed of 5 m/s or more is
measured to be approximately 55% of the year in the west coast
and 65% or more in the coastal area near Jeju Island. Gochang-gun
Yeonggwang-
gun
Anma-do
8.5 or higher
8.0 – 8.5
7.5-8.0
7.0-7.5
6.5-7.0
Figure 3. Locations of offshore wind farm, HeMOSU-1 met mast tower,
6.0-6.5 and site investigation (modified from KEPRI, 2014).
5.5-6.0
5.0-5.5
4.5-5.0
4.0-4.5
4.0 or lower
(unit: m/s)
Figure 1. Wind speed map at the elevation of 100 m from sea level
(modified from www.greenmap.go.kr).
Figure 4. Mean wind speed measured at HeMOSU-1 met mast tower

(KEPRI, 2014).
Northwest
West
Southwest
South
Southeast
East
Northeast
North
Figure 5. Annual wind speed for 30 years estimated by MCP method

using HeMOSU-1 data (KEPRI, 2014).
Figure 2. Main wind direction distribution (modified from www.

greenmap.go.kr)
14
Geotechnical perspective on offshore wind plan, strategy, projects, and research in Korea
2.2 Site investigation wind farm site, and thus, a supporting structure type is selected
by this soft rock layer. Because Jeju Island is an island formed
This section explains the ground conditions of the Southwest by volcanic activity, the geological structure near Jeju Island is
offshore wind farm (verification phase) whose construction almost identical.
inaugurated in early 2017, and the Jeju Tamra offshore wind Figure 7 shows the stratigraphy of the Jeju Tamra offshore
farm (30 MW) which has been producing electricity since 2016. wind farm area. It was confirmed with five boreholes that there
A total of 11 boreholes were drilled in the west coast of Korea is 0–0.3 m thin marine sediments (classified as SP) on the
(8 holes around HeMOSU-1 and 3 holes in the farm area for the seabed and basaltic soft rock or caustic rock layer underneath.
verification phase (as shown in the Figure 3) and the The water content of marine sediments was 24.1% with the
stratigraphy of the site was constructed using the results of specific gravity of 2.73, the fine contents passing #200 sieve
Standard Penetration Test (SPT). was 4.26%, and the N value was 4. However, since the layer is
According to Yoon et al. (2014), there is a marine thin, the physical properties of the soil layer do not seem to
sedimentary layer composed of marine clay (ML, CL) and affect the underlying foundation design. The degree of
marine sandy soil (SM, SP, SW-SM, GP) near the seabed, and weathering and joint spacing of the soft rock layer with basalts
weathered residual soils, weathered rocks, and bedrocks are as parent rocks was widely distributed with the RQD of 54-80%,
followed in order underneath. The thickness of the marine and a hard rock layer. The monopile was designed to penetrate
sediments above the bedrocks is about 24.3–62.5 m and tends to into the soft rock layer whose uniaxial compressive strength
thicken toward the outer sea. The undrained shear strengths ranges from 24.7 to 26.9 MPa and the deformation modulus is
were derived from the Cone Penetration Tests (CPT) performed 2,552 MPa.
near the verification wind farm, displaying 10 kPa or below
near the seabed, yet increasing as it gets deeper, with the
average value of 33.9 kPa.
On the other hand, the results from the drilling investigation
near HeMOSU-1, where the offshore wind farm will be
constructed, are somewhat different. Figure 6 shows the results
of holes Nos. 7 and 8 among the 8 holes, with similar results
from the other boreholes. There is a silt-mixed sand of
approximately 4 m from the seabed, and the N value in the
layer is about 40. Below the layer, there is approximately 25 m
thick very soft silty clay with the N value of less than 10,
followed by a very dense sand layer with the N value of over 50
at depth of 30 m or deeper, and a layer of soft rock is located at
a depth of 40 - 50 m. Therefore, the stratigraphic near the
HeMOSU-1 is very different from the verification wind farm.
Foundation types of offshore wind farm near HeMOSU-1 seem
to be greatly affected by the soil layer with a soft clay of Figure 7. Stratigraphy of Jeju Tamra offshore wind farm (Choi et al.,
thickness of 20 m or more. 2014).
3 CONSTRUCTION RECORDS
Fine sand with silt Fine sand with silt

3.1 Present conditions of wind farms
Loose-very dense Dense/ dark gray/ 6 m
Dark-gray/saturated/6.2 m
Silty clay/firm-very firm
The current status of the Korean offshore wind farms is shown
Dark gray/saturated/silt seam/5.3 m in Figure 8 and Table 1. Currently, a total of 38 MW is in
Fine sand with silt
Medium dense/dark gray-brown/1.5 m Silty clay operation with 35 MW in Jeju Island and 3 MW in Gunsan.
Silty clay with sand, medium stiff/dark Very soft – firm
gray/2.5 m Dark gray
Silt seam
About 2,500 MW in the Southwest Sea is under construction
Silty clay 21 m and 4,686 MW is being prepared.
Firm-very firm
Dark gray
10.9 m
Fine sand with silt

Dense-very dense
Dark gray-light gray Coarse sand with gravel
7.6 m Very dense
Dark gray
12 m
Sand
Very dense
Brown-light gray Weathered rock/2 m
12 m
Weathered soil
Fully weathered from bed rock
Disintegration into silty sand by
impact
Very dense/light green/wet/11 m
Weathered rock/2 m Soft rock
Soft rock
Figure 6. Boring log at Nos. 7 and 8 of southwest sea offshore wind

farm (modified from KEPRI, 2014).
The Jeju Tamra offshore wind Farm located on the

Figure 8. Performance and plan of offshore wind farms.
northwestern coast of Jeju Island has a soft rock layer at
shallow depth, which is different from that of the southwest
15
Table 1. Present and planned wind farms. 3.3 Jeju Tamra
Capa Jeju Tamra offshore wind farm was commissioned by Korea
Substru- Installation
Locations -city Owner
cture Year Southern Power Co., Ltd. and Doosan Heavy Industries &
(MW) Construction consortium completed installation in 2016. As
Korea Institute of shown in Figures 10 and 11, ten units of 3 MW of Doosan
Jeju Island
Energy Research, Heavy Industries & Construction's WinDS3000 are in operation
5 Doosan Heavy Jacket 2011 and the tower length is 65.1 m. The depth of the water is about
(Woljeong)
Industries & 20 m and the upper structure is designed to be supported by the
Construction Co., Ltd. jacket substructure through a pin pile to the shallow depth of
Korea Southern about 15 m of Jeju Island area basalt rock.
Power Co., Ltd.
Jeju Island
30 Doosan Heavy Jacket 2016
(Tamra)
Industries &
Construction Co., Ltd.
KEPCO Research
Gunsan 3 Suction 2016
Institute
Southwest Korean Offshore Jacket,

2,500 2017~
Sea Wind Power Co., Ltd. Suction
Jeju Island Korea Southern Under

200 NA
(Dae-jeong) Power Co., Ltd. preparation
Jeju Island KEPCO

Under
150 Engineering & NA
(Hanrim) preparation
Construction Co., Inc.
Jeju Island Halla Wind Energy Under

105 NA
(East) Co., Ltd. preparation
Busan Korea Southern Under

35 NA
(Mokdo) Power Co., Ltd. preparation
SK Construction Under Figure 10. Jeju Tamra offshore wind farm.

Ulsan 196 NA
Co., Ltd. preparation
Under
Jeonam 4,000 Jeollanam-do NA
preparation
3.2 Jeju Woljeong

Jeju Woljeong offshore wind test site was completed in 2011
using jacket substructures. A total of 5 MW is in operation
including a single unit of 2 MW of the Korea Institute of
Energy Research and a single unit of 3 MW of Doosan Heavy
Industries & Construction. The turbine installed at 2 MW is
HARAKOSAN's Z72 with the tower length of 51 m and weight
of 170 ton. Meanwhile, the turbine installed in 3 MW is Doosan
Heavy Industries & Construction's WinDS3000 with the tower
length of 61 m and weight of 190 ton.
Figure 11. Jeju Tamra offshore wind farm monitoring center.
3.4 Gunsan
Gunsan suction foundation test site was commissioned by
KEPCO Research Institute, and the installation was completed
in 2016 (Figure 12). A 3 MW WinDS3000 by Doosan Heavy
Industries & Construction is in operation and its tower length is
58.5 m. The water depth is about 20 m, and the upper structure
with a total weight of 342 ton was designed to be supported by
the tripod suction foundation in the sandy soil; the tower weight
is 155.8 ton, nacelle weight is 128 ton, hub weight is 28 ton,
and each blade weighs 10 ton. As shown in Figure 13, the
Figure 9. Jeju Woljeong offshore wind test site. suction foundation is composed of three steel suction bases with
a diameter of 6 m, a height of 12 m, and a weight of 148 ton. It
is inclined up to 0.33° during installation as shown in Figure 14,
but by controlling the output of the suction pump, the verticality
is secured up to 0.02° after completion of installation.
16
As of July 2017, the first jacket substructure is under

construction as the first phase (Figure 16). The first verification
phase was designed such that Doosan Heavy Industries &
Construction's WinDS3000 Turbines would be supported by 19
jacket substructures and one tripod suction foundation. The
water depth is about 10 m, and the supporting rock bed appears
at 40 m or deeper.
Figure 12. Gunsan suction foundation test site.
Figure 15. Location of the Southwest Sea wind farm
Table 2. Construction plan of Southwest Sea wind farm.
Phase 2
Phase 1 Phase 3
Phases (Demon- Total
(Verification) (Large scale)
stration)
Figure 13. Panoramic view of tripod suction foundation. Capacity

60 440 2,000 2,500
(MW)
Substructure Jacket, Suction NA NA
Installation
2017∼2019 2020∼2022 2023∼2024
Period
Figure 14. Verticality control using suction pump.
3.5 Southwest Sea wind farm

The Southwest Sea offshore wind farm is located between Wi-
do and Anma-do in Jeollanam-do province and was commi-
ssioned by the Korean Offshore Wind Power Co., Ltd. (Figure
15). As tabulated in Table 2, a total of 2,500 MW are planned to
be constructed with 60 MW in the first phase, 440 MW in the
second phase, and 2,000 MW in the last phase using the
integrated SCADA system developed by the KEPCO Research Figure 16. Panoramic view of the pin pile rock excavation for the jacket
Institute for management. substructure.
17
4 RESEARCH AND DEVELOPMENT OF SUB- substructure system and advanced design technology, high
STRUCTURE IN KOREA durability materials for marine structure and utilization tech-
nology, technology for securing bearing capacity of foundation,
In Korea, research related to wind power generation have been and finally, verified the technology by applying to a test bed.
continued since 1988, and the focus was on domestic wind With regard to the technology for securing support capacity
turbine technology development including development of core of the foundation structure, the analytical technique and the
parts with investing about 271 billion Korean Won (KRW) by support capacity evaluation system of the offshore wind
2010, wind resource evaluation, power plant design and power structure were developed through the laboratory model test, the
grid. In recent years, an interest in offshore wind power has centrifugal model test, and the numerical analysis, and the
increased, and research on the development of offshore wind domestic submarine ground modeling technique necessary for
substructure has been conducted by companies and research the analysis of the offshore wind turbine response. As of 2017,
institutes (KGS 2014). This session introduces the status of the development technology is being applied to the Southwest
research and development related to offshore wind power Sea offshore wind farm (verification phase) in the southwestern
substructure. coast.
4.1 Monopile
A project of "Development of Offshore Wind Energy
Foundation Systems for Large Diameter (more than 5m) and
Deep Sea (less than 60m)" was conducted to develop a large-
diameter, excavation-type monopile system with a diameter of
5 m or more, capable of efficiently supporting large capacity
generators of 3 MW or more in marine ground conditions of 30
m depth or more, where the ground is composed of rock mass
(KICT and KIOST et al. 2015).
In detail, a rock drilling rig that excavates a rock by using a
number of small hammers was developed to improve the rock
Figure 18. 5 MW jacket substructure TP design (POSCO et al. 2017).
excavation speed as shown in Figure 17. The optimal design
guideline for offshore wind power generation monopiles in line
with LRFD-based international design standards was
developed, as well as the platform design/work tip stability 4.3 Hybrid foundation
technology and the optimal form of TP (Transition Piece) for A project titled "Development of Hybrid Substructure System
fast and precise marine construction. for Offshore Wind Turbine" has been developing a new type
In addition, a single and tripod bucket foundation system hybrid support structure system suitable and economical for the
that can be applied economically in thick soil depth (50 - 60 m) Southwest coast of Korea (KICT et al. 2013). In detail, as
was developed with design guidelines for bucket foundation. shown in Figure 19, research topics include development of
The system evolves the limit-state design method based on the economical hybrid support structure system technology through
development of intrusion device and construction system efficient combination of structural materials (steel and concrete)
capable of maintaining the accuracy of vertical angle within 1°. and foundation type, development of foundation/support
The developed large-diameter monopile technology was structure/tower connection technology, foundation type and
adopted in a detailed design for the Jeju Tamra offshore wind development of scour prevention technology for hybrid support
farm project, and the bucket foundation technology was applied structure system, development of response analysis technology
and constructed for the supporting structure of the offshore of hybrid support structure considering fluid-soil-structure
wind turbine in Jeollanam-do Province. interaction, and rapid installation method and maintenance
technology development of hybrid supporting structure.
Consultations are underway to apply the new hybrid support
structures in this R&D project to the construction of offshore
wind farms in the Southwest Sea. By solving the technical
problems to secure the support structure in the soft ground, the
cost reduction for constructing the support structure is expected
to be more than 15%.
Figure 17. Rock drilling rig with multiple hammers (KICT and KIOST
et al. 2015)
4.2 Jacket substructure

In order to support the Southwest Sea offshore wind farm
project sponsored by the Korean government, a project of
“Development of Substructure Systems for Offshore Wind
Power in Shallow Sea Water (Less than 40 m)” funded by
MOTIE (Ministry of Trade, Industry and Energy) developed a
substructure technology for 5 MW wind turbine of less than the
depth of 40 m installed on the South-western coast of Korea, as Figure 19. Offshore wind power hybrid foundation and characteristics
shown in Figure 18, and developed the new concept of fixed (modified from KICT et al. 2013)
18
4.4 Monopod concrete suction foundation

A research and development project titled "Development of
design basis and concrete technologies for offshore wind
turbine support structures" funded by MOF (Ministry of Oceans
and Fisheries) is aiming at the development of the design
standards for Korea's offshore wind support structure as well as
the design/construction guidelines for the new type of concrete
support structure applicable to the HS-139, 5 MW offshore
wind turbine by Hyosung (KR and DAEWOO E&C et al. 2017). In
the case of steel structures, the top generator is sensitive to
deformation and fatigue due to vibration, wind, and wave load,
and is less resistant to corrosion.
To overcome this, the study proposed a new type of concrete
support structure combined with a suction foundation, which is
currently being actively reviewed as one of the alternative of
offshore wind support structures due to its quick and easy
installation and economic advantages (Figure 20).
Furthermore, in this study, three compartments are
implemented inside the suction foundation and the suction Figure 21. Steel tripod suction foundation scale model (Ryu et al. 2017).
pressure can be individually controlled for each compartment so
that vertical control during construction and operation is
possible (Kim et al. 2015, Kim et al. 2016). 4.6 Deep-sea floating substructure
A project “Development of Floating Substructure/Platform for
Offshore Wind Power in Deep Water” aimed to develop the
core technology for the floating substructure and its platform
design/drying/installation/evaluation/verification as well as the
technology for mooring gear that overcomes harsh marine
environment and regarding material/method technology (PNU
et al. 2013).
Its goals are to develop the market-oriented source
technologies in Stage 1 (2011-2013), the competitive core
design skills in Stage 2 (2011-2013), and the practical
applications and monitoring, and the commercialization
technology of the developed technology in Stage 3 (2011-
2013). This study includes the development of anchor system
for deep sea considering ground characteristics which is core
technology of the floating wind power system and development
of the floating wind platform technology.
5 SUMMARY
In this paper, the trends related to offshore wind power
Figure 20. Concrete monopod suction foundation (KR and DAEWOO generation in Korea since the 2000s were introduced; in
E&C et al. 2017). particular, we discussed geological and wind-resources-related
characteristics and the present state of research and
development related to the supporting structures in near shore
4.5 Tripod steel suction foundation area of the Korean peninsula.
In Korea, the three boundary sides are surrounded by the sea,
A project titled “Seashore wind turbine construction & com- and as pointed in this paper, the offshore wind power projects
mercialization embedded suction bucket support structure (Su- are being promoted mostly on the western and southern coast,
CCESS)” has been carried out on the detailed design of support and Jeju Island. This is because the depth of the East Sea is
structure and the transfer/installation of prototype (Rye et al. relatively deep, which makes it difficult to install a fixed
2017). support structure. In the future, when the technology on floating
As shown in Figure 21, a reduced-scale model was cons- support structures would be accumulated, the project is
tructed and tested to verify its applicability. The study focused expected to proceed.
on the development of support structures suitable for mid-water This paper, prepared for the TC209 Offshore Geotechnics
depth and soft seabed of the Southwest coast of Korea, to workshop of the International Conference of Soil Mechanics
achieve economic efficiency by reducing the installation cost of and Geotechnical Engineering (ICSMGE), is expected to
offshore wind support structures and to reduce the business risk contribute to the understanding and role of the geotechnical
by shortening the marine operation period. field for the revitalization of the offshore wind power industry
in Korea.
19
6 REFERENCES
Choi C.H., Kim W.K, and Cho S.D. 2014. Detailed design on monopile
foundation used in offshore wind energy test bed, The 2nd KICT-
YU-CBNU Joint Workshop.
KEPRI. 2014. Test bed for 2.5 GW offshore wind farm at Yellow sea
interim design basis report, KEPRI report.
Korea Institute of Civil Engineering and Building Technology (KICT)
et al. 2013. Development of hybrid substructure systems for
offshore wind power. 1st year research report. Ministry of Land,
Infrastructure, and Transport. Korea.
Korea Institute of Civil Engineering and Building Technology (KICT)
and Korea Institute of Ocean Science and Technology (KIOST) et
al. 2015. Development of offshore wind energy foundation systems
for large diameter (more than 5m) and deep sea (less than 60m).
Registration number: 11-1613000-001295-01. Ministry of Land,
Infrastructure, and Transport. Korea.
Korean Geotechnical Society (KGS). 2014. Geotechnical Design of
Offshore Wind Foundation for Geotechnical Engineers. CIR, Seoul.
Kim B.W., Kim Y.S., Jin B.M., Bae K.T. and Youn H.J. 2016.
Numerical analysis on tilting control of suction pile for offshore
wind power. Journal of the Korean Geo-Environmental Society
17(9), 5-12.
Kim Y.S., Bae K.T., Lee J.P., Joung J.W. and Choo Y.W. 2015. Model
tests for tilting control of suction bucket foundation for offshore
wind turbine. Journal of Korean Society of Hazard Mitigation
17(2), 207-218.
Korean Register (KR) and DAEWOO E&C et al. 2017. Development of
design basis and concrete technologies for offshore wind turbine
support structures. No.20120093. Ministry of Oceans and
Fisheries.
MOTI and KEA. 2016. New & renewable energy white paper.
Registration number: 11-1410000-001321-11, Ministry of Trade,
Industry and Energy, and Korea Energy Agency.
POSCO et al. 2017. Development of substructure systems for offshore
wind power in shallow sea water (less than 40 m). Ministry of
Trade, Industry and Energy.
Pusan National University (PNU) et al. 2013. Development of Floating
Substructure/Platform for Offshore Wind Power in Deep Water.
Ministry of Trade, Industry and Energy, Korea.
Ryu M., Lee J., Kwag D., Bang S. 2017. Design, construction, and
installation of offshore wind turbine with tripod suction bucket
foundation. Proc. of the ASME 36th Int. Conf. on Ocean, Offshore,
and Arctic Engineering, OMAE 2017. Trondheim, Norway.
Yoon G.L., Kim S.B., Kwon O.S., and Yoo M.S. 2014. Partial safety
factor of offshore wind turbine pile foundation in West-South
Mainland sea. Journal of the Korean Society of Civil Engineers
KSCE 34(5), 1489-1504.
20
Marine site characterisation and its role in wind turbine geotechnical engineering
Caractérisation du site maritime et son rôle dans l'ingénierie géotechnique des éoliennes
Michael Rattley, Richard Salisbury, Timothy Carrington, Carl Erbrich & Gary Li
Fugro, [email protected]
ABSTRACT: The compartmentalised approach of separate geophysical and geotechnical investigations, followed by independent
engineering design that has typified offshore site developments over most of the past 50 years is changing. Offshore wind farms
(OWFs) have large spatial extents and the potential to encounter variable soil conditions at numerous turbine locations may lead to
adoption of multiple substructure concepts with differing design considerations. At the data analysis stage it is therefore common to
construct a fully informed ground model to develop the requisite understanding of the OWF site. This paper describes how integration
should be extended to include all the technical aspects of marine site characterisation, but with specific focus on the geotechnical
phases of the process. It will demonstrate the benefits of considering all aspects under a single direction over the lifetime of an
investigation, focussed on close collaboration between different specialists and end users. By this means design can be optimised,
risks can be managed and costs controlled.
RÉSUMÉ : L'approche compartimentée d'enquêtes géophysiques et géotechniques séparées, suivie d'une conception d'ingénierie
indépendante qui a typifié les développements de sites offshore au cours de la plupart des 50 dernières années, change. Les parcs éoliens
offshore (OWF) ont de vastes étendues spatiales et le potentiel de rencontrer des conditions variables du sol dans de nombreux endroits
de la turbine peut conduire à l'adoption de multiples concepts de sous-structure avec des considérations de conception différentes. À
l'étape de l'analyse des données, il est donc commun de construire un modèle de terrain pleinement éclairé pour développer la
compréhension requise du site OWF. Cet article décrit comment l'intégration doit être étendue pour inclure tous les aspects techniques de
la caractérisation des sites marins, mais en mettant l'accent sur les phases géotechniques du processus. Il démontrera les avantages de
considérer tous les aspects sous une seule direction tout au long de la vie d'une enquête, axée sur une collaboration étroite entre différents
spécialistes et utilisateurs finaux. Par ce moyen, la conception peut être optimisée, les risques peuvent être gérés et les coûts contrôlés.
KEYWORDS: site characterisation; ground model; geotechnical engineering
1 INTRODUCTION. • An evolutionary ground model to characterise site

conditions for a range of analyses;
Marine site characterisation is the understanding of geological, • A compiled site characterisation inventory to
geotechnical, environmental and metocean conditions in document and monitor project data acquisition and
relation to the planned development of a site. In concept, it is analysis requirements;
designed to provide the right information at the right project • Early identification of potential constraints and
stage to allow optimal location, design and installation of geohazards;
subsea infrastructure. Large offshore sites are required to be • Early identification of potential implications of site
characterised to develop offshore wind farms (OWFs) and conditions on project schedule and cost;
interconnector cables; the site characterisation process is a • Early identification of data acquisition and
critical component of the project development cycle. interpretation requirements;
The processes associated with the planning and execution of • Early identification of requirements for specialist
geophysical and geotechnical site characterisation for OWFs are analytical studies;
outlined in guidance notes presented by the SUT (2014). This • Prevention of surprises otherwise impacting on
paper describes how an integrated site characterisation should overall project schedule and cost; and
include all technical aspects, including: desk study, fieldwork • Provision of a tool to manage and communicate the
planning, data collection, ground model construction, and development to project stakeholders.
geotechnical parameterisation. In that sense the paper builds on
the approach presented by Thomas (2017), but with a specific The integrated approach is also analogous to the risk
focus on the impact of the wind turbine generator (WTG) and management process described by SUT (2014) with a key
offshore substation (OSS) geotechnical engineering design message: “…the level of (residual) risk is inversely
requirements on the process. Similar discussion was presented proportional to the level of knowledge and, ideally, the risk
by Evans (2011) with regard to the development of economic assigned to any development will decrease with increased
and safe offshore oil and gas facilities in geotechnically knowledge of the development area prior to the development
challenging areas. design being finalised”.
In terms of the specific geotechnical engineering design risk
management and foundation optimisation process, the benefits
2 MARINE SITE CHARACTERISATION listed above become keenest once the outline requirements of
the geotechnical design process are known, or can be
2.1 Integrated Ground Model Approach reasonably established, and can therefore be anticipated as part
Thomas (2017) presents a phased approach for integrated of the data acquisition planning. Given recent developments in
marine site characterisation, with specific reference to large geotechnical design (Byrne et al. 2015) for WTG foundations,
deepwater oil and gas developments. While not all analytical early consultation on the geotechnical design process is key to
studies incorporated within that approach may be relevant for delivering the full benefit of the integrated marine site
an OWF development, the general framework can still be characterisation approach for an OWF development.
successfully implemented. The resulting early project stage Several areas of investigation and specialist study are
benefits are therefore unchanged from those noted by Thomas: outside of the main focus of this paper (that being the ground
21
model and geotechnical characterisation process), but are no The UXO investigation will need to be completed early in
less important in the overall OWF site characterisation process. the overall site investigation process so that locations can be
These areas are considered briefly in the sections below. cleared before the first intervention work takes place. This will
often be geotechnical drilling, testing and sampling
2.2 Unexploded Ordnance
Unexploded ordnance (UXO) refers to explosive devices that
have not fully detonated, and includes mines, air-dropped
bombs, mortars anti-aircraft projectiles and grenades (Figure 1).
These items can be lying on the seafloor or buried at shallow
depth by mobile sediments and carry a risk of detonation from
impact or movement. At risk areas include historic war zones,
firing ranges and dumping grounds. There is an implied risk to
equipment and personnel during the intervention activities at
OWFs including soil testing, turbine installation cable laying
and operation.
Figure 2. The Geowing gradiometer; consisting of five fixed

magnetometers in a rigid, steerable frame. Operated by Fugro
2.3 Metocean Data

As part of the site characterisation it is essential to define the
meteorological and oceanographic (metocean) conditions
prevailing at the location. The winds, waves, current speeds and
water levels will determine the loading on the WTG structure
and hence on the foundations. The metocean requirements for
WTGs are given in standards such as BS EN 61400-3:2009.
The metocean criteria include wind speeds and direction
(Figure 3) at a series of heights above mean sea level, for a
Figure 1. A WW2 mine discovered during survey at the Horns Rev 3
offshore windfarm in 2015. The mine was detonated in 2016
range of averaging times (e.g. 1 hourly mean, 10 minute mean,
(Vattenfall, 2016) 3 second gust); significant and maximum wave heights and
associated wave periods and directions; profiles of current
speeds and directions though the water column; and water
The risk is higher for OWFs than for most other marine levels caused by tides and surge.
developments such as oil & gas fields, because of the extensive
areas of sea floor involved and the large number of structures.
In Europe, many wind farms are located in the Southern North
Sea where past conflicts have left a significant UXO legacy.
The process normally starts with a desk study, based on
available information, to assess the potential threat of UXO
being present in the area and assess the risk to the planned
works. Elements to consider include the type of UXO that
might be encountered, the local marine environment, the type of
works to be carried out and the overall development schedule.
The result of the desk study will normally be a recommendation
to carry out a geophysical survey to detect UXO followed by a
detailed survey specification.
The most common geophysical technique used to detect
UXO is magnetometry. Over the past 10 years this has
developed from using a single magnetometer to multiple
gradiometer systems that measure the magnetic gradient and
provide more accurate, detailed directional information. Figure
2 shows Fugro’s Geowing which deploys multiple
magnetometers on a fixed frame towed behind the survey vessel Figure 3. Wind rose showing joint occurrence of wind speed and
using a remotely operated towed vehicle (ROTV) that offers direction
depth control and steerability.
Data processing and interpretation is a two stage process. Metocean design criteria are typically derived using hindcast
First the data were processed to identify any anomalies, which model data as such models can provide a long enough time
are then correlated with any acoustic information available from series to allow extreme conditions to be determined with
bathymetry and side scan sonar surveys. In the second stage the confidence. Usually at least 20 years of data are used, and an
interpreted data is analysed by UXO specialists who will assess Extreme Value Analysis is performed to extrapolate the data to
the risk of the anomalies representing UXO. the required return periods (e.g. 50 years). Ideally the models
22
would be validated against site-specific measured data, and processes, the monopile eventually became the prevailing
calibrated if necessary. This is particularly important for foundation concept for WTGs, although jacket solutions are still
currents, whose complexity makes numerical modelling more popular in deeper waters (Figure 5) and for OSS structures.
challenging. Site-specific small-scale models can be set up to Monopiles now make up over 70% of foundations for WTGs
accurately account for the local bathymetry at the wind farm (Kallehave et al., 2015) and current market trends indicate that
location. These models are nested within global-scale models monopiles will continue to be the preferred foundation concept
that provide conditions at the model boundary. for WTGs in Europe in the next 10 years. The single-pile
In addition to design criteria, operational statistics are foundation is attractive due to its simplicity and robustness,
required to assist with operational planning during the allowing costs to be reduced through mass fabrication and
installation phase. Knowledge of the seasonal variation in quicker installation.
conditions means that weather-sensitive operations are planned
for times when conditions are most likely to be suitable.
2.4 Environmental study

An environmental study will be needed to address a number of
points that may include habitats, existing infrastructure such as
cables and pipelines, meteorological conditions, oceanography,
fishing, shipping movements, recreational issues, military
exercise areas, dumping grounds and possibly many others.
Precise requirements will vary depending on the regulatory and
licensing regulations and the local conditions.
The environmental study will often be a major undertaking
involving a number of different specialists and not directly
related to the investigation of foundation conditions. However, Figure 5. WTG foundation concepts (a) gravity-based foundation (b)
monopile foundation (c) caisson foundation (d) multipile foundation (e)
there is benefit in maintaining liaison between the multi caisson foundation and (f) jacket foundation (Kallehave et al.,
environmental study and the foundation study so that data 2015)
collection programmes can be coordinated, relevant information
shared and a consistent approach made to the analysis and
presentation of data common to both. For example, an The requirement for increased global OWF capacity has led
environmental baseline or habitat investigation will normally to consideration of bigger WTGs and larger OWFs in the North
involve sampling and photographing the seafloor producing Sea and new rounds of development worldwide. New
information that will be useful in building the initial ground developments coincide with a focus on reducing the cost of
model. Data analysis for environmental purposes may consider wind energy, for which the foundation is a significant
sediment distribution (Figure 4) and seafloor processes, both of contributing factor. These requirements give rise to design
which are also relevant in an engineering context and will be challenges for optimisation of WTG foundation solutions, since
relevant to refining the ground model and site characterisation. the planning, design and execution of an OWF is governed by
the seabed conditions. Efficient geotechnical engineering design
may be restricted where challenging or uncertain soil conditions
are encountered, as they often are. Soil and rock conditions
display an extremely broad spectrum of classification and
mechanical response, and ‘local’ variability can be amplified
across large OWF sites. Seabed conditions could vary from
high strength volcanic and igneous rocks to normally
consolidated, very soft marine sediment.
Nevertheless, optimisation is required and the boundaries of
current geotechnical design practice must be tested in order to
deliver reducing costs for OWF development. Recent industry
driven research has focussed in this area (Byrne et al. 2015) and
new or relatively unproven (for WTGs) foundation concepts,
such as suction bucket jackets (SBJs), are being considered
along with WTG substructure concepts, such as floating turbine
concepts, which will permit OWF development in deeper water
sites. Such methods essentially necessitate the engineering
design strategy for “New Design Solutions” outlined by Evans
(2011). As the complexity of the geotechnical design challenges
increase, so too will the input data requirements associated with
ever more complex geotechnical analyses (Zdrakovic et al,
Figure 4. Example of a seafloor map produced for environmental study 2015). In order to optimise the marine site characterisation
purposes process, and ensure that the likelihood of unforeseen data input
requirements arising at a later stage of the project is minimised,
these input data requirements should be considered and
3 ENGINEERING DESIGN REQUIREMENTS communicated at an early project stage. However, different
foundation design concepts will require different input data and
Early design of foundations for offshore WTGs and OSSs drew geotechnical parameterisation requirements, due to the varying
upon the significant volume of experience existing within the nature of the geotechnical analyses required.
offshore oil and gas industry. It is therefore unsurprising that a Considering monopiles for example, advanced lateral design
large number of the early offshore WTGs and OSSs in the methods (e.g. Peralta, et. al. 2017) require information on soil
North Sea were founded on piled jacket substructures. With stiffness encompassing small to large strain, detailed
greater understanding of the critical WTG load cases and information on cyclic degradation and parameters to define the
development of integrated structural and geotechnical design consolidation rate of the soil (i.e. to define the drainage
23
conditions around the piles). This information is mostly

obtained from carefully conducted laboratory tests.
The pioneering SBJs in sandy soils (see Bye. et. al., 1995)
for oil and gas infrastructure were heavy and the foundations
were only occasionally subject to tensile loads. However, OWF
SBJ’s are much lighter and hence tensile loading is a more
frequent occurrence. With the wind load comprising a much
larger component of the total environmental load compared to a
traditional oil and gas platform, large static load offsets lead to
sustained tensile loads which must be resisted. To address this
design challenge advanced numerical modelling is essential
with detailed and precise geotechnical parameterisation,
particularly of the soil dilational response and the notoriously
variable soil permeability (e.g. see Whyte et. al., 2017). Such
properties can only be obtained in practice through
comprehensive laboratory testing programmes.
A number of recent OWF’s where weak rocks (calcarenites
and/or chalk) are prevalent have encountered different but no
lesser challenges. Once again specialised laboratory testing is
required for optimised design; for example constant normal
stiffness (CNS) testing for assessing axial response in soft rocks
provides critical input into state-of-the-art design methods (e.g.
Erbrich, et. al., 2010a and Augustesen et al., 2015). Lateral pile
design for such geotechnical conditions also requires
specialised tools and laboratory inputs (e.g. Erbrich, 2004 and
Muir Wood et al. 2015) and the envelope is being pushed in
terms of pile driving into such soil conditions, with
accompanying threats of premature refusal and pile buckling
(see Erbrich, et. al, 2010b). The advanced analyses which
should be conducted to appropriately assess these issues require
comprehensive and detailed field information (e.g. cone
penetration testing (CPT) for profiling, seismic CPT for in situ
small strain stiffness) and extensive laboratory testing for
measurement of representative strength and stiffness
parameters. Figure 6. The ultimate design objective is an optimised WTG location
To date, most OWFs have been located in areas with low and design, installed on time and operating reliably (conceptual jacket
seismic activity but as interest increases worldwide, plans for substructure shown)
OWFs in seismically active areas such as North America and
Asia are starting to appear. Seismic engineering will prove a
4 EVOLUTIONARY GROUND MODEL DEVELOPMENT
major new challenge for some of these cases and the foundation
response under seismic loading, and in particular the threat of
4.1 Overview
soil liquefaction, will need to be carefully evaluated from early
stages of development in order to demonstrate project OWF sites cover increasingly large areas of the seafloor as the
feasibility. Conventional simplified design approaches for number of turbines continues to grow and some developments
assessing seismic liquefaction are not likely to be sufficient for plan for over 100 turbines. One OWF site can include multiple
WTG foundations and advanced analyses are anticipated to be geological formations (both laterally and with depth), a range of
essential to capture response in soils that exhibit softening water depths, seafloor gradients, seafloor processes, other
under cyclic loading and/or liquefaction. Giannokou, et. al. constraints and geohazards, all of which can impact on
(2016) and Erbrich et. al. (2016) discuss similar design development planning. A combination of these factors at any
scenarios within oil and gas infrastructure. These will invariably given site requires a coordinated strategy of environmental
require detailed information on soil stress-strain response and baseline or habitat sampling, metocean acquisition, geophysical
cyclic degradation/liquefaction all of which are derived from, or survey, geological and geotechnical sampling, in-situ testing
calibrated to comprehensive laboratory testing. and laboratory testing. Integrating and interpreting these
The above discussion highlights the critical link between the datasets efficiently is essential to assess the engineering
advanced analyses required for optimised foundation design and significance of the seabed conditions. The development of a
the geotechnical input parameters required for these models, parameterised (engineering) ground model is the key to this
which are derived from the field and laboratory testing. The process (Figure 7)
best outcomes will invariably be obtained when the site A successful OWF ground model can identify, map and
characterisation process is advanced with full knowledge of the assess geotechnical constraints and geohazards and provide
design method input requirements. Hence wherever possible the input parameters to facilitate WTG and OSS foundation design,
required geotechnical investigation programmes should be development layout and cable routing. Campbell (1984) first
developed jointly between the engineering designers and the discussed the ground model method in relation to prediction of
characterisation team. seabed conditions for a large offshore development site. Fookes
(1997) describes numerous example onshore projects where
ground models were used as a powerful communication tool to
explain the diversity of the anticipated ground conditions.
Campbell et al (2008) describe the use of a ground model to
develop an optimised site investigation programme and budget.
Successful application of ground models was further
24
demonstrated by Power et al. (2011), Evans (2011) and Hill geotechnical data is collected and integrated, the logical
(2011) as a means to reduce uncertainty in ground condition conclusion is that the ground model will evolve into an
across large scale marine developments. Mason and Smith engineering ground model suitable to form the basis for
(2016) discuss integration of data for power cable routing. In foundation design, for example by establishing critical
the context of an OWF, the ground model is described by SUT subsurface features (Figure 8).
(2014) as an industry standard approach to collating site The process of carrying out a desk study for an OWF is
information as part of the geotechnical risk management described by the SUT (2014) which contains a detailed list of
process. items to be addressed including typical sources of information.
Thomas (2017) further defines an evolutionary ground A great deal of relevant information is typically available in the
model, which is the concept adopted here, in the sense that it is public domain including nautical charts, research papers and
necessary to understand the operation through time of the local experience and knowledge. In addition, there may be
formative or active processes at a site to understand and explain geophysical and geotechnical site investigations and installation
the physical attributes of the site. The terminology also invokes records from existing infrastructure, and although this is likely
a useful inference regarding the construction of the ground to be proprietary information, will be a valuable input to the
model itself, over the course of a site development. desk study if permission for its use can be secured or it can
The components of a preliminary evolutionary ground model otherwise be drawn upon as background experience.
are discussed in the following sections, which represent the site
characterisation process up to the point where a detailed
geotechnical site investigation is initiated. It is during the
course of this preliminary ground model development that the
potential requirements of the geotechnical design process
should be considered in as much detail as the available
information allows.
Figure 8. Example of subsurface features as identified during desk

study for an offshore site
4.3 Geophysical Investigation

In general geophysical investigations precede geotechnical
investigations with their results used both to update the ground
model, and aid selection of geotechnical investigation locations.
The geophysical survey thus provides the initial site specific
dataset from which the preliminary ground model formed
during the desk study is developed. As described by SUT
(2014), geophysical surveys are often carried out in two phases.
A reconnaissance survey over a coarse survey grid with a line
spacing of perhaps several hundred metres, using both seafloor
mapping and sub-seafloor mapping tools, will be designed to
gain a regional understanding of shallow geological conditions.
It should also identify any obvious geohazards and areas likely
to be problematic for installing turbines. Depending on the
licensing system, the reconnaissance survey may be carried out
before the final investment decision has been made.
There are a number of good references that describe the
equipment and techniques for marine geophysical investigations
and their application to studies such as offshore wind farms.
SUT (2014) contains information on this subject and more
Figure 7. The OWF Ground Model detail can be found in the Conduct of Offshore Drilling Hazard
Site Surveys - Technical Notes, IOGP (2015), which although
produced for the oil & gas industry contains much relevant
information.
4.2 Desk study Detailed geophysical surveys are conducted to investigate
The first stage in characterising the geological and more closely any potentially problematic conditions identified
environmental conditions in an area where an OWF is to be at an earlier stage, to refine the ground model and investigate
built is to carry out a desk study. The objective is to make planned turbine locations and cable route alignments. In order
maximum use of existing data and knowledge, by generating a to gain maximum value from the detailed geophysical survey it
preliminary evolutionary, engineering geological ground model. should be carefully planned, and timing within the overall site
This can be upgraded and refined as the project progresses, new characterisation process will be critical. Integration of all
data becomes available and further integration and analysis reconnaissance geophysical and geotechnical site investigation
takes place. Provided that suitable geophysical, geological and data should have been completed and the ground model
25
updated. The survey should be designed to define the 4.5 Geotechnical Design Evaluation
engineering geological units identified and to tie these into the
proposed turbine locations which should have already been In terms of the framework described by Thomas (2017), a
selected (Figure 9). Survey line spacing will typically be relatively detailed consideration of the geotechnical design
reduced compared to the reconnaissance survey and the choice requirements for an OWF development should ideally occur
of equipment directed by the desired resolution that may be during the “System Definition” project stage where the
higher than for the reconnaissance survey. geotechnical evaluation of the site is first considered (Figure
10). This coincides with the development of a preliminary
geotechnical model for the site based on information gathered
during the desk study and (early) reconnaissance geotechnical
investigation stages of the ground model development.
At this stage it is not necessary to be able to anticipate the
final foundation design solution; however, it is likely that the
foundation concept could be reasonably narrowed to
consideration of one or two primary concepts based on prior
review and foundation constraints (Figure 11). The potential
geotechnical design requirements associated with each concept
can then be developed, considering the critical geological
formations as relevant for each foundation type. Even a high
level review which identifies the potential need for advanced
numerical analyses and associated input data requirements is
beneficial at an early stage in order to refine the scope of future
geotechnical investigations, as necessary
The above process should then be subject to review and
iteration as further geotechnical data becomes available and
further geotechnical analyses are performed.
Figure 9. Example interpretation of geological soil units from a

geophysical section
Once the detailed geophysical survey is complete the results

are used to update the ground model. In some circumstances
there may be sufficient confidence in the ground model derived
from detailed geophysical and reconnaissance geotechnical
data, that foundations can be designed on the basis of the
parameters assigned to well defined stratigraphical units. In this
case the detailed geotechnical survey would be directed at
characterising these units and may be limited in its extent
perhaps to the point where a borehole or CPT is not required at
every turbine location. This is considered further by the SUT
(2014) and later in this paper.
4.4 Reconnaissance Geotechnical Investigation

Generally the geotechnical data acquisition for a new
development is best performed as a staged process with one or
more preliminary investigations undertaken to develop the
ground model as an iterative process, and to aid conceptual
design and turbine layout. More information on geotechnical
data acquisition techniques are given in Section 5.
Early reconnaissance geotechnical data feeds into the ground
model development, aids interpretation of the geophysical data Figure 10. Expansion of process flow presented by Thomas (2017) to
and allows foundation concept review and selection. It can also include consideration of geotechnical design requirements
be used as a trial of different site investigation techniques to
confirm that the best techniques are being used to obtain the
required geotechnical data given the site ground conditions. It is
also an opportunity to obtain preliminary data to characterise
the stiffness and cyclic response of key geotechnical units
within the ground model, as identified during the desk study
and following geotechnical design review.
Together with seafloor maps produced during the
geophysical investigation and the environmental study, the
geotechnical data will inform the ground model of areas of
potential seafloor mobility and scour potential, which are
common geohazards for OWF developments.
26
In areas of high current, anchored vessels are a good

solution, but generally dynamically positioned ships are used.
Use of floating vessels then requires incorporation of heave
motion compensators for vessel mounted equipment or the use
of seabed mounted equipment. SUT (2014) provide further
discussion on the advantages and disadvantages of various
vessel and drill platform types, with regard to geotechnical
investigation at OWF sites.
In general, seabed founded cone penetration testing (CPT)
systems and vessel mounted drilling systems are adopted.
Seabed founded drilling systems have not been an economic
option to date and are technically limited for these types of
investigations. However, new innovative systems (Figure 13)
mean that future investigations can now consider hybrid drilling
techniques, such as the SEADEVILTM system (Looijen &
Peuchen, 2017), which can offer improved drilling for difficult
ground conditions and deployment of a full suite of
investigation tools.
Figure 11. Example WTG foundation constraints related to identified Each stage of geotechnical site investigation will normally
subsurface features (Figure 8)
comprise predominantly seabed founded in situ testing to
refusal with a number of locations continued to greater depth
with down-the-hole in situ testing. This is usually
5 DETAILED GEOTECHNICAL INVESTIGATION complimented at a select number of locations with high quality
sampling to enable characterisation of the significant soil units,
5.1 Overview especially regarding strength, stiffness at various strain levels,
and how these vary under cyclic loading. Some in situ
As noted earlier, the geotechnical data acquisition for a new measurement of small strain stiffness, either P-S suspension
development is best performed as a staged process with one or logging or seismic cone testing should be considered. SUT
more preliminary investigations undertaken to feed into and (2014) provide recommendations for best practice geotechnical
develop the ground model as an iterative process and to aid work scope, based on foundation type. Kort et al. (2015) present
conceptual design and selection of foundation concept and a more detailed summary of geotechnical investigation planning
turbine layout. There is usually then a final phase of for GBS foundation design.
geotechnical data acquisition focussed on filling gaps in the
ground model and finalising key geotechnical design
parameters for each structure location.
The selection of site investigation plant is dictated by water
depth, environmental conditions, availability, and cost (day rate
and mobilisation), and the geotechnical data that is required.
Many wind farm developments, especially those in shallower
waters, are in areas of relatively high tidal currents and this can
affect the choice and workability of the site investigation plant.
If the water depth and seabed conditions permit, jack-ups are
generally the most cost effective solution, and permit standard
onshore equipment and site investigation approaches to be used.
If the water depth precludes the use of a cost effective jack up,
then barges or ships will be required (Figure 12). The cross over
is typically in the 15 m to 20 m water depth range.
Figure 13. The SEADEVILTM hybrid vessel-seafloor drilling (Looijen

& Peuchen, 2017)
Figure 12. Geotechnical drilling vessels working at an OWF site in the
North Sea
5.2 Coverage of Investigation
It has previously been mentioned that a typical OWF site will
generally cover a large area of the seafloor, encountering a
27
range of geological formations (both laterally and with depth). in Figure 12). This yields valuable data regarding the
Clearly then, acquiring detailed geotechnical data at all degradation or ‘friction-fatigue’ of soil resistance with relative
structure locations within the OWF area may not appear soil:steel movement, which is a key input into assessing skirt
warranted, especially where a well-developed ground model is penetration resistance for suction bucket foundations and soil
in place; however the requirement for location specific resistance to driving (SRD) for driven piles.
investigation should be considered in the context of the Performing seismic cone penetration tests (SCPT) at a
geotechnical risk assessment for the site. This consideration number of locations, typically 10 to 20%, is a small increment
should be informed by the outcomes of the desk study and the in time and cost over standard CPTs and gives valuable
information contained within the preliminary ground model, but information on in situ small strain stiffness. SCPT is not always
also taking into account the geotechnical design requirements possible due to limited cone penetration, in which case seismic
and track record of the proposed foundation concept(s). High wave velocity (P-S) suspension logging can be used in the
levels of lateral and vertical variability may necessitate unsupported open borehole, so long as borehole stability is
structure location specific investigation, or indeed several sufficient. Often P-S suspension logging is performed at the end
investigation points across a single foundation footprint, in of a sampling or CPT borehole and the borehole quality is not
order to reduce the foundation design and installation risks as very good. Consideration should be given to undertaking the
low as reasonably practicable for the project. logging in a bespoke borehole, or to undertake it at intermediate
The subject of data coverage for different foundation stages during the drilling of a borehole for other purposes so
concepts is discussed by SUT (2014) and is not restated here, that the logging is undertaken in a fresher, better quality hole. In
except to note that ultimately the extent of the geotechnical practice a combination of SCPT and P-S logging is undertaken
investigation may be governed by the national design code to cover the full depth of the foundation. High pressure
requirements applicable for the OWF site under consideration. dilatometer or pressuremeter testing can also be undertaken to
gain insight in soil stiffness and strength at a rage of strain
5.3 In situ testing levels. However this testing is rather slow, and hence
expensive, and only produces discrete data points. The test is
In situ testing is predominantly static CPT and this is a good also often difficult to interpret and therefore may provide
way of getting high quality geotechnical data at many locations limited resolution in strong to hard soils. Only a few number of
in a cost effective manner. If conditions are suitable, seabed tests therefore are generally undertaken and the data
testing units, which are available with up to 20t thrust, can test extrapolated by correlation with other data, e.g. CPT.
up to 40m below seafloor, or beyond. Even if hard ground Future advances are expected to include wider use of
conditions restrict the achievable penetration to an average of, recently developed fibre optic cone technology, which offers
say, 14m, it is usually cost effective to undertake seabed significantly improved accuracy of cone resistance and pore
founded testing and then to progress with down-the-hole CPTs pressure measurements compared to existing strain gauge
in drilled boreholes to cover the full depth of interest for the technology (Fugro, 2017). Other in situ characterisation tools,
foundations. This is normally done at least c. 30% of locations such as the SEADARTTM free fall penetrometer, may provide
providing near continuous data for soil profiling (Figure 14). cost-effective shallow geotechnical data in very soft clays for
It is also possible to extend the measurement potential of the the purpose of cable route investigation, where seafloor
standard CPT cone to include additional sleeves. For example, detection may be of concern (Peuchen et al., 2017). Further
the dual sleeve cone (Figure 15) measures not only initial developments are expected to include cyclic CPT measurements
soil:sleeve frictional resistance, but also the resistance measured under increasing accuracy of control.
after a degree of further relative soil:steel movement
(approximately 0.5 m for the example dual sleeve cone shown
Figure 14. Example measured and interpreted CPT data
28
Figure 15. The dual sleeve cone and example cone resistance and sleeve friction data. Operated by Fugro
high quality coring can be challenging as there is a large

5.4 Sampling unsupported length of coring string compared to onshore coring
operations.
Sampling boreholes will typically be undertaken at 10% of the As with the selection of in situ methods and for the reasons
WTG locations and at each OSS location to obtain samples highlighted earlier in this paper, soil sampling for OWF
(Figure 16) in the significant soil layers to feed into the design. characterisation is subject to the recommendation proposed by
The data obtained from the soil samples is then extrapolated Clayton et al (1995): “…there is a need to match the
across the site by correlation with the in situ test data via the sophistication of sampling to the sophistication of the analysis
calibrated ground model. Sampling boreholes should be taken and design of the project… the constitutive modelling of soil
in the direct vicinity of CPT boreholes to allow best correlation behaviour via finite element or finite difference analyses, will
between CPT and laboratory datasets. require high-quality sampling and testing methods.”
The SUT (2014) provide a summary of seabed sampling
equipment suitability by soil type. For the majority of the 5.5 Laboratory testing
equipment types considered by the SUT, two types of samples
are obtained during marine geotechnical investigation: SUT (2014) recommend the following geotechnical data
‘undisturbed’ and disturbed samples. The sample type describes requirements for foundation and cable installation design:
whether or not the soil is recovered in an intact state, with
regard to the in situ condition. Obtaining undisturbed sand • Description and index classification
samples is not practicable offshore and therefore laboratory • Strength parameters (for different failure modes,
testing on cohesionless soils is by necessity on reconstituted monotonic and cyclic)
samples, as discussed later. Clayton et al. (1995) provide a • Soil modulus and damping parameters
comprehensive discussion on sampling and sample disturbance • Permeability and consolidation parameters
effects for onshore site investigations, which is equally valid for • Liquefaction potential
use of sampler and coring methods offshore. Similar discussion • Thermal conductivity
is presented for very soft to soft clays by Ladd and DeGroot • Chemical composition
(2003). Sample disturbance has a significant impact on
laboratory element testing, as considered later in this paper. In situ measurements alone are not able to meet the data
The soil samples from geotechnical investigation ground requirements for geotechnical analyses since these do not
truth the in situ testing data and also enable the measurement of provide information across the full stress-strain range to be
the soil properties needed for efficient turbine foundation investigated for design and cannot cover a full range of
design, such as the variation of soil stiffness and strength at foundation loading conditions. Laboratory testing is therefore
different strain levels and during cyclic loading. If rock is required to provide information on soil and rock classification
encountered in the foundation zone, then triple barrel rotary and mechanical response under monotonic and cyclic loading
coring can be undertaken. This can also be used to obtain high for foundation design, and chemical and thermal properties for
quality samples in more competent soils. To obtain good quality cable assessments.
core recovery requires very good heave compensation when
working from a floating vessel. Even from a jack-up platform,
29
SUT (2014) provide guidance on the applicability of various

conventional laboratory testing methods for measurement of
different soil parameters in several soil types, including sand,
clay, carbonate/calcareous soils and weak rock. These
guidelines provide a useful point of reference for planning of
laboratory testing programmes, but they should not be
considered as (and are not intended as) blanket
recommendations across all foundation and geotechnical
analysis types. For example, any geotechnical design process
including consideration of advanced numerical analysis is likely
to generate input data requirements outside of the scope of
guidance provided by SUT.
It is common to make a distinction between ‘routine’ and
‘advanced’ testing, where the latter are subject to a higher
degree of interpretation and require an increasingly careful and
detailed specification. Rather than make an arbitrary distinction
based on apparatus type, ‘routine’ laboratory testing should be
considered as all tests scheduled for general soil unit
identification, ground model calibration, basic engineering
classification and parameter profiling, regardless of test type.
‘Advanced’ testing is then defined as those tests from which
detailed and specific soil behavioural models will be developed
for application in advanced geotechnical analyses (e.g. FEA or
similar, site response analysis (SRA) or other state-of-the-art
design methods for monotonic or cyclic foundation response
analysis).
5.5.1 Laboratory Specimens

Scheduling, specification and interpretation of laboratory
testing should always consider the methods employed to obtain
and prepare the geotechnical samples being tested. As discussed
previously, sample disturbance has a significant impact on the
laboratory measured response of both soft clays (Lunne et al.,
2006) and overconsolidated clays (Berre, 2014). Disturbance
can also be introduced during preparation of intact soil
specimens. X-ray examination of sample cores can be
considered ahead of testing, to ensure samples are of suitable
quality and free of inclusions which may otherwise compromise
the quality of the test data.
The soil fabric created during soil deposition and subsequent
processes in situ contributes to mechanical response to the
extent that the effect of structure can be of equal importance to
those of soil state (void ratio) or effective stress (Mitchell and
Soga, 2005). Given the range of depositional and post-
depositional processes acting on soils offshore, laboratory
reconstitution is unlikely to result in test specimens which are
fully representative of the in situ soil state. The laboratory
measured response of reconstituted soil specimens will then
depend on the preparation method employed (Ishihara, 1993).
The effect on drained strength is less significant due to rapid
changes in fabric during shearing and tends to diminish with
increasing relative density. However, the effect on undrained
response of sand is particularly pronounced and has been
investigated extensively in relation to the liquefaction potential
of cohesionless soils (Vaid et al. 1999; Sze and Yang, 2014).
The procedures necessary to minimise the impact of
disturbance and specimen preparation should therefore be
carefully considered during specification of laboratory testing.
In addition the designer’s interpretation of the laboratory data
should also be carefully considered. For example, the
specification of direct simple shear (DSS) testing as input to
pile design should consider the boundary conditions of the
problem as replicated during testing. Application of normal
stresses representing in situ effective vertical stress would be
relevant for lateral pile design (Erbrich et al., 2010). For axial
pile design it may be necessary to consider a normal stress
Figure 16. Conceptual illustration of the (API drilling mode) downhole which is representative of the effective horizontal stress
sampling process
30
condition, imposed on a test specimen cut at 90 to vertical from consolidated undrained (CU) triaxial compression and
a sample core (Figure 17, Rattley et al., 2017; Sim et al., 2013). extension tests and DSS tests, can be performed onshore on
undisturbed and reconstituted soil specimens. For routine
(profiling) effective stress triaxial testing (Figure 18), the
specimen is tested under conditions that replicate as closely as
possible the (pre-sampling) in situ stress conditions at the
sample depth. Performing a range of strength testing enables
undrained strength to be measured under different failure
modes, providing an indication of strength anisotropy of the
soil. From these measurements the variance in the cone factor
(Nkt) can be estimated for each mode of failure, allowing near-
continuous undrained shear strength profiling from CPT cone
resistance data. Note that it is a common oversight not to
perform undrained strength testing on cohesionless soil
samples. While this data may not be required for jacket pile
design, a drained soil response assumption is unlikely to be
valid for the majority of large diameter monopile and suction
bucket design cases. For these foundations response may be at
least partially drained under environmental loading, even for
relatively uniform sand profiles (Peralta et al., 2017; Whyte et
al., 2017) and the undrained strength of sands soils is of
practical interest for design. For the same reason, laboratory
measurements of soil permeability may also be required.
As a part of the ‘routine’ triaxial test, bender elements can
be included in the platens of the apparatus to measure
(vertically propagating) shear wave velocity for determination
of shear modulus at very small strain (Gmax). This parameter is a
key input for geotechnical design of WTG foundations and
frequent laboratory measurement allows reliable correlation
with in situ data for parameter profiling. The measurements can
Figure 17. Schematic of stress conditions in a soil element adjacent to a also be used to assess the reliability of laboratory reconstituted
pile shaft as translated in the DSS test for axial pile design (see also Sim
et al., 2013)
specimens by comparison to in situ data.
5.5.2 Routine Laboratory Testing

Laboratory test schedules for routine testing should be assigned
by an experienced engineer based on the soil conditions
sampled during geotechnical investigation and considering the
evolutionary ground model developed for the site. The latter is
essential for a coherent site characterisation process. A
laboratory testing schedule should contain sufficient testing to
allow classification of the soils general engineering
characteristics and to enable soil unit identification for
calibration of the ground model. Laboratory testing should
therefore be scheduled on a soil unit basis. Where the same soil
unit is encountered at a range of depths at a number of borehole
locations, it is recommended to schedule tests at different
depths to provide a composite, near-continuous data profile
across the soil unit. More frequent testing, including thermal
conductivity tests, may be required on soil near the seafloor for
Figure 18. ‘Routine’ triaxial testing systems. Operated by Fugro
cable route assessments. In addition to index testing, tests
should be scheduled for derivation of general design parameter
profiling with depth, at each sampled location, for input to
Consolidation tests (i.e. constant rate of strain (CRS) or
foundation design.
incremental oedometer tests) are undertaken to assess the one-
For cohesionless material, minimum and maximum density
dimensional compression response of the soil. The tests are
(min/max) tests are typically undertaken to define the
particularly important for soils with a variable geological stress
approximate limits of the void ratio which may be present in-
history. In the North Sea, for example, many soil formations
situ. Results from these tests are used to calculate target
have experienced significant variation in past stress levels due
densities for laboratory reconstituted soil specimens, based on
to geological processes including: desiccation and glacial
relative density values determined from the CPT data. While
advancement and retreat, and other process such as creep
this method is generally considered standard practice for
(ageing) and physicochemical actions. Geological history
offshore site characterisation, it should be noted that the
therefore has an important influence on the mechanical
approach is subject to some known limitations surrounding
response of marine soils. Evans (2011) describes how the
determination of maximum in situ density (Blaker et al ,2015).
framework proposed by Burland (1990), Chandler (2000) and
Consolidated drained (CD) triaxial compression tests are
Cotecchia and Chandler (2000) can provide an essential
typically performed for determination of drained shear strength
interpretation basis for understanding the influence of soil
(friction angle).
structure on the response of clay soils, allowing for
Undrained shear strength is measured offshore using index
sedimentation structure and post-sedimentation effects. Both
test methods. More accurate test methods, where effective
intact and reconstituted tests are therefore recommended.
stresses can be controlled (to varying degrees), such as
31
Understanding stress history in clay soil units can also provide

valuable insight into the past loading of underlying sand units,
since the variation of stress conditions in situ (i.e. evolution of
in situ K0) cannot be assessed by direct measurements on
disturbed sand samples.
5.5.3 Advanced Laboratory Testing

Jardine (2014) provides an overview of advanced laboratory
testing which focusses on the application of highly
instrumented testing to increase understanding of foundation
response. Jardine illustrates the practical value of such testing,
but also demonstrates that the aims of the foundation analysis
must be well understood in order that the laboratory testing is
carefully specified and controlled to maximise the applicability Figure 20. Clay specimen during cyclic shearing in the CSS apparatus
of the resulting data. Jardine ends his discussion with an and resulting stress-strain data. Operated by Fugro
observation which is now fundamental to the WTG foundation
design process: “Advanced laboratory testing is vital to
advancing all difficult geotechnical engineering problems The soil response measured in any stress or strain controlled
where the outcomes depend critically on the detailed laboratory test is a direct function of the boundary conditions
constitutive behaviour of the ground.” applied by the test apparatus and the processes that preceded the
The complexity of geotechnical analysis being routinely shearing stage of the test and the application of such processes
applied for WTG foundation design is increasing. To achieve should be carefully considered. For example, the following are
foundation optimisation under complex loading it is necessary just a small subset of considerations which may be pertinent to
to perform advanced analysis (FEA or similar) and such specification of advanced (or routine) laboratory testing:
analysis requires careful calibration of a suitable constitutive
model to predict soil response. Advanced laboratory test • Specimen reconstitution method
programmes are therefore required to evaluate specific aspects • Saturation of heavily overconsolidated (swelling)
of soil response as considered critical for detailed geotechnical clays
analysis. These generally include, but are not limited to: • Preconsolidation regime
• Consolidation stress paths
• Cyclic behaviour • Consolidation creep intervals
• Nonlinear stiffness (stress/strain dependent) • Pre-shearing stages
• Rate dependency and creep • Cyclic loading ratios, etc.
• Anisotropy
• Critical states The above discussion reinforces the requirement for the
• Yield surface and plastic potential advanced laboratory testing programme for an OWF
development to be considered with full understanding of the
The laboratory testing required to investigate the above laboratory testing methods and knowledge of the design method
includes dynamic testing such as the cyclic triaxial (CTXL) input requirements. It is therefore recommended that a clear
(Figure 19), cyclic direct simple shear (CSS) (Figure 20) and distinction is made between the routine testing and advanced
resonant column (RC) tests. However it can also include testing programmes (although scheduling is interdependent),
carefully planned series of tests performed on routine testing with the former being developed jointly between the
apparatus such as the triaxial and oedometer apparatus, with engineering designers and the site characterisation team. It is of
variations in test specification lying outside of those that would limited value to schedule a generic suite of ‘advanced’
be considered for routine (profiling) testing. Triaxial testing laboratory testing simply to satisfy a general preconception of
may also include non-standard measurements, such as high applicability to design. To ensure project value advanced
resolution local strain measurements, vertical and horizontal BE laboratory testing must be coordinated with a clear design
determinations and mid-height pore water pressure strategy and ideally based on knowledge gained from
measurements to increase the scope of the test data for preliminary geotechnical analyses using soils data generated
constitutive model calibration. from the preliminary ground model. This approach will ensure
adequate opportunity for a thorough gap analysis approach to
test scheduling.
6 GEOTECHNICAL INTERPRETATION
Geotechnical interpretation and parameterisation forms the link

between the ground model and the engineering design process
(Figure 5) and is interrelated to each. As part of the
geotechnical interpretation it is necessary to define soil
unitisation and design parameter profiling with depth for each
WTG and OSS structure location. Soil behavioural models,
considering aspects of response as discussed earlier, are
developed. If required, constitutive model calibration will also
be performed as input to engineering analyses (Figure 21).
Figure 19. Cyclic triaxial testing systems. Operated by Fugro
32
Figure 21. Schematic example of the geotechnical interpretation and analysis process for WTG monopile analysis (up to FEA)
If the integrated site characterisation process has been well should be developed jointly between the engineering designers
coordinated, taking input from all parties, then the geotechnical and the characterisation team.
interpretation will proceed successfully without highlighting
data gaps which lead to: excessive conservatism in parameter
selection, extended parametric analyses, or a requirement for 8 REFERENCES
further geotechnical investigation. Ultimately it is likely that
Augustesen A., Leth C., Ostergaard M., Moller M., Duhrkop J. and
some extrapolation of data is required as part of the Barbosa P. 2015. Design methodology for cyclically and axially
geotechnical interpretation; this should be performed with care loaded piles in chalk for Wikinger OWF. Proc. Int. Symp. on
and taking due consideration of all available published and Frontiers in Offshore Geotechnics, Oslo, Norway.
proprietary experience available. Common with development of Berre T. 2014. Effect of sample disturbance on triaxial and oedometer
geotechnical investigation programmes, geotechnical behaviour of a stiff and heavily overconsolidated clay. Canadian
interpretation should be progressed jointly between the Geotechnical Journal, 51(8), 896-910.
engineering designers and the characterisation team. Blaker O., Lunne T., Vestgarden T., Krogh L., Thomsen N., Powell J.
and Wallace C. 2015. Method dependency of determining maximum
and minimum dry unit weights of sands. Proc. Int. Symp. on
Frontiers in Offshore Geotechnics, Oslo, Norway.
7 CONCLUDING REMARKS Burland J.B. 1990. On the compressibility and shear strength of natural
clays. Geotechnique, 40(3), 327-378.
The large spatial extents of offshore wind farms mean that an Bye A., Erbrich C., Rognlien B. and Tjelta T. 1995. Geotechnical
effective site characterisation is essential to achieving optimised Design of Bucket Foundations. Proc. Offshore Technology
geotechnical design. A compartmentalised approach of separate Conference Houston, USA. OTC 7793.
Byrne B.W., Mcadam R., Burd H. and Skov Gretlund J. 2015. New
geophysical and geotechnical investigations followed by design methods for large diameter piles under lateral loading for
independent engineering design can lead to undesirable project off-shore wind applications. Proc. Int. Symp. on Frontiers in
surprises which impact on overall project schedule and cost. Offshore Geotechnics, Oslo, Norway.
The solution is a phased approach to integrated marine site Campbell K. 1984. Predicting offshore soil conditions. Proc. Offshore
characterisation where all technical aspects are considered. The Technology Conference Houston, Texas. OTC 4692.
benefits of this approach become keenest when the Chandler R.J. 2000. Clay sediments in depositional basins: the
requirements of the geotechnical design process are considered geotechnical cycle. Q. J. Eng. Geology, 33, 7-39.
at an early stage of the characterisation process. The best Clayton C., Matthews M. and Simons N. 1995. Site Investigation.
outcomes will invariably be obtained when the site Blackwell Science, Oxford.
Cotecchia F. and Chandler R.J. 2000. A general framework for the
characterisation process is advanced with full knowledge of the mechanical behaviour of clays. Geotechnique, 50(4), 431-447
design method input requirements. Erbrich C. 2004. A New Method for the Design of Laterally Loaded
To ensure project value, detailed geotechnical investigation Anchor Piles in Soft Rock. Proc. Offshore Technology Conference,
must be coordinated with a clear design strategy, ideally based Houston, USA. OTC 16441.
on knowledge gained from preliminary geotechnical analyses Erbrich C., O’Neill M., Clancy P. and Randolph M. 2010a. Axial and
using soils data generated from the preliminary ground model. Lateral Pile Design in Carbonate Soils. Proc. Proc. Int. Symp. on
This approach will ensure adequate opportunity for a thorough Frontiers in Offshore Geotechnics, Perth, Australia.
gap analysis approach to investigation planning. Wherever Erbrich C., Barbosa-Cruz E., and Barbour R. 2010b. Soil-Pile
Interaction During Extrusion of an Initially Deformed Pile. Proc.
possible the required geotechnical investigation programmes
Int. Symp. on Frontiers in Offshore Geotechnics, Perth, Australia.
33
Erbrich C., Wallbridge P. and Yamamoto N. 2016. Numerical Vaid Y.P., Sivathayalan S. and Stedman D. 1999. Influence of
Modelling of Seismically Induced Settlement for Ichthys Riser specimen reconstituting method on the undrained response of sand.
Support Structure. Proc. Offshore Technology Conference OTC Geotechnical Testing Journal, 22(3), 187–195.
Asia 2016. OTC-26778-MS. Vattenfall. 2016. Mines found at Horns Rev 3 site [online]
Evans T. 2011. A systematic approach to offshore engineering for http://news.vattenfall.com/en/article/mines-found-horns-rev-3-site
multiple-project developments in geohazardous areas. Proc. Int. Whyte S., Rattley M., Erbrich C.E., Burd H.J. and Martin C.M. 2017. A
Symp. on Frontiers in Offshore Geotechnics, Perth, Australia. practical constitutive model for soil structure interaction problems
Fookes P. 1997. Geology for engineers: the geological model, involving dense sands. Proc. Int. Conf. Offshore Site Investigation
prediction and performance. The First Glossop Lecture. Q. J. and Geotechnics, London, UK.
Engineering Geology, 30, 293-424. Zdravkovic L., Taborda D.M.G., Potts D.M., Jardine R.J. and Sideri M.
Giannakou A., Makra A., Chacko J. and Poudens O. 2016. Evaluation 2015. Numerical modelling of large diameter piles under lateral
of Kinematically Induced Demands on Offshore Platform loading for offshore wind applications. Proc. Int. Symp. on
Foundations due to Liquefaction. Proc. Int. Conf. on Natural Frontiers in Offshore Geotechnics, Oslo, Norway.
Hazards & Infrastructure, Chania, Greece.
Ishihara K. 1993 Liquefaction and flow failure during earthquakes.
Geotechnique, 43(3), 351-451.
Jardine R.J.J. 2014. Advanced laboratory testing in research and
practice. Geotechnical Research, 1(1), 2-31.
Kallehave D, Byrne B., Thilsted C., Mikkelsen K.K. 2015.
Optimization of monopiles for offshore wind turbines. Phil.
Transactions, Royal Society Publishing.
Kort D., Pederstad H. and Nowaki F. 2015. Planning of soil
investigation for GBS foundation design. Proc. Int. Symp. on
Frontiers in Offshore Geotechnics, Oslo, Norway.
Ladd C.C. and DeGroot D.J. 2003. Recommended practice for soft
ground site characterization. Proc. PanAmerican Conf. Soil and
Rock America, USA
Lunne T., T Berre and S Strandvik. 1997. Sample disturbance effects in
soft low plasticity Norwegian clay. Proc. Conf. on Recent
Developments in Soil and Pavement Mechanics, Rio de Janeiro,
Brazil.
Lunne T., Berre T., Andersen K.H., Strandvik S. and Sjursen M. 2006.
Effects of sample disturbance and consolidation procedures on
measured shear strength of soft marine Norwegian clays. Canadian
Geotechnical Journal, 43(7), 726-750.
Mason A. and Smith C. 2016. Integration of geophysical and
geotechnical data for the routing of power cables. Conf. Applied
Shallow Marine Geophysics. Near Survey Geosciences, Barcelona,
Spain.
Mitchell J.K. and Soga K. 2005. Fundamentals of soil behaviour. Wiley
& Sons, USA.
Muir Wood A., Mackenzie B., Burbury, D., Rattley M., Clayton C.,
Mygind M. and Liingaard M.A. 2015. Design of large diameter
monopiles in chalk at Westermost Rough offshore wind farm. Proc.
Int. Symp. on Frontiers in Offshore Geotechnics, Oslo, Norway.
Perelta P., Ballard J.C., Rattley M. and Erbrich C. 2017. Dynamic and
Cyclic Pile Soil Response Curves for Monopile Design. Proc. Int.
Conf. Offshore Site Investigation and Geotechnics, London, UK.
Looijen P. and Peuchen J. 2017. Seabed investigation by a novel hybrid
of vessel-based and seafloor-based drilling techniques. Proc. Int.
Conf. Offshore Site Investigation and Geotechnics, London, UK.
Peuchen J., Looijen P. and Stark N. 2017. Offshore characterisation of
extremely soft sediments by free fall penetrometer. Proc. Int. Conf.
Offshore Site Investigation and Geotechnics, London, UK.
Power P., Clare M., Rushton D. and Rattley M. 2011. Reducing Geo-
risk for Offshore Development. Int. Symp. on Geotechnical Risk
and Safety.
Rattley M., Costa L., Jardine R. and Cleverly W. 2017. Laboratory test
predictions of the cyclic axial resistance of a pile driven in North
Sea soils. Proc. Int. Conf. Offshore Site Investigation and
Geotechnics, London, UK.
Sim W., Aghakouchak A. and Jardine R.J. 2013. Cyclic Triaxial Tests
to Aid Offshore Pile Analysis and Design. Geotechnical
Engineering, 166 (2), 111-121.
Society for Underwater Technology (SUT). 2014. Offshore Site
Investigation Committee, 2014, Guidance Notes for the Planning
and Execution of Geophysical and Geotechnical Ground
Investigations for Offshore Renewable Energy Developments. SUT,
London.
Sze H. and Yang J. 2014. Failure Modes of Sand in Undrained Cyclic
Loading: Impact of Sample Preparation. Journal of Geotechnical
and Geoenvironmental Engineering, 140(1), 152-169.
Thomas S. 2017. A Phased and Integrated Data Interpretation
Approach to Site Characterisation. Proc. Int. Conf. Offshore Site
Investigation and Geotechnics, London, UK.
34
Design aspects for monopile foundations
Aspects du dimensionnement pour les fondations monopieux
Harvey J. Burd, Byron W. Byrne, Ross A. McAdam, Guy T. Houlsby, Chris M. Martin, William J.A.P.
Beuckelaers
Department of Engineering Science, Oxford University, UK, email: [email protected]
Lidija Zdravković, David M.G. Taborda, David M. Potts, Richard J. Jardine

Department of Civil and Environmental Engineering, Imperial College London, UK
Ken Gavin, Paul Doherty, David Igoe

Formerly University College Dublin, Ireland
Jesper Skov Gretlund, Miguel Pacheco Andrade

DONG Energy Wind Power
Alastair Muir Wood

Formerly DONG Energy Wind Power
ABSTRACT: This paper describes the outcome of a recently completed research project – known as PISA – on the development of a
new process for the design of monopile foundations for offshore wind turbine support structures. The PISA research was concerned
with the use of field testing and three-dimensional (3D) finite element analysis to develop and calibrate a new one-dimensional (1D)
design model. The resulting 1D design model is based on the same basic assumptions and principles that underlie the current p-y
method, but the method is extended to include additional components of soil reaction acting on the pile, and enhanced to provide an
improved representation of the soil-pile interaction behaviour. Mathematical functions – termed ‘soil reaction curves’ – are employed
to represent the individual soil reaction components in the 1D design model. Values of the parameters needed to specify the soil
reaction curves for a particular design scenario are determined using a set of 3D finite element calibration analyses. The PISA
research was focused on two particular soil types (overconsolidated clay till and dense sand) that commonly occur in north European
coastal waters. The current paper provides an overview of the field testing and 3D modelling aspects of the project, and then focuses
on the development, calibration and application of the PISA design approach for monopiles in dense sand.
RESUME : Ce papier décrit les résultats d’un projet de recherche terminé récemment – connu sous le nom de PISA – portant sur le
développement d’une nouvelle procédure pour le dimensionnement des fondations monopieux pour les structures support des
éoliennes offshore. La recherche PISA s’est intéressée à l’utilisation d’essais in-situ sur pieux et d’analyse (3D) par éléments finis
afin de développer et de calibrer un nouveau modèle de dimensionnement unidimensionnel (1D). Le modèle de dimensionnement 1D
obtenu est basé sur les mêmes hypothèses et principes fondamentaux à la base de la méthode p-y actuelle. Cependant, la méthode est
étendue afin d’inclure des composants supplémentaires à la réaction du sol sur le pieu, et améliorée afin de fournir une meilleure
représentation du comportement de l’interaction sol-pieu. Des fonctions mathématiques – appelées ‘soil reaction curves’ (courbes de
réaction du sol) – sont utilisées afin de représenter les composants individuels de la réaction du sol dans le modèle de
dimensionnement 1D. Les valeurs des paramètres requis pour spécifier les courbes de réaction du sol pour un scenario de
dimensionnement donné, sont déterminés au moyen d’un ensemble d’analyses de calibration 3D par éléments finis. La recherche
PISA s’est concentrée sur deux types de sols spécifiques (argile glaciaire surconsolidée et sable dense) communément présents dans
les eaux côtières du nord de l’Europe. Le présent papier offre une vue d’ensemble des caractéristiques des essais in-situ sur pieux et
des modélisations 3D du projet, et se concentre ensuite sur le développement, la calibration et l’application de l’approche de
dimensionnement PISA pour les monopieux dans du sable dense.
KEYWORDS: monopile design, 1D model, soil reaction curves
1 INTRODUCTION models of the performance of individual wind turbine structures

are employed within a computational framework to optimise an
The monopile is the dominant foundation system for current entire windfarm. For a wind turbine foundation model to be
and planned offshore wind farm developments in shallow useful in this context, it must be fast to compute and sufficiently
coastal waters, particularly in Europe. Monopile foundations in accurate and reliable for design purposes. Analyses based on
this application are typically designed with the aid of simplified the p-y method can be computed rapidly, but current forms of
analysis approaches, such as the ‘p-y’ method and its variants, the p-y method – which have their origins in the design of long,
in which the foundation is modelled as an embedded beam, with relatively flexible piles – are widely regarded as being
the lateral load-displacement interaction between the soil and unreliable for the design of monopiles with relatively small
pile represented by non-linear functions known as p-y curves. values of L/D (where L is embedded length and D is pile
Simplified computational models of this sort facilitate the diameter), see e.g. Doherty and Gavin 2011.
development of multiscale optimisation procedures in which
35
A recent project – known as PISA (PIle Soil Analysis) – sand. At an early stage in the project, potential sites for the field
employed field testing and computational modelling to develop tests were sought, with approximately homogeneous profiles
a new design approach for monopile foundations for offshore consisting of each of these two soil types. The outcome of this
wind turbine applications. In this new approach, the underlying process was that a site at Cowden in the north east of England
simplicity of the p-y method – in which the pile is modelled as was selected as the clay site and a site at Dunkirk in northern
an embedded beam – is retained, but additional components of France was selected as the sand site.
soil reaction are incorporated within the design model to The current paper provides an overview of the design
improve its performance. approach that was developed during the PISA project, together
The PISA project was conducted, between 2013 and 2016, with a description of the process that was employed to calibrate
as a joint industry/university study. The scientific programme the 1D design model for a representative offshore soil profile
was developed by an Academic Working Group (AWG) with consisting of dense sand. The development of representative
members drawn from Oxford University, Imperial College and soil conditions and the model calibration process were based on
University College Dublin. The research was supported by the the results of the Dunkirk field tests and associated 3D finite
project partners listed in the Acknowledgements section of this element modelling.
paper; Dong Energy acted as the lead partner and main A similar process to calibrate the design model for a
contractor. PISA consisted of three related research strands: (i) representative offshore clay till site (not discussed in the current
reduced-scale field testing of monopile foundations (with paper) is described in Byrne et al. 2017.
associated site investigation and laboratory soil testing) (ii)
three-dimensional (3D) finite element modelling and (iii) the
development of a new one-dimensional (1D) modelling 2 THE PISA DESIGN APPROACH
approach for design.
The PISA research was focused on monotonic lateral 2.1 Formulation of the PISA design model
loading, as this was identified to be the area where substantial
gains could be achieved. Although it is acknowledged that The components of the PISA design model are illustrated in
improved design methods for cyclic loading are needed, it is Figure 1. A monopile foundation is represented in the model as
first necessary to have robust procedures for monotonic loading an embedded beam with moment and horizontal force
before addressing more complex cyclic loading issues. The applied to the pile at the ground surface. Four separate
methods developed in PISA are capable of future extensions to components of soil reaction are assumed to act on the
cyclic loading, and additional cyclic testing was conducted embedded monopile. Consistent with the standard p-y method, a
during the field testing phase to aid such extensions. distributed lateral load, p (units of force / length) acts on the
To limit the scope of the research the project was concerned pile. Additionally, a distributed moment, m (units of
specifically with two sets of soil conditions that are commonly force length / length) is applied; this distributed moment is
encountered in north European coastal waters: (i) a stiff caused by the vertical tractions that are induced at the soil-pile
overconsolidated clay till and (ii) a dense to very dense marine interface when local pile rotations occur, as indicated in Figure
Tower
(a)
(a) (b)(b)
MG
MG
HG HG
Ground level
z v
Distributed
lateral load Distributed
lateral load
Monopile
p(z,v)
Distributed
moment
m(z,)
Vertical shear
tractions at pile- Base
soil interface horizontal Base
force HB(vB) moment
MB(B)
Horizontal force and

moment applied at the pile
base.
Figure 1. PISA 1D monopile model (a) assumed soil reactions acting on the monopile (b) 1D design model. In the left figure, the soil reactions are
shown in the directions that they are likely to act, given the applied loads that are indicated. In the right figure, the indicated directions of the soil
reactions are consistent with the cordinate directions shown.
36
1a and Figure 2. Gavin 2011, Lam 2013).

A horizontal force and a moment acting on the The design model in Figure 1 has been implemented as a 1D
base of the pile are also included in the design model. A four- finite element model adopting a Galerkin formulation. The
component model of this sort has previously been employed for implementation is based on standard procedures for non-linear
the design of drilled shafts for onshore applications (e.g. Lam finite element analysis. The functions representing the soil
2013) and has previously been described in the context of the reaction curves are embedded in the 1D model (in much the
PISA research by Byrne et al. 2015a. same was as a constitutive model is embedded in a standard 3D
In the current implementation of the model, the monopile is finite element program) with parameters specified by the user.
represented by Timoshenko beam theory, which allows the
shear strains in the pile to be incorporated in the analysis in an 2.2 Selection and calibration of the soil reaction curves
approximate way. Since the influence of the shear strains on the
overall pile deformation is likely to increase as L/D is reduced, The functions selected to represent the soil reaction curves are,
the use of Timoshenko theory provides a means of maintaining to an extent, arbitrary. They should, however, be capable of
the robustness of the approach as the embedded length of the representing the soil reactions for behaviour ranging from small
monopile reduces (or the diameter is increased). displacements (needed, for example, to predict the dynamic
Consistent with the conventional p-y method, the soil response and natural frequencies of a wind turbine structure) to
reactions are applied to the embedded beam on the basis of the the large displacement response (required for the calculation of
Winkler assumption, i.e. in which the p and components the Ultimate Limit State, ULS). The conventional cube root
are specified to be functions only of the local pile displacement, function for p-y curves in clay (DNV 2016) is unsuitable
, and m and are specified to be functions only of the local (unless modifications are introduced) since it implies an infinite
pile cross-section rotation, . Functions relating the soil initial stiffness. The two-parameter hyperbolic tangent function
reactions and the local pile displacements (or rotations) are typically adopted for p-y curves in sand (DNV 2016) employs
termed ‘soil reaction curves’. Although the Winkler approach parameters to control both the initial stiffness and the ultimate
neglects the coupling that inevitably occurs between adjacent value of the distributed lateral load. However, it is not possible
soil layers, it provides a convenient basis for design to tune this function to match the shape of the response between
calculations, as demonstrated by the widespread adoption of the these two limits, or to specify the magnitude of displacement
p-y method. needed to mobilise the ultimate value of distributed load.
It should be noted, however, that soil reaction curves In the current work, a four-parameter conic function –
determined on the basis of the Winkler approach are unlikely to described in further detail later – is employed for each of the
be unique. Appropriate soil reaction curves may depend, for soil reaction curves in the design model. This four-parameter
example, on the relative magnitude of the translational and function appears to provide a reasonable compromise between
rotational movements of the pile. It is desirable, therefore, to ability to represent the soil reactions at an appropriate level of
calibrate the soil reaction curves using pile deformation modes detail and the desirability (from a practical perspective) of
that are representative of those that are expected to be minimizing the total number of parameters in the model.
experienced by full-scale wind turbine monopile foundations. Ideally, the soil reaction curves employed in the PISA
The PISA design model reduces to the standard p-y approach design model would be calibrated directly, using the results of
when m, and are set to zero (and appropriate choices field testing on full-sized monopiles. However, conducting
are made on the relationship between p and the local lateral pile experiments on full-sized structures would be prohibitively
displacement, ). Experience has shown, however, that m, expensive and it would be impractical to devise a test
and become increasingly important as L/D is reduced programme that encompasses all of the relevant soil, pile and
(Byrne et al. 2015a). The distributed moment component, for loading parameters. Also, considerable technical difficulties
example, depends on the pile diameter; it increases in exist in devising instrumentation systems to determine the
magnitude as the pile diameter is increased. Similarly the force various soil reaction components from the measured pile
and moment reactions and at the base of the pile performance.
become increasingly significant as L/D is reduced. The four- In PISA, an alternative approach was adopted, in which the
component model in Figure 1 therefore provides a rational way calibration of the soil reaction curves in the 1D design model
of addressing a feature of the p-y method, that has come to be was related, indirectly, to the results of the field testing
known as the ‘diameter effect’, in which the standard p-y curves campaigns conducted during the project, as illustrated in Figure
(e.g. API 2010, DNV 2016) are typically found to become 3. This model calibration process, which employed three
increasingly unreliable as the pile diameter is increased, or the separate activities (field testing, 3D finite element modelling
pile length is reduced (e.g. Alderlieste et al. 2011, Doherty and and 1D model development) is summarized in general terms
below.
Field tests involving lateral loading of monopiles were
conducted, at a reduced scale, at the two selected test sites
(Cowden and Dunkirk). Bespoke 3D finite element models
Rotation, The vertical shear traction

varies around the pile Site
perimeter, leading to a investigation
distributed moment, m. Calibrate
dz Validate
3D finite
Constitutive
Field tests element
modelling
analysis
Pile
Calibration of 1D
design model
Figure 2. Diagrammatic view of the vertical tractions acting at the soil-
pile interface for an elemental length of pile, dz, These tractions are
assumed equivalent to a distributed moment reaction, m. Figure 3. PISA design model development process.
37
were developed for several of the field test piles. The The piles were instrumented with a range of above and below
constitutive models that were selected for the analyses, to ground instrumentation, including inclinometers, optical fibre
reproduce the behaviour of the two soils, were calibrated using Bragg grating strain gauges and extensometer strain gauges.
the results of advanced site investigations employing pre- Testing was conducted using the arrangement shown
existing data combined with intensive new laboratory and field diagrammatically in Figure 4. The test piles were loaded at a
investigations conducted during the PISA project (Zdravković height h (termed ‘load eccentricity’) above ground via a
et al. 2015). Data obtained from the field tests were compared hydraulic ram reacting against a reaction pile. Most of the field
with the 3D finite element results to confirm the veracity of the tests employed a protocol in which the applied horizontal
3D modelling procedures; this process is indicated as ‘validate’ load, , was controlled to apply a constant ground-level
in Figure 3. A key feature of this validate process is that velocity of D/300 per minute to the pile. At various stages
artificial means of achieving a match between the numerical during these constant velocity tests, the applied load was held
model and the field tests were rigorously rejected. constant, to allow observations to be made of time-dependent
Once the 3D finite element simulations of the field tests had behavior (e.g. due to creep and/or consolidation). A test was
been completed, a separate 3D finite element parametric study considered complete when (i) the ground-level pile
(referred to below as the ‘calibration analyses’) was conducted. displacement exceeded /10 and (ii) the ground-level
These calibration analyses adopted homogeneous clay and sand pile rotation exceeded 2°. A few tests were conducted at
profiles that were based on the soil conditions at each of the two elevated displacement rates and a limited amount of cyclic
test sites, but adjusted to obtain soil profiles that are testing was also conducted. The 3D finite element models of the
representative of offshore conditions. The soil at the onshore field tests, however, were validated only with respect to the
Dunkirk test site, for example, was interpreted to have a (small) constant velocity tests. Further details of the field testing
suction near the surface. Since this is unrepresentative of campaign are given in Byrne et al. 2015b.
submerged offshore conditions, a surface layer with negative
pore pressures was not included in the 3D finite element
calibration analyses employed to develop the PISA sand model. 4 3D FINITE ELEMENT MODELLING
The calibration analyses adopted a range of pile geometries and
loading conditions that were judged to span the likely design 4.1 3D finite element analysis of the Dunkirk tests
space for full-sized monopiles.
Numerical data from the 3D finite element calibration 3D finite element models were developed for several of the test
analyses were used, directly, to calibrate the soil reaction curves piles at the Dunkirk site. Analyses were conducted using the
employed in the PISA 1D model. This approach provides an finite element program ICFEP (Potts and Zdravković 1999),
indirect link between the field test data and the 1D model, via adopting the modelling procedures outlined in Zdravković et al.
the use of 3D finite element modelling procedures. Data 2015.
processing activities that would be infeasible for the field data The chosen constitutive model was a bounding surface
(such as extracting the individual soil reaction curves from the plasticity model (Taborda et al. 2014). This is a critical state
soil-pile interface tractions) can be conducted straightforwardly model within the state parameter framework, capable of
when applied to the 3D finite element results. reproducing the stress level- and void ratio-dependent
The current paper provides a description of the application behaviour of sands. The model was calibrated using the earlier
of this process to calibrate the PISA 1D model for a research by Imperial College, combined with recent triaxial test
homogenous sand profile with relative density DR = 75% based data on Dunkirk sand reported in Aghakouchak 2015 and
on the Dunkirk field tests, employing the procedures indicated Aghakouchak et al. 2015 as well as the work conducted for
in Figure 3. The paper also demonstrates the application of PISA described by Liu et al. 2017. The interface between the
the 1D model to two design cases with pile parameters that pile and the soil was represented by an elasto-plastic Mohr-
differ from those employed in the 3D finite element calibration Coulomb model with zero cohesion and a friction angle of 32o.
process. Detailed consideration of the near-surface ground conditions
indicated that the 3 m thick hydraulic fill layer developed
higher CPT qc resistances than had been found in the mid-to-
3 DUNKIRK FIELD TESTING CAMPAIGN
3.1 Site details

The Dunkirk test site is located in a coastal area in northern TEST PILE REACTION PILE
H
France, near to the town’s Port Ouest. Earlier laboratory and
field research described by Chow 1997, Kuwano 1999, Jardine
et al. 2006, Ahgakouchak 2015 and Ahgakouchak et al. 2015
showed that the site consists principally of a dense Flandrian
sand with a surface layer (about 3m thick) of dense h
hydraulically-placed sand that has the same geological origin as
the deeper Flandrian deposit. New CPTu and seismic cone
soundings conducted for PISA and new advanced laboratory
tests described by Liu et al. 2017 show that the hydraulic fill
and Flandrian sand layers of the soil present relative densities of
100% and 75% respectively, with a critical state friction angle z
of 32 . The laboratory tests provide extensive L
information on the sands stress-dilatancy behaviour and highly
non-linear stiffness characteristics.
3.2 Testing details D
Monopiles with diameter D = 0.273 m, 0.762 m and 2.0 m, and

Figure 4. Illustration of the pile testing system employed at the Cowden
values of L/D between 2 and 10 were installed at the test site.
and Dunkirk sites.
38
late 1990s investigations described by Chow 1997 and Jardine hold periods that were prescribed during the testing process.
et al. 2006. It was postulated that the additional shallow Also shown in Figure 5a are an unload-reload loop conducted
resistance was a result of either ageing in the hydraulic fill, soon after the start of the test, and the unloading response of the
local spatial variations, light cementation between the sand pile at the end of the test. The response computed using the 3D
grains, or of suctions due to partial saturation. Detailed finite element analysis is also shown in Figure 5a; this shows
investigation to fully explain these variations was not feasible. good agreement with the field data. Comparisons of a similar
However, piezocone tests did show clear signs of suctions quality were obtained for the field test data and the
developing at shallow depths and the analyses therefore corresponding finite element models for the two large diameter
employed a realistic limited suction (i.e. pore pressures less (D = 2 m) piles tested at the site.
than hydrostatic) in the soil above the level of the water table Figure 5b shows the measured bending moments induced
(estimated to be at a depth of 5.4 m below ground level). It is in pile DM4, deduced from the strain gauge instrumentation, at
noted that the additional complexities involved in modelling the start of the hold period for 192 kN. The distribution of
unsaturated surface layers are absent from offshore sites, where bending moments determined from the 3D model at this same
soils are typically assumed to be fully saturated. value of applied horizontal load is also shown in the figure. The
Example field data for a D = 0.762 m, L = 4 m pile agreement between the two sets of data appears reasonable.
(identified as pile DM4) tested at the Dunkirk site, are shown in The finite element results were found to be less consistent
Figure 5. Figure 5a shows the relationship between the applied with the field data for the shortest D = 0.762 m pile (with L =
lateral load H and the lateral displacement of the pile at ground 2.3 m) that was tested at this site. For this relatively short pile,
level, . The constant load steps in the field data indicate the the finite element results are highly dependent on the initial
conditions that are assigned to the unsaturated surface layer.
Since some uncertainty existed on the appropriate conditions to
(a) 300 apply to this surface layer (as a consequence of limitations in
Field data the available site investigation data) the finite element results
are regarded as being less robust for this particular pile.
250 3D FE The broadly satisfactory comparison between the field test
data and the numerical analysis supports the use of the 3D finite
200 element model to calibrate the soil reaction curves in the 1D
design model, as described below.
H (kN)
150 4.2 3D finite element calibration study

The calibration analyses were based on the soil conditions at the
100 Dunkirk test site, together with the constitutive model and
associated constitutive parameters that were employed in the
finite element analyses of the field tests. Certain adjustments to
50 the Dunkirk soil profile were required, however, to ensure that
the profile employed in the calibration analyses was
0.1D representative of an offshore homogeneous sand site. In
0
particular, in the calibration study, hydrostatic pore pressures
0 20 40 60 80 100 were assigned to the entire soil profile (noting that a region with
vG (mm)
a small suction above the water table was adopted to model the
Dunkirk field tests). Although the Dunkirk site profile shows
(b) variations in sand state with depth, uniform relative density of
Moment (kNm) 75% was assigned to the soil for these calibration analyses.
0 500 1000 1500 2000 2500 Eleven calibration analyses were conducted for monopile
0 dimensions and load eccentricities in the range 5m 10m,
2 ⁄ 6, 5 ⁄ 15. A typical mesh employed for the
0.5
1
1.5
Depth (m)
2
2.5
3
Surface M
3.5 Strain gauges
Spline fit
4
3D FE
4.5
Figure 5. Comparisons between the field measurements and 3D finite

element results for pile DM4 (D = 0.762 m, L = 4 m) tested at Dunkirk;
(a) horizontal load, H, vs. ground displacement, , response (b) below-
ground bending moments at 192 kN, 41 mm. In (b) ‘Surface
M’ indicates the bending moment determined from the value of the
horizontal load, ‘Strain gauges’ refers to bending moments inferred
from the fibre optic Bragg grating strain gauges, ‘Spline fit’ indicates a Figure 6. 3D finite element mesh for pile C4 (D = 10 m, L = 60 m, h =
spline that has been fitted to the bending moment data. 50 m) in the calibration analysis set.
39
calibration analyses is shown in Figure 6. in terms of the normalized variables listed in Table 1. The four-
parameter conic function used to represent the soil reaction
curves is illustrated in Figure 7, where ̅ signifies a
5 1D MODEL DEVELOPMENT FOR SAND normalized displacement or rotation variable and signifies
the corresponding normalized soil reaction component. The
5.1 Soil reaction curves conic function is calibrated by the specification of four
parameters ( , , ̅ , , each of which has a straightforward
Numerical representations of the soil reaction curves (referred interpretation. The parameter specifies the initial slope;
to below as ‘numerical soil reaction curves’) were determined is the ultimate value of the normalized soil reaction and ̅ is
from the 3D finite element calibration analyses using a process the normalized displacement (or rotation) at which this ultimate
in which nodal forces acting at the soil-pile interface, and value of soil reaction is reached. The parameter (0 1
stresses in the interface elements between the pile and the soil, determines the shape of the curve.
were extracted. Numerical data on the distributed lateral load, p, Values of the parameters defining the soil reaction curves
were determined by integrating the x-components (using the for each of the soil reaction components were determined via an
coordinate system in Figure 6) of the horizontal tractions acting automatic optimization process conducted over the complete set
on the pile at discrete values of depth along the pile. Data on the of eleven 3D finite element calibration analyses. In conducting
distributed moment were obtained by integrating the vertical this optimization, the soil reaction curve parameters for ̅ and
soil-pile tractions, accounting for the distance between the pile were assumed to vary linearly with depth along the pile.
neutral axis and the point on the pile perimeter where the Initial values for the model parameters were determined by
traction is applied. The force and moment reactions at the pile least-squares fitting of the numerical soil reaction curves,
base were obtained by integration of the stresses in the layer of moderated by eye. The calibration was further improved by
soil elements immediately below the pile base. allowing adjustments to these parameters to optimize the fit
To proceed, it is necessary to use appropriate non- between the vs responses computed using the 3D finite
dimensional parameters to process the numerical results and element model and the 1D model for 0 0.1 .
also to formulate the soil reaction curves. The non-dimensional
forms employed for the PISA sand model are listed in Table 1, 5.2 Example soil reaction curves
where is the local value of initial vertical effective stress,
G is the local value of small-strain shear modulus, and Two contrasting examples of the process of fitting the four-
are the local pile lateral displacement and cross-section rotation parameter conic function to the data extracted from the 3D
respectively. The normalization process for the distributed finite element calibration analyses are shown in Figure 8 and
moment follows a pattern that differs from the other three soil Figure 9.
reaction components. It was identified, when reviewing the Figure 8 shows example data on normalized distributed
numerical soil reaction curves, that the distributed moments lateral load, ̅ , at various pile depths, z, for a calibration
appeared to scale conveniently with the current value of the calculation (pile C4) with D = 10 m, L = 60 m. At shallow
local distributed lateral load, p. Since the vertical tractions depths, where soil displacements are relatively large, a peak is
induced on the pile arise as a consequence of friction at the soil- typically found to occur in the response, followed by post-peak
pile interface, it seems plausible that the magnitude of m is softening. An example of this type of response, for ⁄
correlated with the normal tractions applied to the pile. The 0.23 is shown in Figure 8a. This behaviour is closely linked to
normal tractions are, themselves, closely associated with the the dilational characteristics of the soil. Since softening
distributed lateral load, p; it therefore seemed appropriate in the behavior cannot be represented with the four-parameter conic
current modelling to normalize the distributed moment, m, with function employed in the 1D model, it was necessary to make
the local value of the distributed load, . The use of the non- an arbitrary choice on the ultimate value ̅ of the normalized
dimensional form in Table 1 implies that the distributed distributed load for incorporation in the model. In the current
moment m is a product of the current value of p and a separate work, when post-peak softening was observed, ̅ was taken
function of the local rotation, . Although this adds to the as an intermediate value between the peak and final values
complexity of the 1D model, this form of soil reaction curve is determined from the 3D finite element calibration data. The
incorporated straightforwardly within the 1D finite element resulting parametric soil reaction curves (determined by
formulation used in the model. optimizing over the complete set of eleven calibration analyses)
The soil reaction curves employed in the 1D model, referred are plotted for a range of depths, z, for the full range of soil
to below as the ‘parametric soil reaction curves’ are formulated displacements developed in the calibration analysis in Figure
8a, and for small displacements in Figure 8b. Differences are
Table 1. Parameter normalization. shown to exist between the data from the 3D calibration
analysis and the resulting calibrated soil reaction curves. As the
Normalized variable Non-dimensional form
parametric curves are tailored to provide a representation of the
Distributed lateral load, ̅
Lateral displacement, ̅
Distributed moment,
Pile rotation,
Base shear load,
Base moment, Figure 7. Four-parameter conic function employed to represent the soil
reaction curves.
40
3D finite element data across the complete set of calibration calibration data for individual piles at a local level.
analyses, they therefore can sometimes exhibit the tendency, In contrast, Figure 9 indicates the match between the
apparent in Figure 8, to depart significantly from the 3D parametric soil reaction curve and the numerical data for the
base moment for pile C4; in case of this component of soil
reaction, the two data sets are seen to agree well.
(a)
30
5.3 1D model analysis of the calibration cases
Normalized distributed laod p/σ'viD
25 Example comparisons of the computed vs response for

two of the D = 10 m diameter calibration piles are shown in
20 Figure 10; pile C1 has length 20 m and pile C4 has length 60 m.
In spite of the apparently poor performance of the four-
parameter function in representing aspects of the computed
15
lateral soil reaction curves for pile C4, as shown in Figure 8, the
overall performance of the 1D model – in terms of the extent to
10 z/D=0.23 which 1D predictions of the H vs response agree with the
z/D=2.08 3D calibration data for piles C1 and C4 – is seen to be
5 z/D=4.83 excellent. Similarly close comparisons between the 1D and 3D
results were obtained for all of the other calibration analyses.
z/D=5.97
This exercise indicates that the 1D model is able to
0 reproduce the overall behavior of the calibration piles, although
0 50 100 150 200 at a local level, significant differences can exist between the
Normalized displacement vG/Dσ'vi calibration data and the parametric soil reaction curves. This
well-conditioned aspect of the 1D model is considered to be due
to the overall performance being obtained by integrating the soil
(b) reaction curves along the entire length of the foundation.
10
z/D=0.23
Normalized distributed laod p/σ'viD
9
z/D=2.08
8 (a)
z/D=4.83 35
7 3D FE
z/D=5.97
30 1D design model
6
5 25
4
20
H (MN)
3
2 15
1
10
0
0 1 2 3 4 5 5
Normalized displacement vG/Dσ'vi 0.1D
0
Figure 8. Example soil reaction curves for normalized distributed lateral 0 0.2 0.4 0.6 0.8 1 1.2
load for calibration pile C4 (D = 10 m, L = 60 m). Solid line are data vG (m)
from the 3D calibration analysis of the pile, dashed lines are the
calibrated soil reaction curves; (a) large displacements, (b) small
displacements. (b)
250
3D FE
0.06
1D design model
200
Normalized moment MB/σ'viD3
0.05
0.04 150
H (MN)
0.03
100
3D FE
0.02
Soil reaction curve
50
0.01
0.1D
0 0
0 5 10 15 20 25 0 0.2 0.4 0.6 0.8 1 1.2
vG (m)
Normalized rotation ψG/σ'vi
Figure 9. Parametric soil reaction curve (shown as a dashed line) and Figure 10. Comparisons between the 3D finite element calibration
3D finite element calibration data (shown as a solid line) for the base analyses and the 1D model computed response (a) for pile C1 (D = 10
moment MB, for pile C4 (D =10 m, L = 60 m). m, L = 20 m) (b) for pile C4 (D = 10 m, L = 60 m).
41
Provided that significant systematic errors are absent from the The calibration process has been demonstrated for an example
1D model, the averaging process employed in the modelling offshore site where the soil is a uniform, dense sand. In this
procedure appears to have the consequence that the model is example, the 1D model is shown to provide a close
remarkably tolerant of imperfect fitting of the data at a local representation of the overall pile behaviour for each of the
level. calibration analyses. Comparing the 1D model with the data
that were used to calibrate it does not, in itself, provide any
evidence of its predictive capability. This comparison exercise
6 DESIGN EXAMPLE does, however, indicate that the various approximations and
assumptions inherent in the 1D model do not detract
Once the 1D model has been calibrated – or ‘trained’ – it can be significantly from its reliability.
used to determine the performance of a monopile foundation for The predictive capability of the 1D model has been
arbitrary values of geometry and loading parameters that lie demonstrated by means of two independent design examples
within the calibration space. To demonstrate the predictive with pile parameters that differ from those employed in the
capability of the model, two separate example design analyses calibration set (although within the bounds of the calibration
have been considered. The geometric configurations adopted space).
for these test cases, specified in Table 2 (where t is pile wall It is suggested that the PISA modelling approach could be
thickness), were selected to fall within the parameter space employed for monopile design in one of two ways. For initial
adopted for the calibration analyses, as indicated in Figure 11. design calculations, it may be appropriate to employ pre-
defined functions and parameters to represent the soil reaction
Table 2. Pile parameters selected for the two design examples. curves based on the soil profiles established for any given site
Reference D (m) H (m) L (m) t (mm)
D1 7.5 37.5 22.5 68 (a)
D2 8.75 8.75 35 91 30
The load-displacement responses computed using the 1D

model and subsequently the 3D finite element model for both of 25
the design examples, D1 and D2, are shown in Figure 12. The
results indicate a close match between the two sets of data over
the full range of applied loading (up to a ground level pile 20
displacement of 0.1 ). Figure 13 shows the lateral
H (MN)
displacements induced in the embedded portions of the piles for 15

an applied horizontal load of 0.75 , where is
the value of horizontal load at 0.1 determined from the
3D finite element analysis. The two sets of data are seen to 10
agree well. The close agreement between the computed 3D FE
responses obtained using the 3D and 1D models for these two 1D design model
design cases is consistent with the assumption implicit in the 5
PISA methodology, that the 1D model provides an efficient
means of interpolating the overall pile response computed using 0.1D
the 3D calibration calculations to other pile geometries within 0
the calibration space. 0 0.2 0.4 0.6 0.8
7 DISCUSSION vG (m)
A method has been presented to calibrate a 1D model of

monopile behaviour using a suite of 3D calibration analyses. (b)
60
20
18 50
16
40
14
Load eccentricity, h
H (MN)
12 30
Field tests
10 Calibration analyses
8 20
Design cases 3D FE
6 1D design model
10
4
0.1D
2 0
0 0.2 0.4 0.6 0.8 1
0
vG (m)
0 2 4 6 8 10
L/D
Figure 12. Comparisons between load-displacement responses
Figure 11. Parametric geometry space for the field tests, the calibration computed using the 3D finite element model and the 1D model for the
analyses and the design examples. two design examples (a) D1 (b) D2.
42
as characterized by relatively simple index or other testing. 8 ACKNOWLEDGEMENTS

(This process is broadly similar to current forms of the p-y
method). In the PISA terminology, this approach is referred to The PISA Project was funded by the UK Department for
as the ‘rule-based’ method. Energy and Climate Change (DECC) and the PISA Industry
For more detailed and robust design calculations, an Partners under the umbrella of the Offshore Wind Accelerator
alternative route is proposed in which the site conditions are (OWA) program which was designed and is led by the Carbon
investigated more intensively with advanced sampling, Trust. The Authors acknowledge the provision of financial and
laboratory and field techniques being applied to establish the technical support by the following project partners: Alstom
detailed behavior of the principal soil types present at any given Wind, DONG Energy, E.ON, EDF, Iberdrola, innogy, SSE,
wind farm location. Bespoke 3D finite element calculations Statkraft, Statoil, Van Oord and Vattenfall. The Authors also
may then be performed that span the likely ranges of soil acknowledge the main test contractor for the field test
profiles and the parameter space that will control the final campaign, ESG. Christelle Abadie contributed to the production
design. Soil reaction curves may then be extracted from the of the data presented in the paper on the two design examples.
finite element analyses predictions for the soil-pile interface
tractions. Procedures for this latter approach – which is termed
the ‘numerical-based’ method – are demonstrated in the current 9 REFERENCES
paper for a homogeneous dense sand site. The numerical-based
approach is essentially a procedure to train a relatively simple Aghakouchak A. 2015. Advanced laboratory studies to explore the
axial cyclic behaviour of driven piles. PhD thesis, Imperial College
calculation (the PISA 1D model) using data from more detailed
London.
3D finite element analyses. The 1D calculation is rapid to
compute, with accuracy that is linked to the fidelity of the 3D Aghakouchak A., Sim W.W. and Jardine R.J. 2015. Stress-path
laboratory tests to characterise the cyclic behaviour of piles driven
finite element models that are employed in the calibration
process. This approach means that full use can be made of any in sands. Soils & Foundations. 44(5), 917-928.
site investigation data (via the constitutive model employed in Alderlieste E. A., Dijkstra, J. and Van Tol A. F. 2011. Experimental
the 3D calibration analyses) in the formulation of the soil investigation into pile diameter effects of laterally loaded mono-
reaction curves. Moreover, the method can evolve with future piles. In ASME 2011 30th International Conference on Ocean,
developments in site investigation, constitutive modelling and Offshore and Arctic Engineering. American Society of Mechanical
finite element analysis. Engineers, 985-990.
The PISA modelling process is suitable for assessments of API 2010. RP 2A-WSD – Recommended Practice for Planning,
the ULS performance as well as for predicting the small- Designing and Constructing Fixed Offshore Platforms.
displacement dynamic performance of an offshore wind turbine Washington: American Petroleum Institute.
structure. The current paper is limited to the modelling of Byrne B.W., McAdam R.A., Burd H.J., Houlsby G.T., Martin C.M.,
monopile behaviour in a dense sand (although a similar Beuckelaers W.J.A.P., Zdravković L., Taborda D.M.G., Potts
calibration study, not reported here, has been completed for an D.M., Jardine R.J., Ushev E., Liu T., Abadias D., Gavin K., Igoe
overconsolidated clay till). Further development work is needed D., Doherty P., Skov Gretlund J., Pacheco Andrade M., Muir Wood
to extend the method to other soil types, included layered soils, A., Schroeder F.C., Turner S. & Plummer M.A.L. 2017. PISA: new
and to include the effects of cyclic loading. design methods for offshore wind turbine monopiles. Proceedings
of the Society for Underwater Technology Offshore Site
Investigation and Geotechnics 8th International Conference on
Lateral pile displacement, v (m) “Smarter Solutions for Future Offshore Developments”, London.
‐0.2 0 0.2 0.4 0.6 Byrne B. W., McAdam R., Burd H. J., Houlsby G. T., Martin C. M.,
0 Zdravković L., Taborda D. M. G., Potts D. M., Jardine R. J., Sideri
M., Schroeder F. C., Gavin K., Doherty P., Igoe D., Muir Wood A.,
Kellahave D. and Skov Gretlund J. 2015a. New design methods for
5 large diameter piles under lateral loading for offshore wind
applications. Proceedings of Third International Symposium on
Frontiers in Offshore Geotechics 1, 705-710.
10 Byrne B. W., McAdam R., Burd H. J., Houlsby G. T., Martin C. M.,
Gavin K., Doherty P., Igoe D., Zdravković L., Taborda D. M. G.,
Potts D. M., Jardine R. J., Sideri M., Schroeder F. C., Muir Wood
Depth (m)
15 A., Kellahave D. and Skov Gretlund J. 2015b. Field testing of large

D1 diameter piles under lateral loading for offshore wind applications.
Proceedings of XVI European Conference on Soil Mechanics and
20
Geotechnical Engineering, Edinburgh, 1255-1260.
Chow F.C. 1997. Investigations into displacement pile behaviour for
25 offshore foundations. Ph.D Thesis, Imperial College London.
DNV, 2016. DNVGL-ST-0126 – Support structures for wind turbines.
Doherty P. and Gavin K. 2011. Laterally loaded monopile design for
30 offshore wind farms. Proceedings of the ICE – Energy 165(EN1),
3D FE
7-17.
1D design model Jardine R.J, Standing J.R and Chow F.C. 2006. Some observations of
35 D2 the effects of time on the capacity of piles driven in sand.
Géotechnique 55(4), 227-244.
Kuwano, R. 1999. The stiffness and yielding anisotropy of sand. PhD
40 Thesis, Imperial College London.
Lam, I.P.O. 2013. Diameter effects on p-y curves. Deep Marine
Foundations - A Perspective on the Design and Construction of
Figure 13. Below-ground lateral displacements for design examples D1 Deep Marine Foundations.
and D2 computed using the 3D finite element model and the calibrated
1D design model. The D1 and D2 results correspond to H = 22.1 MN
and 36.15 MN respectively.
43
Liu T.F, Aghakouchak A., Taborda D.M.G. and Jardine R.J. 2017. 160.
Advanced laboratory characterization of a fine marine sand from Zdravković L., Taborda D. M. G., Potts D. M., Jardine R. J., Sideri M.,
Dunkirk, France. Proc. 19th ICSMGE, Seoul. Schroeder F. C., Byrne B. W., McAdam R., Burd H. J., Houlsby G.
Potts D. M. and Zdravković L. 1999. Finite element analysis in T., Martin C. M., Gavin K., Doherty P., Igoe D., Muir Wood A.,
geotechnical engineering: theory. London: Thomas Telford. Kellehave D. and Skov Gretlund J. 2015. Numerical modelling of
Taborda D.M.G., Zdravković L., Kontoe S. and Potts D.M. 2014. large diameter piles under lateral loading for offshore wind
Computational study on the modification of a bounding surface applications Proceedings of Third International Symposium on
plasticity model for sands. Computers and Geotechnics 59, 145- Frontiers in Offshore Geotechnics 1, 759-764.
44
Design Aspects of Suction Caissons for Offshore Wind Turbine Foundations
Aspects de conception des caissons d’aspiration pour les fondations de turbines éoliennes en mer
Sturm, Hendrik
Computational Geomechanics, Norwegian Geotechnial Institute (NGI), Oslo, Norway, [email protected]
ABSTRACT: This paper provides an introduction to the geotechnical design of suction caisson foundations for Offshore Wind Turbine
(OWT) foundations. It summarizes the experience gained in a number of projects from across the world and proposes a guidance for the
design of future projects. The paper is structured in a logical manner; the first section introduces the general design approach of suction
caisson foundations, whereas the individual design aspects are discussed in detail in the subsequent sections. Therein, all relevant
aspects are covered, including design basis, installation-, capacity- and serviceability-analysis, assessment of the foundation stiffness,
and soil reactions. In the last section other aspects such a grouting, integrated analysis, and application of the presented approach to
complete wind farms is briefly discussed.
RÉSUMÉ: Ce papier introduit la conception géotechnique de fondations de caissons de succion utilisés dans les fondations des turbines
des éoliennes en mer. Cet article résume l’expérience acquise au cours de projets menés à travers le monde et propose quelques conseils
pour l’élaboration de projets futurs. Ce papier est structuré en trois sections. Dans la première partie, différentes approches utilisées
lors de la conception des caissons de succion des fondations sont présentées de manière générale. Les aspects individuels et particuliers
de la construction sont expliqués en détails plus loin dans cette même section. Tous les aspects pertinents sont couverts allant de la
conception à l’analyse de l’installation, de la capacité et de la maintenance à l’évaluation de la rigidité de la fondation et des réactions
du sol. Dans la dernier section, d’autres aspects, tels que le ciment, la conception intégrée, et l’application de l’approche présentée à un
parc éolien complet sont discutés.
KEYWORDS: suction caissons, offshore wind, design
MOTS-CLES: caissons de succion, éoliennes en mer, design
1 INTRODUCTION rently developed standards and guidelines for certification and ap-
proval. Not included in this contribution are detailed descriptions
All major offshore wind energy developers worldwide are cur- of design methodologies as they are widely discussed in the many
rently investigating alternatives to the Monopile concept, which is other publications. However, some references to relevant design
widely used for the foundation of Offshore Wind Turbines (OWT). methodologies are included. Main focus is to outline OWT-specific
This effort is driven by technical considerations – mainly increas- design aspects, for both caissons for jackets and mono-caissons.
ing turbine capacities and deeper waters at future wind parks – as
well as environmental and economical considerations. A promis- 1.1 General design approach
ing foundation concept is the so-called Suction Caisson; a hollow
steel cylinder closed at the top and opened at the bottom. Suction Suction caissons are used since the 1980s in the Oil & Gas (O&G)
caissons are installed by means of the self-weight of the structure industry as the foundation of both bottom fixed and floating off-
and a suction pressure applied inside the caisson. Once installed, shore structures. It is estimated that by the end of 2010 more than
they resist environmental loads like an embedded shallow founda- 1000 permanent offshore suction caissons and anchors were in-
tion, but can also temporarily mobilize considerable suction, which stalled.
further increases the capacity and stiffness. In the last decades a vast amount of articles and journal pa-
pers were published presenting results of research work and prac-
Though suction caissons are already used since several decades,
tical experience with suction caissons and anchors. Most of these
practical experience with the short- and long-term behavior of
are addressing particularly deep-water application cases. While in
these foundations used for OWTs is limited so far. Notwithstand-
the early years mainly suction caissons in clayey soils were con-
ing the lack of experience, a number of projects have been initiated
sidered, also sandy and layered soils came into the focus in the
where suction caissons have been or will be applied. The Norwe-
more recent years. Most publications present theoretical and nu-
gian Geotechnical Institute (NGI) has been involved in most of
merical studies as well as small-scale 1g or N g model tests (e.g.
these projects, including Borkum Riffgrund 1 (BKR01), Borkum
Byrne 2000, Johansson et al. 2003, Kelly et al. 2006, Jostad et al.
Riffgrund 2 (BKR02), Hornsea 1 (HOW01), Aberdeen Offshore
2015a). Only limited measurement data is found from actually
Wind Farm (EOWDC), Hywind Scotland Pilot Park, and South-
built structures. Some examples of installation data are report by
west Offshore Demonstration Wind Farm (SWK), providing vari-
Sparrevik (2002), Colliat et al. (2007), Aas et al. (2009), Langford
ous services such as laboratory testing, geotechnical design, suc-
et al. (2012), Solhjell et al. (2014), Saue et al. (2017), and in-place
tion installation support, and health monitoring systems. The ex-
measurement data on prototypes by Schonberg et al. (2017), Svanø
perience gained in these and other projects forms the basis for the
et al. (1997).
presented work.
The experience gained in the last 30 years from the O&G in-
The objective of this paper is to provide an overview of the
dustry provides a good basis for the design of suction caissons for
particular design-requirements and -challenges of suction caissons
OWTs. However, there are a number of important aspects, which
for the foundation of OWTs, and should assist decision makers to
are different, and which require particular consideration in the de-
consider this foundation concept in future wind farm projects. The
sign of caissons for OWTs:
presented design aspects and recommendations can be directly ap-
plied in ongoing and future projects, and provides a basis for cur- • Most offshore wind farms are located in relatively shallow
45
waters where the sub-surface has been exposed in the more re- both the geotechnical analysis and other involved disciplines. The
cent geological history to significant environmental changes structural designer may update the properties of the caisson and
such as glacial periods, dry periods and floods, yielding pro- the sub-structure, the turbine manufacturer may update the (cyclic)
nounced soil layering comprising a large range of different loads, and the soil layering and properties may be complemented
soil types and properties (e.g. Cotterill et al. 2017, Dove et al. by updated field and laboratory test data, to name a few.
2016). As a result, soil profiles may vary significantly both in The workflow of the (geotechnical) design approach illustrated
depth and horizontally. in Figure 1 is not very much different to that of any other foun-
dation. However, it is important to be aware of the interdepen-
• The loading conditions are different for OWT foundations. dency, as this pose a natural limitation on the achievable optimiza-
With increasing turbine size operational and other load cases tion. A typical project comprises different phases; e.i. feasibil-
can govern the geotechnical design, being potentially more ity study, pre-FEED1 , FEED and Detailed Design. Each of these
severe than a conventional 50-, or 100-years storm event, phases can comprise one or several iteration(s). Current research
which is typically used in the design of offshore O&G struc- aims to solve some of the activities in an integrated manner (e.g.
tures. In addition, these design-critical load cases may have Krathe & Kaynia 2016, Page et al. 2016, Skau et al. 2017). That
considerably recurrence rates during the lifetime of an OWT. means it is tried to model the complete OWT in one analysis to
capture the interdependency. However, all parts, and in particular
• The response of the sub-structure of an OWT is very sensi-
the soil-foundation-system, is often represented in these analysis
tive to the foundation behavior, i.e. stiffness and (differential)
in a simplified way in order to limit the required calculation time.
settlements. Although this can be an important design as-
Thus an integrated analysis may not be suitable for an optimiza-
pects for O&G structures, it is in general more important for
tion, but can be very beneficial for other aspects, in particular for
OWTs due to the high-cyclic loading conditions during oper-
the assessment of loads.
ation and the sensitivity of the turbine on a tilt.
To complicate matter, the supposed conservative assumptions 1.2 Interface between disciplines
made in the geotechnical design in order to cope with these and The iterative design approach illustrated in Figure 1 requires a
further challenges are not necessarily conservative for the struc- physical interface between the different disciplines at which in-
tural design – and vise versa, for apparently conservative assump- put, or output, respectively, is exchanged. There are in principal
tions made in the structural design. Thus, input and assumptions two types of information which need to be exchange between the
in both the geotechnical and the the structural design need to be geotechnical and structural designer:
aligned and consistent.
• The geotechnical designer gets loads and delivers back the
Design Basis: Soil layers and properties, Loads, etc. corresponding deformations, i.e. load-deformation curves.
These curves are practically represented by lumped stiffness
(Assume) foundation dimensions values describing the response of the soil-foundation-system
Iteration with other disciplines
Geotechnical Foundation Design
in one point. The stiffness values are typically provided in

Assess (cyclic) soil design profiles matrix form and can comprise of linear secant stiffness val-
ues or non-linear tangential stiffness values.
Calculate foundation capacity
• The structural designer requires for the caisson design dis-
tributed loads and/or deformations acting on the skirts and
Check installation lid. These distributed loads/deformations are often denoted
Soil Reactions as they describe the response of the soil. Soil
Assess serviceability reactions can be provided as unit loads, total loads or linear
springs (i.e. Winkler-type springs).
Calculate foundation stiffness and soil reactions
Figure 1: Schematic presentation of the iterative and interdependent work- Shaft

2 Lid
flow of suction caissons design
Grout
0.0 m
The consistency is achieved by an iterative design approach as 1

illustrated in Figure 1. The geotechnical design of a suction cais-
son foundation comprises 5 main activities: 1) Assessment of the
cyclic soil properties for the given boundary conditions, i.e. load 3
conditions, foundation geometry, and soil layering and properties;
r z
2) Foundation capacity assessment for short- and long-term load-
ing; 3) Prediction of the installation resistance and corresponding
required suction pressure; 4) Serviceability assessment, i.e. short- Symmetry axis caisson
and long-term settlement, displacement and rotation; and 5) Cal-
Figure 2: Possible interface points for the geotechnical and structural de-
culation of the foundation stiffness including corresponding soil signer
reactions. The activities are interdependent and typically need to
be solved in an iterative manner in order to optimize the caisson
Practically, three different points could be imagined for the
geometry.
load-stiffness exchange, as illustrated in Figure 2. Each point has
Furthermore the geotechnical design is embedded into a design
advantages and disadvantages.
loop interacting with other disciplines. The basis for the geotech-
1
nical design will be continuously updated based on the results of Front End Engineering Design (FEED)
46
1. Traditionally, Point 1, located on the symmetry axis of the above described problem of finding an appropriate interface point.
caisson at mudline, is very often used. However, the structural The structural designer will still need stiffness values at the bottom
designer needs to establish loads at a point which is not con- of the sub-structure.
nected to the structure. In order to do that, he needs to intro- In principal, stiffness values and soil reactions could be estab-
duce a so-called super-element, connecting the structure with lished from the same analysis as they are actually describing the
the ground in this point. Given that the structure – in this case same response. However, the extraction of soil reactions from FE
the caisson lid and grout – is significantly stiffer than the soil analysis is difficult and very sensitive to the modeling technique,
for the considered load level, simplified, linear elastic prop- element-type and -size. As the soil reactions are only used for the
erties can be assigned to the super element. If the flexibility caisson design, but neither for the load assessment nor the design
of the structure is considerably larger and a interaction with of the sub-structure, it has been found most appropriate to establish
the soil behavior may be expected, more complicated proper- reasonable ranges for the distributed loads acting on the skirts and
ties need to be assigned to the super element. However, these lid based on empirical considerations.
properties are very difficult to assess, which may not be pos-
sible. Experience from recent projects has shown, that both 2 DESIGN BASIS
the lid and skirt flexibility is important and an optimization of
the caisson geometry is difficult, for which reason, Point 1 is The design basis is the input to the geotechnical design before any
not recommended to be used in future projects. interpretation or processing is done. It comprises soil properties,
loads, structural properties, guideline requirements, and other rele-
2. Point 2, located at top of the caisson lid in the interface be- vant boundary conditions such as weight- and size-limitations due
tween the shaft of the sub-structure and the caisson, has been to logistical considerations.
used in more recent projects. The advantage is, that Point 2
is also often an interface for the structural design, as the de-
2.1 Site and soil parameters
sign of the caisson and sub-structure is often done separately.
Loads are assessed by the load- or structural-designer using The loading regime acting on a suction caisson requires special
integrated analysis where only the sub-structure is modeled. attention with respect to the soil parameters used in the design.
The soil is therein often represented by set of springs in Point The impact of cyclic loading on the soil strength and stress-strain-
2. That means no super-element is required, but the geotech- behavior needs to be quantified by a thoroughly planned laboratory
nical designer needs to include the lid accurately in his anal- testing program of all relevant soil layers. The following list out-
ysis. lines the recommended minimum site- and soil-investigation pro-
The load-deformation response is complex, meaning that a gram to establish the required soil-profiles and -parameters:
reasonable stiffness matrix describing the load-deformation
• From a geotechnical perspective, a geophysical survey is rec-
of the soil-caisson-system will have both diagonal and off-
ommended to identify the number and depth of the soil layers
diagonal components. However, most programs used for in-
at the OWT location(s). The geophysical survey should pro-
tegrated analysis cannot cope with a full stiffness matrix but
vide an overview of the soil profile variability at a location,
can take only the positive diagonal terms. That means that the
which is in particular relevant for multi-legged sub-structures
soil-foundation response can be only considered in a simpli-
having three or more caissons. In some recent projects, two
fied manner when using Point 2.
surveys have been conducted. In a first survey the complete
3. In order to overcome the shortcoming of using a simplified offshore wind farm was screened, whereas in a second survey
stiffness matrix in the structural analysis, the stiffness matrix high resolution 3d seismic scans of the shallow soil has been
could be provided for the so-called decoupling point, which performed. The advantage of the latter survey is, that it al-
is illustrated in Figure 2 by Point 3. The decoupling point can lows to find also small boulders, which can be critical for the
be assessed in the stiffness analysis as described in Section 7, installation.
and is characterized by the fact that incremental horizontal,
• Minimum one seabed Cone Penetration Test (CPT) per loca-
vertical or moment loads yield only displacements or rota-
tion with a minimum investigation depth za , measured from
tions in the corresponding loading direction. That means that
the skirt tip, where za is the maximum of
the stiffness matrix comprises only positive diagonal terms.
If the geotechnical designer includes in the stiffness analysis – the depth below the caisson where the additional
the caisson with its correct properties, and applies the loads in stresses ∆σv0 due to the permanent weight of the struc-
Point 2, the structural designer can use a rigid super element ture does not exceed 15% - 25% of the in-situ stress
connecting Point 2 with Point 3 and apply the stiffness matrix prior to the installation of the caisson. Assuming a load
in the integrated analysis in Point 3. Though Point 3 seems spread angle of 1:3, a submerged foundation weight be-
to be the most appropriate point for the interface, the problem tween 5 to 7MN, a caisson diameter between D = 8
is, that the location of the decoupling point is not constant but and 10m, a submerged unit weight of the soil of 10 m kN
3,
depends on the load-level, combination of load components and a skirt depth of s = 0.6 · D, the required depth
and load-deformation response. s + za (measured from mudline) varies between 15 and
Based on experience from recent projects, it is recommended 20m.
that the structural designer provides the caisson model and the – the depth of the governing failure mechanism in a bear-
loads in Point 2, and the geotechnical designer delivers back a stiff- ing capacity analysis, which is a function of the caisson
ness matrix in Point 2 and Point 3 as well as the coordinates of diameter D, the number of footings and distance of the
Point 3. legs, and the loading regime. A rotational failure is ex-
In Section 7 is introduced the concepts of a global model. pected for mono-caissons, whereas a compression fail-
Though this model is a considerable improvement as both the sub- ure is expected for caissons supporting a jacket. In both
structure, caisson and soil is modeled, it does not overcome the cases, the depth measured from the skirt tip level is less
47
than the caisson diameter, given that there is no inter- Table 1: Recommended soil data for suction caisson design (after
action between the footings of multi-legged structures. Andersen et al. 2013)
For the dimensions indicated above, the required depth
s + za (measured from mudline) varies between 10 and Soil parameter Clay Sand
14m.
Frictional characteristics
Even though, neither combined deep failure mechanisms of
Peak drained friction angle, ϕ0 x
multi-legged structures, nor exceptionally high weights, have
been observed in past projects, it is recommended to check Residual / critical drained friction angle, ϕ0c x
in the FEED study, whether the values given above are not Undrained friction angle, ϕ0u x
exceeded. That means, it needs to be ensured that the ad- Dilatancy angle, ψ (x)
ditional stresses are not larger, nor the actual failure mode
Slope of DSS drained failure line, α0 x
giving the lowest foundation capacity reaches deeper than as-
sumed. If the required investigation depth cannot be achieved Slope of DSS undrained failure line, αu x
by the seabed CPT, complementary downhole CPT should be Interface friction angle, δpeak and δresidual x
performed. Monotonic data
At sites and turbine locations where highly variable soil con- Undrained shear strength, sC DSS
u , su , sE
u x x
ditions are expected, several CPTs should be conducted. Initial shear modulus, Gmax x x
In general, it is recommended to perform the CPTs outside the Cyclic data (triaxial and DSS)
actual caisson location, to avoid open holes which will poten-
Undr. shear strength, τf,cy = f (τa , τcy , N ) x x
tially affect the caisson installation and may even prevent the
caisson to reach the target penetration depth. Pore pressure, up = f (τa , τcy , N ) (x) x
up = f (τcy , log N ) for τa = τ0 , (x) x
• Sufficient boreholes at the site in order to extract samples of
Stress strain data, γa , γp , γcy = f (τa , τcy , N ) x x
all relevant soil units. Number and locations of the boreholes
should be selected based on the review and interpretation of γcy = f (τcy , log N ) for τa = τ0 x x
the geophysical and CPT data, preferable on basis of a ground Damping x x
model (e.g. Forsberg et al. 2017) Consolidation characteristics, intact soil
• Laboratory tests of all relevant soil layers within the CPT Preconsolidation stress (and OCR) x x
depth. Andersen et al. (2013) provide a comprehensive list Un- and reloading constrained moduli x x
of required parameters for various foundation concepts. A Permeability, k (x) x
summary of parameters for suction caissons is listed in Ta-
Remoulded soil data x
ble 1. The crosses in brackets indicate parameters, which are,
according to the author’s experience, somewhat less relevant. Sensitivity, St x
In order to determine the required parameters, drained and Undrained shear strength, sDSS
u x
undrained, monotonic and cyclic DSS, triaxial compression Cyclic undrained shear strength, τf,cy x
and triaxial extension tests need to be performed. Further, oe- Constrained modulus (x)
dometer tests, bender element tests, and interface tests should Permeability (x)
be included in the testing program. For layers with few
decimeter thickness, triaxial tests may be omitted. The num- Thixotropy (x)
ber of tests depends on the loading conditions, available data
from previous investigations at similar material, and the ap-
plied design methodologies. A representative set of labora- It is important to perform the tests at a stress and density
tory tests per soil layer may comprise or OCR, respectively, representative for the expected in-situ
conditions before and after installation. Three zones need to
– 2 oedometer tests, be distinguished; inside the caisson, outside the caisson, and
– 1 monotonic undrained DSS test and 1 monotonic below the caisson. While the soil state outside the caisson will
undrained triaxial compression test, as well as corre- be less affected by the installation, the soil at the inside may
sponding drained tests when testing sands, undergo considerable shearing, which will affect the density
and stresses. The soil below the caisson will be less affected
– 3-5 cyclic undrained DSS tests,
by the installation, but the weight of the OWT will yield an
– 4-6 cyclic undrained triaxial tests increase of the vertical effective stresses (with time).
In addition, other tests such monotonic as drained triaxial ex- In addition, index parameters such as relative density Dr ,
tension, or resonant column tests may be conducted where plasticity coefficient Ip , water content w, and grain size distri-
necessary. Of particular importance is the soil-skirt interface bution should be determined. These are in particular relevant
strength. It may be best represented by a remolded DSS test in an early stage of the project for the feasibility study and
consolidated to a stress equivalent to the lateral in-situ stress preliminary sizing, where not all laboratory tests have been
after installation. The stress level needs to be estimated. Rea- initiated yet, and where strength and stress-strain-behaviour
sonable stress ratios may be 0.5 and 1.0 times the vertical has to be assessed based on correlations using index data and
in-situ stress σv0 = γsoil
0
· z. Larger values may be less likely CPT soundings. Andersen (2015) proposes a comprehensive
due to set-up effects and arching, but may need to be decided set of correlations, which can be used as a first estimate of the
project specific. expected soil parameters.
48
In addition, information of scour development and/or scour pro- More complicated is the identification of the load cases which
tection is required. Type, thickness, submerged weight, and infor- should be used for the serviceability analysis. Two scenarios have
mation on the stability of the planned scour protection need to be to be distinguished; a maximum deflection and rotation during a
considered in the geotechnical analysis. severe load event, and accumulated average long-term deforma-
tion and rotation. The peak deflection may be assessed using the
loads used in the capacity analysis. For assessment of the long
2.2 Loads term deformations and rotations, cyclic loads are required. Ide-
ally, all loads during the lifetime of the OWT should be considered
The geotechnical designer needs to consider two different load
in chronological order. However, as this cannot be applied in a
sets. One set is required for the actual geotechnical design, i.e.
geotechnical analysis, simplified load histories are required.
capacity and serviceability analysis. The other set is used in the
load-stiffness iteration (outer loop in Figure 1). Some load cases It can be supposed that large cyclic load amplitudes will con-
may be included in both sets. But in general, the loads cases are tribute most to the accumulated deformations and rotation. Thus
different in both sets, since the governing design-loads and -criteria focusing on a series of storm events may be a reasonable simpli-
are typically different in the structural and the geotechnical design. fication. One option could be to use the 35-hrs design storm and
That means each discipline has to identify the relevant load cases, assuming a Gumble distribution to extrapolate the peak amplitudes
and need to define them such that everyone involved in the design of other storms with different return periods. The accumulated av-
process has a common understanding. Since this is a very critical erage displacements and rotations can be calculated for each scaled
aspect of a successful project, a load document should be prepared, 35-hrs design storm separately and then superimposed depending
which is continuously updated. This has been proven beneficial in on the expected number of occurrences of each storm during the
many projects. lifetime of the OWT.
The main challenge is to derive from the load-time-series the
Most design guidelines distinguish between loads for the Ul-
actual load amplitudes and corresponding mean values, and num-
timate Limit State (ULS), Serviceability Limit State (SLS), and
ber of occurrences, both of the maximum- and the cyclic-load
Fatigue Limit State (FLS)2 . ULS loads are required by both the
events. Most commonly the so-called rainflow-counting-algorithm
geotechnical and the structural designer. However, SLS loads are
is applied. Though this algorithm is widely used in structural fa-
mainly relevant for the geotechnical analysis, whereas FLS loads
tigue analysis, it is important to be aware of its limitations:
are mainly relevant in the structural analysis. All load cases are
assessed by the load or structural designer, and the geotechnical
• It is assumed that the loads are independent, meaning that the
designer need to provide input to these.
order of load cycles is not important.
Identifying or defining the required loads needs an experienced
designer. A reasonable starting point for the capacity analysis is to
• The information of the load frequency, that means the cyclic
look at the load cases comprising the maximum amplitudes; that
period, gets lost.
means maximum compression, tension, moment, etc. The maxi-
mum load amplitudes often adhere a load event which is embedded
• Since only the peak values are counted (that means actually
into a cyclic load history, which can be a storm for example. The
half-cycles are counted), no information can be directly de-
German Bundesamt für Seeschifffahrt und Hydrographie (BSH) in-
rived of the actual corresponding mean load.
troduced in the standard BSH (2015) a 35-hrs design storm based
on a composition of the Design Load Case (DLC) 6.1 proposed in
Depending on the soil type, drainage properties and boundary
the IEC standard IEC (2009). This cyclic event shall be applied to
conditions, these information can be crucial. Thus, if these infor-
assess the cyclically (degraded) soil strength, which is to be used
mation would need to be considered, other counting methods may
in the (subsequent) geotechnical analysis. Practically, this event
be applied where possible; for example the method proposed by
has also been also applied outside Germany, due to the lack of al-
Norén-Cosgriff et al. (2015). They apply high- and low-pass filters
ternatives, since the DLC’s defined in the IEC standard are 10 or
and determine the amplitude of each half-cycle from adjacent max-
60 minute long load-time series, which cannot be directly used in
ima and minima, which belong to the same load cycle. In addition,
a geotechnical design.
the proposed method keeps track of the corresponding average load
In more recent projects, where turbines with larger capacity and may also keep the information of the load period (frequency).
were considered, it has been found that also other events can be The authors compared their method with the rainflow-counting-
critical, such as an (emergency) shut-down at relative high wind- algorithm and showed that the calculated cyclically degraded soil
speeds. In the event of an (emergency) shut-down, the OWT strength using the example of a normally consolidated clay can be
swings and the load spectrum corresponds to a damped vibration. significantly different.
Depending on the degree of damping, which affects the decay rate, Cyclic load histories are often provided in from of a Markov
subsequent load cycles with smaller amplitudes can be critical due Matrix comprising cyclic load amplitudes and corresponding mean
to the cyclic degradation of the soil, induced by the previous larger load value as well as number of occurrences. Since these are of-
load cycles. Another event found critical for the foundation ca- ten established using the rainflow-counting-algorithm, it is recom-
pacity analysis of multi-legged structures is the prolonged tension mended that the geotechnical designer reviews also the original
load case, which typically occurs during operation of the turbine at load-time-series from which the Markov Matrix has been estab-
high wind speeds. lished. This in particular applies to the load-time-series compris-
In addition to the in-place loads, there may be further situations ing the maximum load values used in the geotechnical capacity
which needs to be considered in the design. These can be load analysis. The load cycle yielding the maximum load values may
cases during installation, maintenance, and decommissioning of sometimes appear to have a considerable offset from the rest of the
the OWT. cyclic loads history and it requires geotechnical judgment to de-
2
cide on the load cycle which the soil actually experience. But also
The author questions the appropriateness of the expression limit state in this con-
text. However, since it is widely used, it is – due to convenience reasons – also adopted a critical review of the mean load value is important, as the soil be-
in this contribution. haves essentially different symmetric and asymmetric cyclic loads.
49
It is recommended that permanent and environmental loads are Selecting appropriate safety factors for the design is difficult.
provided separately, and both as characteristic values, as occasion- Solely the DNV standard proposes a consistent safety concept for
ally, different partial safety factors need to be applied to the differ- capacity analysis considering the particular offshore conditions. In
ent load components in the geotechnical and structural analysis. general, the strength of the soil shall be reduced or carefully es-
timated for capacity and serviceability analysis. However, for the
installation analysis, a higher strength is more critical, which is not
2.3 Structural properties
considered in any standard. Sturm et al. (2015) proposes safety
As outlined in Figure 2, it may be important to include structural factors for installation analysis of suction caissons in sand, which
components in the geotechnical analysis. With increasing com- were established based on probabilistic analysis. Similar type of
plexity of the structural model, the stability and accuracy of nu- analysis may be performed for other design aspects. No safety
merical analysis may be quickly challenged. Thus, if structural factors should be applied in the serviceability-, stiffness-, and soil
models shall be included in a geotechnical analysis, they may be reaction-analysis as detailed in the corresponding sections.
simplified as appropriate. Beam and plate elements should be pre- Due to the lack of long-term experience, it is recommended
ferred over continuum elements. Structural components such as to consider a comprehensive monitoring system as part of the so-
stiffeners and stays may be omitted where possible. called observational method.
For capacity analysis, a rigid structure may be assumed, as the
the strength and stiffness of the soil at failure is several magnitudes 3 CYCLIC STRESS-STRAIN BEHAVIOR
smaller than the strength and stiffness of the structure, given that
the yield stress of the caissons material is not exceeded at any time. The loading condition of an OWT is of inherent cyclic nature.
For installation purposes, the properties of the skirts are of fun- Thus, all components including the soil, need to be designed ac-
damental importance and need to be considered in the penetration cordingly. The general supposition is, that cyclic loading yields a
analysis as accurate as possible. In general, the skirt tip resistance decrease of strength and stiffness, often denoted as cyclic degrada-
increases with increasing wall thickness. If stepped skirts are con- tion. This applies to all soil types and foundation concepts.
sidered, i.e. where the skirt wall thickness varies over the height, A number of authors have proposed methods for assessing the
the skirt friction may be affected considerably, which also will af- effect of cyclic loading on the suction caisson foundation response.
fect the in-place behavior. It is also important to consider compart- Therein two main approaches are followed; an empirical approach
ments3 and stiffeners in the penetration analysis if present. and an analytical/numerical approach.
• The empirical approach is typically based on model test

2.4 Guidelines and safety factors where the soil-foundation system is considered as one entity.
A dedicated standard or guideline for the design of suction caissons The caisson is subjected to cyclic loading and the response
for OWT applications does not exist. In the absence of such a doc- in the loading point is measured. The actual behavior of the
ument, other non-dedicated standards and guidelines need to be ap- structure and soil is not considered separately, hence it is a
plied in the design. This requires to define a code hierarchy, where phenomenological approach. The results can be presented in
in general national standards rank highest, followed by offshore interaction diagrams4 or failure envelopes in the HVM space,
wind related standards as well general offshore standards, and fi- where HVM is the horizontal, vertical, or moment load com-
nally other standards, guidelines and publications, which rank low- ponent, respectively. Failure envelopes allow a more detailed
est. Some examples are presented in the following. description of the foundation response compared to interac-
tion diagrams. In addition, a failure envelope diagram can
The IEC has proposed a series of documents addressing the par-
be extended to describe the actual load-displacement behav-
ticular design aspects of onshore and offshore wind turbines. For
ior by introducing a stack of HVM envelopes to which the
the load assessment and corresponding partially load factors, typi-
corresponding displacement components are assigned. Since
cally IEC standard 61400-3 is applied (IEC 2009). Other standards
these diagrams are based on interpolation of some few data
published by the IEC consider structural and geotechnical design
points, they are essentially empirical. Many, so-called macro-
aspects. However, these documents are so generally formulated,
elements, are based the empirical approach. Some Macro el-
with respect to geotechnical requirements – and in particular suc-
ements are mathematical complex and can describe very de-
tion caisson design – that other standards need to considered.
tailed the load-deformation behavior of a caisson subjected
To the author’s knowledge, all countries where OWTs are
to general cyclic loading. A number of authors have devel-
considered, have own national standards for geotechnical design.
oped macro-elements for suction caissons, (e.g. Nguyen-Sy
However, since these standards originate form onshore design re-
2005, Nguyen-Sy & Houlsby 2005, Salciarini & Tamagnini
quirements, the application of the recommended methods and pro-
2009, Salciarini et al. 2011, Foglia et al. 2014, Skau et al.
cedures to offshore structures can be critical. Thus, some countries
2017). Macro-elements are well suited in integrated analysis
are in the process of establishing national standards particularly for
for structural design and load assessment.
OWTs. This has been done by the German BSH for example. The
US Bureau of Ocean Energy Management (BOEM) and the Ger- • In the analytical/numerical approach the response of the
man DIN are also working on corresponding documents. soil-foundation system is assessed by modeling the actual
As most OWTs need to be certified due to financial and insur- soil-structure interaction under consideration of the structural
ance reasons, some certifiers have published their own guidelines, flexibility and stress-strain-behavior of the soil. This requires
which are frequently used in the design. Most relevant is the DNV a detailed description of the skirt-soil- and lid-soil-interface
GL standard 0126 (DNV-GL 2016). This document provides valu- behavior. In an analytical approach, the distribution of aver-
able recommendations and includes also a section on suction cais- age and cyclic loads – or actually stresses – along the skirts
sons. However, it is very generally formulated and neither particu- need to be assumed, whereas the distribution is automatically
lar methods nor procedures are proposed. calculated in a numerical approach. The assessment of the
3 4
Compartment mean that the caisson lid area is divided into different cells Similar to diagrams used for cyclic axially loaded piles
50
cyclic stress-strain behavior and strength of the soil needs to

be described by using appropriate soil models. The analyt-
ical/numerical approach is well suited for the geotechnical
sizing of the caisson, but may also be used for assessment
for the serviceability and calibrating of the input parameters
to a macro-element.
NGI has developed a method for describing the behavior of

cyclically loaded soil elements using so-called cyclic contour di-
agrams. The method, originally proposed in the early 70th , which
was continuously developed further, has been presented in a nu-
merous publications; the most recent and comprehensive one is
the article by Andersen (2015). Cyclic contour diagrams span a
3-dimensional space and provide a general relation between aver-
age and cyclic shear stresses and corresponding average and cyclic
Figure 3: Finite element analysis of a suction caisson subjected to com-
shear strains as function of number of applied cycles. Diagrams bined vertical and horizontal cyclic loading using the NGI soil model PD-
are established for one soil type and density or OCR, respectively. CAM. The contour plot shows the excess pore pressure at the end of the
One complete set of 3d-diagrams for one soil unit comprises typi- peak phase during a 35-hrs design storm.
cally of 4 diagrams; 1 strain and 1 pore pressure diagrams for tri-
axial and DSS conditions, respectively. In many practical applica-
tion cases, only some representative 2-dimensional cross-sections 4 FOUNDATION CAPACITY
of the 3-dimensional space are required. This simplifies the ap-
The foundation capacity needs to be ensured for all possible load
proach and reduces the number of cyclically laboratory tests. The
combinations. Two main load scenarios should be distinguished,
selection of appropriate cross-sections requires some experience
which are detailed in the following.
and assumptions.
In combination with a cyclic load history, the cyclic contour di-
agrams can be used in the so-called cyclic accumulation procedure. 4.1 Short-term loading
The cyclic degradation due to the cyclic loading is calculated and Short-term loading is characterized by a loading duration being
the effect can be expressed by the so-called Equivalent number of so short that the soil behaves essentially undrained, meaning that
cycles (Neq ). the soil response depends on the undrained shear strength only. In
As cyclic contour diagrams provide a relationship between sandy soils, the caisson may mobilize considerable suction below
stresses and strains, but the cyclic loads are given as forces, as- the lid and negative pore pressure in the soil, causing an increase in
sumptions on the load transfer and stress distribution has to be mean stresses and hence higher shear strength. Due to the shallow
made, which is best done using the Finite Element Method (FEM). water depth at typically OWT sites, particular attention requires
This is in particular the case where complicated boundary condi- the cavitation limit. The cavitation limit cannot be exceeded by the
tions, soil layering and drainage conditions are analyzed, which is suction or negative pore pressure, respectively. That is in partic-
in general the case for suction caissons for OWTs. NGI has im- ular important to consider when deriving the shear strength from
plemented the cyclic accumulation procedure using cyclic contour laboratory tests where considerable back-pressures may have been
diagrams in an FE code. Jostad et al. (2014) present the procedure applied, as these tests can potentially exceed the maximum achiev-
for fully undrained conditions during the considered cyclic load able pore pressure and hence strength compared to the actual in-
history (UDCAM)5 , whereas the procedure for partially drained situ conditions. The theoretical cavitation limit pcav,max in a soil
conditions (PDCAM)6 is presented by Jostad et al. (2015b). The element is the sum of, the depth z of that element below mudline
cyclic accumulation is done for each integration point. The ad- plus the water depth ws , multiplied with the unit weight of water
kN
vantage of using the FEM is, that the stress redistribution is con- γw = 10 m 3 , and the atmospheric pressure patm = 100kPa, viz.
sidered accurately and continuously updated if relevant, and that
strain continuity is ensured. Furthermore, a output of such an 0
analysis is not only the cyclic stress-strain behavior and degraded pcav,max = (z + ws ) · γw + patm (1)
strength and stiffness, but also the accumulated displacements and
At NGI, the short-term capacity analysis is often done using a
rotations, which are required for the serviceability analysis.
total stress approach. Figure 4 shows a potential failure mecha-
Though the soil-structure interaction is modeled in detail (nu-
nism of a suction caisson under combined compression and mo-
merical approach), the description of the soil behavior using cyclic
ment loading. The undrained strength in the failure zone is de-
contour diagrams is an empirical approach.
scribed by the strength measured in undrained DSS tests, or in a
An example of a PDCAM analysis of a suction caisson sub- trixial tests where different Total Stress Paths (TSP) are followed.
jected to a combination of vertical and horizontal cyclic loading is Cyclic contour diagrams can be used for assessing corresponding
shown in Figure 3. A suction caisson with 8m diameter and 6m cyclic shear strength values.
skirt length in a homogeneous soil deposit with an average soil Figure 5 illustrates the four main different total – and corre-
permeability of k = 1 · 10−5 ms is modeled. At the peak phase of sponding effective – stress paths, using the example of a medium
an 35-hrs design storm according to BSH (2015), the soil at skirt dense to dense sand specimen consolidated to a stress state of
tip level accumulates considerable excess pore pressure. Due to σ0
k = σh0 = 0.5 at a vertical effective stress of σv0 = 200kPa.
the symmetric soil and load conditions the predicted pore pressure v
field is also almost completely symmetric. The difference between the TSPs is the way the shear strength has
been applied. For path 1 and 6 the cell pressure in a triaxial test has
5
UnDrained Cyclic Accumulation Model been decreased or increased, respectively, whereas for path 4 and 2
6
Partially Drained Cyclic Accumulation Model the vertical pressure has been increased or decreased, respectively.
51
M 250
Norm. loads Fx/Fx,ref, Fz/Fz,ref, My/My,ref [%]

V
H Grout Fx
0.0 m Filter layer (optional) My
Mt
TSP 1 TSP 6 200 Fz
150
TSP 2
z
TSP 4 100
DSS
DSS
DSS
50
Figure 4: Possible failure mode of a caissons subjected to combined com-

pression and moment loading 0
0.001 0.01 0.1 1 10 100 1000
Norm. rot. stiff. kϕϕ/kϕϕ,ref [-]
In addition, the total and effective stress path in direction 4 for a
specimen consolidated to σv0 = 20kPa is shown. Figure 6: ULS loads as function of the rotational stiffness of caisson sup-
porting a three-legged jacket.
400 TSP4
ESP4
TSP1
300 ESP1 stiffness components.
TSP6
ESP6
200 TSP2
ESP2
CSL 200
100 CSL
q [kPa]
TSP4 LS
ESP4 LS
0
Norm. FoS [%]
-100 150
-200
-300
0 50 100 150 200 250
100
p’ [kPa]
Figure 5: Total and Effective stress path in trixial tests where the shear
stress is applied in different ways.
50
0.001 0.01 0.1 1 10 100 1000
From Figure 5 becomes apparent that the soil strength of a sand
Norm. rot. stiff. kϕϕ/kϕϕ,ref [-]
specimen for a given initial density and stress state is depending
on the loading path. The difference between the total and effec-
tive stress for the different paths equates the corresponding pore Figure 7: Normalised Factor of Safety (FoS) as function of the applied
pressure. The maximum negative pore pressure cannot exceed the loads shown in Figure 6
cavitation limit. Whether the NGI method or any other method is
applied, it is important that the dependency of the stress path and The effect of the load combination shown in Figure 6 on the
the cavitation limit is considered accurately when assessing the soil caisson capacity is shown in Figure 7, where the normalized Fac-
strength profile. tor of Safety (FoS) is plotted on the abscissa. As it may be expected
The stress path dependency is equally relevant for clay speci- from conventional bearing capacity analysis, the normalized FoS is
mens. In additon, due to the viscosity of clays, the dependency lower for larger moments, that means for a rotational failure mode.
of the shear strength on the shear rate needs to be considered. That applies also to a mono-caisson foundation, which is essen-
The shear rate in laboratory tests may be different compared to in- tially subjected to environmental horizontal and moment loading
situ loading rate for short-term loading, meaning the shear strength only.
may need to be corrected accordingly. In offshore foundation design of multi-legged jacket structures,
The capacity of suction caissons to short-term loading is es- it is often assumed that the rotational stiffness of a foundation at
sentially governed by the load combination, that means horizon- ULS loading is considerably lower than the rotational stiffness of
tal, vertical and moment loading. As illustrated in Figure 1, the the corresponding leg of the sub-structure. Hence, the local mo-
design basis, including the loads, is continuously updated. Fig- ment loading at failure may be omitted in the capacity analysis.
ure 6 shows the dependency of the ULS loads on the rotational However, the relatively high jackets stiffness can be an issue for
stiffness of a suction caisson at the example of a multi-legged sub- the fatigue design of an OWT, as the goal is, that the first eigen-
structure. The loads of a leg in compressions, are normalized with mode shall be in the range between 1P and 3P; e.g. typically be-
the reference loads provided in the 1st iteration. The predicted tween 0.25 and 0.35 Hz for turbines with 6 to 8MW. Thus, the
corresponding rotational stiffness – also normalized – is shown at structural designer tries to make the jacket more flexible, meaning,
the abscissa where all load components are crossing. Though the that omitting the local moment may be too optimistic.
global loads acting on the OWT are constant, the local loads can To complicate matters, the local load components at a leg of a
vary considerably depending on the response of the caisson. The jacket do not scale proportionally with the global load amplitude,
higher the rotational stiffness, the lower the vertical and torsional even though the global loads may be applied linearly increasing.
loads. Similar effects, but less pronounced is found for the other Thus, the ULS load components provided in the design basis may
52
not be scaled proportionally with a load factor. However, as the to the upward flow reducing the effective vertical stresses – and a
soil will be always softer than the jacket leg in rotation when being small suction pressure below the lid, which is required to maintain
at failure, overestimating the local moment may yield lower FoS as a constant flow. The friction capacity needs to be further reduced
shown in Figure 7. Nevertheless, it is recommended to check the to account for the relative vertical movement of the caisson, which
FoS for differently scaled local loads, that means lower load factor reduces the vertical stresses in the soil and hence the shear stresses
applied to the local moment and a larger load factors to the vertical, in the soil-skirt-interface.
horizontal and torsional load components. For mono-caissons, a The difficulty is to decide upon the load and resistance factors
redistribution of the local loads is not expected and the same load which shall be applied. If a load case can potentially cause a failure
factor should be applied to all load components. of the structure, the full load and resistance factors according to the
For suction caissons subjected to tension loading, the same con- considered standard should be applied. However, if the loads for a
siderations discussed above apply. The TSP strength used in the considered load case can be controlled, for example by the turbine
analysis need to account for the different loading and hence stress operation, the load factors may be reduced somewhat to acknowl-
conditions. edge for the reduced uncertainty in the actual load amplitude. But
Gapping at the outside of the caisson may need to be consid- also the failure mechanism may justify to apply somewhat lower
ered in the capacity analysis, if previous load conditions or stepped safety factors. In case of a suction caisson in sand subjected to
skirts may have generated a gap. Due to the short-term loading, the long-term tension loading, the structure may not experience a sud-
drainage time may not be sufficient to generate a new gap during den failure, but may be pulled out gradually. If reduced load and
the considered load event. This depends of course on the load com- resistance factors are applied, the serviceability needs to be en-
bination and soil type and may need to be checked. sured at any time, and an appropriate monitoring system should be
Of particular importance is the scour development and scour installed, in order to apply the observational method. In addition,
protection. The stress and density state of the soil can be consid- mitigation measures need to be prepared.
erably affected, which can have an impact on the foundation ca- As the loading conditions of OWTs is of essentially cyclic na-
pacity. Whether to include or omit the effect of a scour and scour ture, also the long-term tension loading is actually a cyclic load
protection should be discussed with the operator, as the presump- case. Thus, an appropriate cyclically degraded shear strength pro-
tion of a permanent scour protection may require more frequent file and corresponding stress-strain response need to be used. For
on-site inspections, which can have an impact on the Operational that purpose assumptions need to be made on the distribution of
and Maintenance (O&M) costs. the average long-term tension load and the cyclic amplitude. De-
pending on the considered load case, it may be assumed that the
4.2 Long-term loading skirt-soil-interface at the outside of a caisson in clay may take the
cyclic component and the soil below and inside the caisson may
Suction caissons have considerable capacity under short-term load-
take the average component. Where this distinction should not be
ing conditions. However, the resistance to long-term loading, can
possible, an equally degraded strength profile may need to be as-
be very low, as the possibly mobilized suction may dissipate. This
sumed.
is in particular relevant for suction caissons supporting a jacket
As the cyclic load components have relatively short period, the
structure. During operational load cases the caisson(s) may expe-
soil response of a caisson in sand will be essentially undrained
rience considerable tension loading, which can last for hours or
to this component only. Thus, for a caisson in sand, the capac-
even days. The tension capacity of suction caissons is a function
ity needs to be checked for at least two cases; the resistance to
of the skirt wall friction and the soil permeability.
the average tension load, and the resistance to combined cyclic
For caissons in clay the soil permeability will be low, meaning
and average load using an appropriate cyclic shear strength pro-
that the capacity can be calculated similar to the long-term capac-
file. When using the NGI framework based on cyclic contour dia-
ity, but the shear strength needs to be reduced to account for the
grams, the strength and stress-strain response can be derived from
slow loading rate. In the absence of suitable tests, the decrease in
diagrams where the average shear stress was applied drained in the
shear strength may be estimated using
corresponding laboratory test. Further information can be found in
Andersen (2015).
Iv
γ̇slow The same considerations made for the short-term bearing ca-
su,slow = su,ref · (2)
γ̇ref pacity analysis on whether to include or to omit the effect of scour
or scour protection, applies to the long-term bearing capacity anal-
where the su,ref is the shear strength measured in the laboratory ysis as well.
at a shear rate of γ̇ref . γ̇slow is the shear rate representative for the
considered load case. Iv is a viscosity coefficient which typically
varies between 0.03 and 0.07 for a silty or fat clay, respectively 5 INSTALLATION
(Leinenkugel 1976). Iv can be determined with Equation 2 from
an undrained static laboratory test, where the shear rate is varied. The installation is considered by many as one of the most chal-
If previous load cases, structural boundary condition or any lenging aspects of suction caisson application. However, experi-
other causes may have generated channels or gaps at the outside ence from actual installations has demonstrated that installation
and inside of the caisson in the clay, only the skirt wall friction can in many different soil types and profiles is feasible. Moreover,
be considered in the tension capacity analysis. the predicted penetration resistance and hence the required suction
For caissons in sand, the soil permeability is considerably pressure agrees often reasonably well with the actual measured
higher, meaning that a continuous flow of water from the outside values (e.g. Sparrevik 2002, Colliat et al. 2007, Aas et al. 2009,
to the inside can be expected, given that the tension load exceeds Langford et al. 2012, Solhjell et al. 2014, Saue et al. 2017).
the resistance calculated by integrating the fully drained skirt wall The governing mechanisms are well understood and several au-
friction over the skirt area at inside and outside of the caisson. In thors have developed calculation methods. Most methods can be
this case, the capacity is the sum of the drained skirt wall fric- applied in uniform and homogeneous soil conditions or soil pro-
tion at the outside, a reduced drained friction at the inside – due files with perfectly horizontal layering. A general discussion of the
53
installation process and calculation methods is presented in Sub- (Driving loads) W´

section 5.1.
t
All existing calculation procedures have limitation, and there
are a number of aspects which need particular attention during the Ureq
0.0 m
actual installation, since they cannot be considered by the existing
calculation models. Some of the most relevant aspects are pre-
sented in Subsection 5.2. Possible mitigation measures are dis- z∙γ´soil z∙γ´soil
cussed in Subsection 5.3. z
z∙γw∙i(z) z∙γw∙i(z) Qwall,out

5.1 Calculation methods
The often reasonably accurate predictions of the penetration resis- Qwall,in

tance and hence required suction pressures is a result of extensive
research in this field. A number of authors have proposed meth- σ´h,out σ´h,in
ods for calculating the penetration resistance and required suction Qtip
(In-situ stresses) (Resistance)
pressure in both clay, silt and sand layers; particularly noteworthy
are the models proposed by Houlsby & Byrne (2005a,b), Andersen Figure 8: Forces and stresses acting on a caisson during suction installation
et al. (2008) and Senders & Randolph (2009). These are based on in a high permeable soil; from Sturm et al. (2015)
model tests, field tests and prototype installations.
The penetration resistance is a function of the skirt tip resis-
tance Qtip and the skirt wall friction Qwall . Qtip may be esti- • the submersion depth of the pump, given that the pump sits
mated using a bearing capacity based approach or correlations with on top of the caisson lid, or
measured CPT resistances. Qwall is a function of the skirt-soil-
• the mudline depth, given that a closed system is established,
interface strength τfric and the effective skirt wall area. τfric can
where one hose is connecting the caisson with the pump and
be assessed by means of laboratory tests, such as DSS tests or ring
another hose returns the water from the outlet of the pump
shear tests. Alternatively, τfric can be estimated using correlations
back to the mudline.
with measured CPT resistances.
If the total penetration resistance Q = Qtip +Qwall exceeds the Though the pressure is theoretically higher for the latter case, it
submerged weight of the caisson and sub-structure W 0 = Wcais. 0
+ is technically more challenging. Furthermore, a considerable head
0
Wsubstr. , an additional driving force needs to be applied in order loss can be expected due to the length of the hoses, which reduces
to penetrate the caisson to the required Target Penetration Depth the efficiency of the second solution.
(TPD). This is done by applying a relative under- / suction-pressure The actual maximum achievable pressure p0suc,cav (z) is practi-
psuc at in the inside of the caisson. The additional driving force cally somewhat less than the calculated value psuc,cav (z), since the
is calculated by integrating the applied suction pressure over the pump may not be able to go as low as to the theoretical pressure.
horizontally projected area Asuc to which the pressure is applied. Thus, a reduction of 20 to 50kPa of psuc,cav (z) may be considered
The maximum achievable penetration depth is reached when the in the design, where the reduction should be adjusted based on the
total resistance Q exceeds the driving forces W 0 + psuc · Asuc . pump specifications.
Two main scenarios need to be distinguished; an undrained The actual allowable suction pressure psuc,all (z) ≤ p0suc,cav (z)
penetration and a drained penetration. A penetration is undrained may be limited by geotechnical and structural stability considera-
if the soil permeability k of the penetrated layer is so low, that no tions. The skirt needs to take the load without to buckle. In the
significant amounts of pore pressure will dissipated during the ac- initial phase when applying the first time a suction pressure right
tual installation process. In contrast to an undrained penetration after the self-weight penetration phase, the caisson is exposed to
is the pore pressure dissipation considerably in a drained penetra- buckling failure due to the lack of any soil support above mudline.
tion, which will affect the the stress regime in the soil. Due to the This is in particular critical for penetration in stiff clays at shallow
applied suction pressure, a seepage flow through the soil from the depths. But also in the course of further penetration when the re-
outside to the inside will develop in a high permeable soil layer. quired suction pressure p0suc,req (z) increases with depth, the cais-
The upward flow in the soil plug inside the caisson causes a de- son may be exposed to buckling failure, if the inside soil support
crease of the vertical effective stresses σv0 and hence a decrease of is low. This is typically the case for penetration in high permeable
the inside side friction τfric . Furthermore, also the tip resistance soils due to the upward flow of pore water in the soil plug reducing
will decrease due to the potentially high gradient around the skirt the stresses and hence strength.
tip. Both yield a considerable reduction of the penetration resis- Geotechnical limitations which can potentially affect
tance, meaning that a suction pressure has a twofold effect in a psuc,all (z) are reverse bearing failure, primarily when pene-
drained penetration; it increases the driving force and reduces the trating in low permeable soils, and hydraulic heave failure,
resistance in high permeable soils. Figure 8 illustrates the driving primarily when penetrating in high permeable soils. Some authors
forces (top), stresses in the soil (left bottom) and resulting reaction have included in their calculation models criteria and functions to
forces (right bottom) acting on a suction caisson during installation ensure that these failures are avoided.
in a high permeable soil. Somewhat more complicated is the penetration in layered soil
The maximum possible suction pressure psuc,cav (z), which can profiles. Two scenarios need to be distinguished; sand over clay
be applied inside the caisson, is limited by the cavitation pres- and clay over sand, where sand is a high permeable layer and clay
sure. As detailed in Section 4, the cavitation pressure depends a low permeable layer. Sand over clay is a common profile in many
on the pump configuration, and is given by the sum of the atmo- areas of the North and Baltic Sea, and the penetration through these
spheric pressure patm = 100kP a and the unit weight of water do not pose a particular challenge. However, clay over sand is sub-
0 kN
γw = 10 m 3 times the depth of either ject of ongoing discussion. Some authors have found in centrifuge
54
a) Soil plug lift b) Piping c) Excessive loosening d) Stucking
Loosening
Low permeable Flow
Void/Gap layer channel
ki ko
High permeable
dense layer
e) Uneven seabed or heave f) Inner erosion g) Tilt h) Boulders
kn
kn
ko
i) Insufficient SWP depth j) Sliding k) Inclined layers or lenses l) Lost of soil support (buckling)
Figure 9: Some possible failure during installation, which cannot be predicted or insufficiently predicted with the available installation analysis models.
tests and/or small scale model tests, that penetration in the under- necessary. However, to generate a gradient in the sand layer cov-
laying sand layer may not be possible without triggering a plug-lift ered by the clay, a seepage flow must have been developed. As the
failure (e.g. Cotter 2009). They recommend to stop the penetration water cannot flow out through the soil plug in the caisson, the sand
above the sand layer, where the maximum allowable penetration layer below the clay layer needs to take the water volume, meaning
depth into the clay is given by the shear strength of that layer be- that the sand will reduce its density. Thus despite the fact, that the
low the skirt tip and the caisson geometry. However, installations trial installations demonstrated that a penetration in layered soils
of suction caissons in such layered soil profiles have demonstrated, is possible, it is recommended to penetrate relatively fast to avoid
that a penetration is in principal possible without a measurable soil excessive loosening (soil plug heave) or eventually a soil plug lift.
plug-lift. In installations, where pore pressure sensors were placed
at the in- and outside of the skirt walls above tip, it was found that
5.2 Challenges
the pressure gradient in the sand layer around the skirt tip, equates
the gradient measured in installations in homogeneous clean sand The methods mentioned in Subsection 5.1 are applicable for ide-
deposits. That supports the assumption that a plug lift failure is not alized conditions, i.e. uniform and homogeneous soil conditions
55
or perfectly horizontal layering, vertical and parallel skirts, and no increases. Installations with single caissons and anchors showed
structural imperfections, to name but a few. However, there are that a caisson is a self-stabilizing system, meaning that it recti-
a number of situations which are not covered. Some of the most fies due to the lateral soil resistance. However, if the caisson is
common ones are illustrated in Figure 9. constrained – for example when attached to a jacket – the loads
Soil plug lift is a failure often discussed in connection with pen- can become critical for the sub-structure. Thus it is important to
etration in layered soils. In contrast to soil plug heave, soil plug lift ensure a minimum degree of verticality of all caisson of a multi-
will generate a water filled void or gap in the ground. That needs legged sub-structure during the fabrication.
to be avoided in order to not negatively affect the in-place behavior Boulders and other large obstacles can prevent the caisson to
of the suction caisson. Furthermore soil plug lift may prevent the reach the TPD as the penetration resistance will increase consider-
caisson from penetrating to the TPD as the caisson will be filled ably. If not identified in due time by the pump operator, the cais-
up with soil. Practical experience form installations in layered soil son skirts may be damaged or buckled. Small boulders may flip
profiles suggest to apply a minimum penetration rate in order to or pushed to the inside due to the suction pressure. Boulders can
reduce the amount of water flowing into the soil plug and potential be detected by means of suitable geophysical site investigations. If
void. boulders are met, the caisson may be retrieved and relocated, given
Piping is a critical failure, as the volume of water per time flow- that the structure has not been damaged.
ing from the outside to the inside will increase considerably. If the If the submerged weight of the caisson and substructure is too
water volume exceeds a certain amount, the pump may not be able low, the self-weight penetration may not be sufficient to ensure a
to apply the required suction pressure and the TPD may not be seal at skirt tip level, which is necessary to apply a suction pressure.
reached. Furthermore, piping channels generated during installa- Sliding during the lowering and touch-down phase of the cais-
tion can negatively affect the in-place performance, as the tempo- son may remove soil in the vicinity of the skirt tip, preventing suf-
rary suction during short-term loading will dissipate much faster ficient seal, which is necessary to apply a suction pressure. Hence,
which can potentially decreases the capacity significantly. Piping allowable sea states for the installation should be assessed in the
can be triggered by obstacles below the skirt tip which are dragged design.
down while penetrating the caisson. These obstacles can leave a Particularly challenging is the penetration of profiles compris-
highly disturbed zone along the skirt wall. But also locally varying inclined layers and lenses. In case of an inclined clay layer or
ing soil properties in combination with penetration at high suction lens below or in a sand layer, respectively, the pore pressure gradi-
pressures and hence penetration rate can trigger the generation of ent at skirt tip level may become critically high, since the changed
piping. drainage conditions will affect the seepage flow pattern. That can
Excessive loosening may occur in installation in permeable potentially trigger a local failure or piping along the skirt at the
soils. Due to the reduced vertical stresses and additional shear- side of the caissons which is still in the sand. In case of an in-
ing of the material inside the caisson, the soil will dilate. That clined sand layer or lens below or in a clay layer, respectively, the
will affect the soil permeability and hence the seepage flow pat- penetration resistance may considerably increase since a seepage
tern, which can prevent the caisson to reach the TPD, since the flow, as described for perfectly horizontally layered profiles, may
required flow gradient in the soil cannot be achieved. Experience not be established. Furthermore, the soil resistance will be asym-
from installations in homogeneous sand deposits indicate that the metric and potentially causing a tilt of the caisson or local moment
degree of loosening correlates positively with the installation time, in the leg of the sub-structure, respectively. However, the deeper
meaning that penetration at higher rate may potentially avoid ex- the caisson has penetrated the more soil support at the outside of
cessive loosening. Sturm et al. (2015) proposes safety factors for the caisson is available, which can compensate for the asymmetric
the penetration analysis capturing the uncertainty of an excessive penetration resistance.
loosening. Imperfections or buckling at skirt tip level can increase the
Embedded and thin granular but relatively low permeable soil penetration resistance considerably and also affect negatively the
layers and lenses may cause the caisson to stuck, if the required in-place behavior of the suction caisson. Thus the allowable suc-
suction pressure exceeds an allowable value and if no seepage flow tion pressure should not be exceeded and a maximum tolerance for
can be mobilized in that layer, which would reduce the tip resis- imperfections and misalignments shall be considered in the fabri-
tance considerably. cation.
An uneven mudline may prevent the caisson to reach the TPD,
if not considered in the design of the so-called free height, which
5.3 Mitigation measures
is the skirt length in addition to the calculated required penetra-
tion depth. The free height is typically measured from the original In case that the penetration resistance is higher than predicted, the
mudline and need to accommodate the soil plug heave, grout, and required suction pressure to penetrate the caisson will be higher as
pre-installed filter material if applied, and seabed elevation. An well. Where it is not possible to apply the required suction pres-
uneven mudline can be also critical for the self-weight penetration sure due to geotechnical, structural or technical limitations, one
phase, if the penetration resistance is locally too high preventing may consider to abort the penetration or apply mitigation measure
the whole caisson circumferences to penetrate and to establish a in order to try to penetrate further until reaching the TPD. The deci-
sealing, which is required to apply a suction pressure. sion should depend on the achieved penetration depth as well as on
Soil layers with a gap graded grain size distribution curve, the course of the penetration process. If for example, the caisson
where the large diameter grains can form a stable matrix, are sensi- has penetrated 80 or 90% of the TPD and the penetration resistance
tive to inner erosion. Fine grained particles are washed out of the had been continuously higher than predicted in the design, it may
soil due to the applied suction, and a very high permeable grain indicate that the foundation has already sufficient capacity for the
skeleton remains in the ground. Since the amount of water volume actual reached penetration depth. More challenging is the impact
flowing into the caisson per time increases, the pump may not be of the stiffness for a lower penetration depth.Sturm & Mirdamadi
able to apply the required suction pressure, meaning that the TPD (2017) propose a reliability based method for assessing foundation
cannot be reached. stiffness, which can be used during installation, on which basis a
Tilt of the caisson can be critical, as the penetration resistance decision can be made if the caisson(s) need to be penetrated further
56
Figure 10: Pressure gradient in the soil for a clean sand profile (left), a sand profile with a clay layer a skirt tip (center), and sand profile with a clay layer
a skirt tip with a stepped skirt (right)
by means of applying mitigation measure. have a negative affect on the resistance in normal and low consol-
Two categories of mitigation measures need to be distinguished; idated clays, as the soil may partially drain and by that increases
preemptive and reactive mitigation measures. Preemptive methods its strength. The soil sensitivity may provide an indication of the
are those which have been considered before the actual installation. expected efficiency of cycling.
Reactive methods are applied during the actual installation and do
The effectiveness of cycling depends further on the cyclic dis-
not require any particular structural considerations.
placement amplitude which is also indicated by the results pre-
A simple but often effective reactive mitigation is to ballast the
sented by (Cudmani & Sturm 2006). The larger the displacement
structure to increase its weight. This can help in many situations
amplitude the more effective the cycling, which can be explained
discussed in Subsection 5.2, for example in case of piping, inner
by an increased shearing of the soil. In addition, the cycling rate
erosion, stucking, and insufficient self-weight penetration.
may have an effect on the effectiveness as it allows the soil to drain
Another reactive mitigation measure is to cycle the suction
somewhat.
pressure, which is illustrated in Figure 11. Cycling has been ap-
plied in many installations to successfully penetrate to the TPD. A preemptive mitigation measure is a stepped skirt. A stepped
skirt has different wall thicknesses over the height. Similar to driv-
4.0 ing shoes used for piles, a stepped skirt, as considered herein, will
be thicker at the tip compared to the rest of the skirt wall. This will
generate a thin gap or disturbed zone along the skirt, which needs
4.2
Penetration depth [m]
to be at the inside of the caisson, in order to be effective for the

Reduction at
equal depth penetration. The stepped skirt functions as a friction breaker.
4.4
Figure 10 shows the required pore pressure field to penetrate
equal pressure
a caisson in a clean sand deposit (left figure). When penetrating

Reduction at
4.6
through a thin clay layer embedded in the sand, the seepage flow
is prevented and the required gradient around the skirt tip cannot
be achieved (center figure). The caisson cannot be penetrated fur-
4.8 ther. However, when using a friction breaker, a gap or disturbed
150 100 50 0 -50 -100 -150 -200 -250 -300 zone along the skirts, and in particular in the thin clay layer, may
Differential pressure [kPa] be generated, which allows to establish a seepage flow from the
outside to the inside. Due to the different seepage flow pattern, the
Figure 11: Suction pressure versus vertical displacement during cycling as
a reactive mitigation measure actual required suction pressure to achieve the same pressure gra-
dient at skirt tip is less than the required suction pressure in a clean
sand profile (right figure). This indicates that a friction breaker
Due to the cycling of suction pressure, the caisson will move
can be a very effective mitigation. However, due to the disturbed
somewhat up and down, which will affect the soil in the vicinity
zone, the in-place performance of the caisson may be negatively af-
of the skirts. Over-consolidated clays will be remolded due to the
fected, since the suction generated during short-term loading will
cycling and the shear strength will decrease. This will mainly af-
dissipate faster. And also the resistance to long term loading may
fect the skirt wall friction Qwall . Cycling when penetrating in sand
be reduced compared to a caisson with constant wall thickness.
layers can be beneficial as well, as the soil below the skirt tip will
dilate due to the unloading, which will decrease the tip resistance Another preemptive mitigation measure is the water injection
(see Cudmani & Sturm 2006). The effectiveness of cycling can system, where at a pipe with nozzles through which water can be
be described by considering the reduced suction pressure at equal injected into the soil is arranged at the skirt tip. Purpose of the wa-
penetration depth or the achieved additional penetration depth at ter injection system is to reduce the penetration resistance. This is
equal suction pressure; both illustrated in Figure 11. achieved in sand by a loosening the soil at skirt tip, and in clay by
A systematic evaluation of NGI in-house installation data, remolding the soil along the skirts. Injection of water appears to be
where the suction pressure was cycled, showed that both measures most effective in combination with cycling, where the amount of
are equivalent, though more practical relevance has the increase injected water is adjusted to the void generated by the skirt when
in depth at equal pressure. Further, a general tendency can be moving upwards. This will have a minimal effect on the soil state
observed that the effectiveness of cycling increases with increas- after installation. Water should be injected in any case at low pres-
ing penetration resistance. This may be expected as the decrease sure to avoid excessive soil disturbance, which can potentially neg-
in strength due to remolding is higher for over-consolidated ma- atively affect the in-place behavior of the caisson. Aas et al. (2009)
terial than for normal consolidated material. In fact cycling may reports results of a water injection system used in layered profiles.
57
6 FOUNDATION SERVICEABILITY displacements for a given cyclic history, but at different scaling
factors. The accumulated total displacements can be than deter-
The foundation serviceability is probably one of the most impre- mined by summing up the the calculated displacements for the dif-
cisely predictable aspects in geotechnical engineering. Service- ferent storm events multiplied with the number of occurrences of
ability in this context means settlements, lateral displacements, and the corresponding event.
rotation or tilt, respectively. Most critical is the tilt of an OWT as it However, both approaches miss out important aspects. Differ-
affects the operation of the turbine. Pure settlements are typically ent to engineering materials such as steel or concrete, soils are sen-
less critical, though some secondary steel components such as the sitive to the order of cyclic loading. While large cyclic load ampli-
J-tube or the boat lander may be affected. The lateral displace- tudes can cause a degradation of the soil strength and stiffness, can
ments are typically small, and have practically no relevance in the the soil regain strength and stiffness when subjected to lower cyclic
projects considered so far. Thus main focus is given in the follow- load amplitudes, which can be described as self-healing. The influ-
ing on differential settlements or tilt of multi-legged substructures ence of varying strength and stiffness of the soil on the settlement
or mono-caissons, respectively. and tilt depending on the cyclic loading conditions is described in
In order to assess the Serviceability Limit State (SLS), corre- Sturm (2009) and Sturm (2011) at the example of skirted shallow
sponding limit values need to be defined. These are typically given foundations. It is introduced the concept of the so-called cyclic
by the turbine supplier. In addition, the maximum tilt may be lim- attractor, which is a value being asymptotically approached by a
ited in order to reduce operational loads, which is in particular rel- given cyclic load history with constant amplitude. Given that the
evant for multi-legged OWTs; increased average tilt yield typically foundation is stable for all relevant cyclic load histories, the value
an increased average tension load. of the cyclic attractor is proportionally to the composition and in-
Three different types of settlement/tilt components need to be tensity of the cyclic load history. Thus, for the assessment of the
distinguished: cyclic accumulated average tilt of a stable OWT, only the cyclic
attractor for the largest cyclic load event needs to be determined,
• Static settlement/tilt due to the submerged weight of the meaning that only one cyclic load history needs to be considered
OWT. in the design. Cyclic attractors can be found for the accumulated
average tilt of shallow foundations. However, no attractors exist
• Peak settlement/tilt due to a ULS loads.
for vertical settlements of shallow foundations.
• Accumulated (average) settlement/tilt due to cyclic loading
from wind, wave and operation loads. 7 FOUNDATION STIFFNESS
Following traditional geomechanics, the static settlement/tilt The local foundation stiffness is the link between the geotechnical
can be further distinguished into immediate-, consolidation- and and structural designer. Foundation stiffness is an output of the
creep-settlements/tilt. The corresponding values can be computed geotechnical analysis, but is not part of the actual sizing, i.e. ca-
using well established geotechnical calculations procedures. pacity serviceability and installation analysis. However, the results
The peak settlement/tilt can be assessed by means of a mono- of the stiffness analysis will affect the design basis as illustrated in
tonic pushover FE analysis. The soil model needs to be cali- Figure 2. As detailed in Subsection 1.2, foundation stiffness can
brated in order to reproduce the correct stress-strain-behavior of be provided as single secant stiffness values, nonlinear tangential
the soil. Where necessary, the decrease of strength and stiffness stiffness values, or full linear or non-linear stiffness matrices in-
due to cyclic loading needs to be included. This may be done for cluding coupling terms if necessary. This needs to be agreed in
example by using a total-stress-based model with adjusted stress- upfront with the involved disciplines and may be included in the
strain curves based on cyclic contour diagrams, or an effective- load document. Further, it need to be agreed on the load cases for
stress-based model to which a pore pressure field is superimposed; which the foundation stiffness shall be assessed.
see also Section 3. The peak settlement/tilt represents actually the Foundation stiffness can be established using simplified analyt-
maximum expected value, meaning that the load case considered ical methods or advanced FEM based methods. Gazetas (1991) has
is in general the same used in the ULS capacity analysis, but with- proposed a large number of closed form equations for assessing the
out applying load and resistance factors. Practically, this value is stiffness of different foundation types and ground conditions. In
less relevant, as the settlement – and more important the tilt – will contrast to the simplified methods, which consider linear soil prop-
immediately decrease again in the subsequent unloading. Further, erties, the FEM allows to capture the non-linearity of the soil and
the OWT may not be in operation during the ULS event, for which the flexibility of the structure, i.e. the soil-structure-interaction.
reason the allowable serviceability limit criteria may not apply. The methods used for assessing the foundation stiffness should be
Most relevant is the assessment of the accumulated average set- adjusted based on the stage of a project and anticipated degree of
tlement/tilt, which, however, is also one of the most challenging optimization. In an early stage of a project, i.e. feasibility and con-
components. Thereto, different strategies can be applied. One of cept study, simplified analytical methods may be used, whereas in
the most conservative assumptions is to take all load cycles which a FEED and Detailed Design the FEM may be more appropriate.
occur during the lifetime of an OWT and sort them in ascending Typically the foundation stiffness is provided as a range with
order. This sorted cyclic load history can be applied in a calcula- high-, best- and low-estimate. The width of the range should be
tion procedure, for example in the NGI method (Jostad et al. 2014, narrowed down during the project and every design iteration. No
2015b), or in an FE analysis using the high-cyclic accumulation attempts should be made by the geotechnical designer to assume
model (Niemunis et al. 2005, Wichtmann et al. 2010). any particular soil profile which may be conservative for the struc-
Since small load cycles will typically not contribute signifi- tural design. The selected soil profiles should rather reflect the
cantly to the accumulated total displacements, a different approach inherent uncertainties of the soil state after installation and load
has been followed in more recent projects. The design storm used conditions.
in the ULS analysis, which is based on a 50 years wind wave event, Two different type of stiffness values need to be distinguished;
has been extrapolated to other storm events with different recur- stiffness values for the structural utilization (denoted in the follow-
rences using a Gumble distribution. That enables to calculate the ing ULS load case) and stiffness values for the load assessment,
58
9 9
9 ηLE
8 tLE
8 ηHE 8 tHE
7
7 7
6
6 6
5
5 5
4
4 4
3
3 3
2 2
1
1 1
2 50 100 150 200 0 100 200 300
Sketch of jacket Norm. ULS utilization η/ηBE [%] Norm. fatigue life t/tBE [%]
Figure 12: Impact of the HE and LE ULS and FLS stiffness on the jacket response
Figure 13: Example of a global model used for assessing local foundation stiffness
structural fatigue analysis and eigenmode analysis (denoted in the end of the first iteration of a generic FEED study. The correspond-
following FLS load case). In general the ULS stiffness is non- ing values are normalised by the best-estimate values. The corre-
linear due to the high mobilization of the foundation, wheres the sponding normalized stiffness values are listed in Table 2.
FLS stiffness appears often to be linear due to the significant lower It becomes apparent that the structural utilization scales approx-
load amplitudes. The soil profiles used in both the simplified and imately proportional with the ULS foundation stiffness. The vari-
the advanced analysis need to reflect the different loading condi- ation in utilization, however, is not reflecting the relatively large
tions. The FLS load case is governed by the cyclic amplitude, range in ULS foundation stiffness values, meaning that the jacket
where the average or mean load is of less important. is less sensitive to variations in foundation stiffness. In contrast to
that, is the impact of the high- and low-estimate FLS stiffness on
Table 2: Variation of the HE and LE ULS and FLS stiffness nor- the fatigue life very pronounced. Even though, the low estimate
malized by the corresponding Best Estimate (BE) stiffness FLS stiffness is 69% of the best-estimate FLS stiffness, the fatigue
life decreases to less than 10%.
ULS FLS This example illustrates, that an optimization of an OWT can be
challenging, if the range of foundation stiffness values is too large.
High Estimate (HE) 288% 189%
Furthermore the implications of assumptions in the geotechnical
Low Estimate (LE) 37 % 69% design on the structural design can be hardly estimated without
performing corresponding structural analysis.
Figure 12 shows the impact of the high-, best- and low-estimate When assessing the foundation stiffness for a mono-caisson us-
foundation stiffness on both the fatigue life and the structural uti- ing the FEM, the loads provided in the design basis can be di-
lization of a jacket supported by three suction caissons. Basis for rectly applied to the caisson. When using the calculated foundation
this analysis are the structural and geotechnical properties at the stiffness values in the subsequent structural analysis the updated
59
loads are typically of similar order and ratio. That means the load- 70
Integrated lid pressure / vertical load [%]

stiffness iteration – outer loop in Figure 1 – converge relatively L01
L02
fast. 60 L03
This is somewhat more complicated for multi-legged sub- L04
50 L05
structures. Depending on the footprint width, caisson dimensions,
load conditions, and ground conditions, the loads can be redis-
40
tributed between the different legs due to both the flexibility of
the sub-structure – i.e. the jacket – and the interaction of the cais-
30
sons in the ground. Using FE models of single caissons only will
not capture the redistribution correctly. More accurate would be 20
to model both the caissons, the sub-structure and the soil. This is
denoted global FE model and is shown in Figure 13. The differ- 10
ence between a global model and an integrated model is the type of
analysis. A global model is typically used in monotonic push-over 0
-40 -20 0 20 40 60 80 100
analysis, whereas an integrated model is used in a time domain
Vertical load / ULS compression load [%]
analysis.
Advantage of a global model is, that the loads and foundation Figure 14: Ration between the loads carried by the lid and the skirts as
stiffness values can converge relatively fast in just some few load- function of the load amplitude
stiffness iterations. However, such analysis are time consuming, 100
and – depending on the stage of a project – single caisson models
Vertical load / ULS compression load [%]

may be used instead, though the accuracy is less good. Based on 80
recent experience, it is recommended to use global FE model in
FEED and detailed design at some representative locations of an 60
offshore wind farm. The identification of relevant locations can
be reasonably well done using the simplified methods or single FE 40
models, as the error is in general proportional.
20
It may be noted that global FE models are particular relevant
for assessing ULS foundation stiffness due to the large mobiliza-
0 L01
tion. For FLS load cases, single caissons models are sufficient. An L02
exception is the assessment of foundation damping, both for ULS -20 L03
L04
and FLS. If FE analysis is used for determine foundation damping, L05
the complete soil may be modeled to capture the interaction and -40
larger soil mass. -0.01 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Lid deflection [mm]
In addition should be mentioned that attempts are undertaken to
use macro elements in structural analysis. However, the macro ele- Figure 15: Deflection of the lid as function of the load amplitude
ments require a calibration of the particular site and caisson geom-
etry, which can be done for some models using the above described
methods. applied load. Small load amplitudes are taken solely by the skirts.
The increase in lid pressure is linear and larger for tension loading,
8 SOIL REACTIONS though a considerable scatter can be observed on the tension side.
Figure 15 shows the displacements of the caisson at the load-
Soil reactions are, like foundation stiffness, an output of the ing point, which is in this case the top of the lid. It appears to be
geotechnical design, but are not considered in the geotechnical siz- very linear for low and medium sized load amplitudes, whereas
ing of the caisson. However, the results of the stiffness analysis it becomes pronounced non-linear for larger loads. Variations
can affect the design basis as illustrated in Figure 2. Soil reactions of the lid and caisson stiffness showed, that the linear response
are typically provided as loads distributed over the skirt(s) and lid at low and medium sized loads is directly proportionally to the
which is in contact with the soil. The assessment of load reactions caisson stiffness. The soil non-linearity becomes first visible at
is difficult and depends on many factors, such as the flexibility of larger loads. This demonstrates the importance to select the correct
the caisson, the soil layering, the recent cyclic load history, and the geotechnical-structural interface point discussed in Section 1.2.
actual applied load for which the soil reactions shall be provided. Thought the results are encouraging, the calculation of the ac-
Soil reactions are used for the structural design of the caisson, tual values using the FEM is very time consuming and sensitive
and need to be provided for two different cases; for installation and to the modeling. The stresses depend on the size, number, shape,
for in-place conditions. The soil reactions during installation will and type of continuum element used in the analysis. Further, the
typically govern the required thickness and shape of the skirt wall – tip resistance is difficult to assess due to the ratio of wall thickness
assessed in buckling analysis – whereas the in-place soil reactions to caisson diameter, which requires exceptionally small elements
will primarily govern the design of the caisson lid. Of particular along the skirt and below the tip. Thus, the FEM may be used as a
importance is the distribution of the loads carried by the lid and the complementary method for assessing the soil reactions.
skirts. In conventional offshore geotechnical engineering, soil reac-
Figures 14 and 15 show the result of FE analysis of five suction tions are established based on engineering judgement and are pro-
caissons with different geometries and soil conditions, subjected vided as so-called unit loads. Unit loads scale proportionally with
to short-term compression or tension loading, respectively. The the applied total load. Assumptions are made on the distribution;
load conditions are representative for a compression or tension leg similar to the one shown in Figure 14. Upper and lower estimates
of a multi-legged sub-structure. From Figure 14 becomes appar- of the distribution need to be provided.
ent, that the load taken by the lid scales proportionally with the Soil reactions for the installation can be derived from the ac-
60
tual penetration analysis, where both skirt wall friction, skirt tip 9.3 Earthquake loading
resistance, and required suction pressure are calculated. More
In some parts of the world, earthquake loading and earthquake in-
complicated is to assess the soil support during installation. It
duced liquefaction needs to be considered in the design. Both can
may be assumed that almost no support is provided for penetra-
be considered in the design using existing methods. The loads from
tion in sand where significant flow gradients in the soil plug is
the earthquake represent just another load case to which particular
expected (drained penetration). Some support may be assumed for
soil conditions need to be assigned in the corresponding analysis.
undrained penetration, which may be estimated based on suitable
Kaynia (2017) provides a comprehensive introduction to the de-
laboratory tests.
sign of OWTs subjected to earthquake loading.
9 OTHER ASPECTS
9.4 Observational method
9.1 Grouting OWTs supported by caissons are a relatively new concept and
An issue often discussed is the necessity of grouting. Grout is used long-term experience does not exist yet. Thus, the observational
to fill the void between the lid and the soil at the inside of the method may be considered in current projects. It can be applied
caisson. Most suction caissons and anchors installed so far were during both installation and operation. In order to use the obser-
grouted with only some few exception. Main reason of using grout vational method, it is important that the failure is ductile, which
for suction caissons of bottom fixed OWTs is to reduce or avoid allows to initiate mitigation measures(s) in time.
potential (differential) settlements, and pumping-effects. Due to The observational method is a combination of predictions and
cyclic vertical loading, the water cushion below the lid is exposed measurements. The behavior of the OWT is calculated using ex-
to continuous pressure pulses, which can trigger a local piping fail- isting methods. Further, ranges of allowable values need to be de-
ure along the skirts. In addition, a lack of soil/grout support below fined. If exceeded, mitigation measures need to be initiated, which
the lid will cause large local stresses and moments in the lid. All need to be planned in the forehand.
loads need to be transferred trough the lid into the skirts. This re- Examples of mitigation measures during installation are pre-
quires a thick, massive lid, to avoid large deflections and fatigue sented in Subsection 5.3. The decision value is typically the re-
issues. quired suction pressure, which shall be provided as a range with
In order to improve the bearing behavior of the lid and and high and low estimate. If the high estimate value is exceeded, the
to avoid the afore mentioned negative effects when omitting mitigation measures may need to be applied. The same concept
grout, structural components may be applied to replace the grout. can be applied for the serviceability. When a maximum tilt is ex-
Stopper-pods, which are elements made of steel, hard rubber, or ceeded the OWT may need to be rectified.
composite materials, can be attached under the lid. The caisson For a successful application, it is important to plan both an ap-
need to penetrate until the pods are in contact with the mudline propriate health monitoring system and mitigation measures. As-
enabling to transfer loads from the lid into the soil. Alternatively, pects of monitoring systems are presented by Sparrevik & Strout
small ribs or T-beams may be welded under the lid dividing the (2015). The usefulness of such systems is presented by Schonberg
base into compartments. The structural elements may be slightly et al. (2017) at the example of the Borkum Riffgrund 1 Suction
cone-shaped to allow partially penetration into the ground in order Bucket Jacket.
to compensate for inclined mudline or uneven soil heave. If, an
uneven soil surface is expected, a jetting system may be used to 9.5 Wind farm design
flush the upper soil and by that generating a slurry mixture which
So far, only single OWT foundations were considered herein. An
slowly consolidated during the final phase of the installation.
iterative approach as outlined in Section 1.1 at each turbine loca-
Disadvantage with using structural and jetting systems is, that
tion of an Offshore Wind Farm (OWF) would require considerable
the soil which is in contact with the structural components is soft,
time, which may not be possible in the given project time frame.
and the stresses in the lid may be concentrated to some few points
Thus, a clustering may be introduced. Typically, the clustering
only. Based on current experience, the use of grout seems to be
is a based on the water depth, since the loads are expected to be
appropriate to optimize the lid geometry. However, the cost sav-
very similar for a given depth. Foundation capacity and installa-
ings due to an optimized lid geometry, needs to be compared to the
tion analysis can be relatively quickly performed. The results of
costs of the additional offshore work for the grouting.
these can be used for the sizing. If FEM is used for the capacity
analysis, as described by Jostad & Andersen (2015), the foundation
9.2 Integrated design approach
stiffness can be qualitatively estimated. Based on that, the softest
As mentioned in the several sections, so-called integrated analysis and stiffest location within a cluster can be identified. These two
are performed in OWT design (e.g. Krathe & Kaynia 2016, Page can then used in the stiffness and soil reaction analysis, represent-
et al. 2016, Skau et al. 2017). Such analysis are particularity suited ing the parameters used in the iteration process, given that the same
for structural analysis, such as in the load or eigenmode assess- sub-structure will be applied in the cluster.
ment. Integrated analysis are not appropriate for the foundation
sizing, though some macro-elements may indicate this possibil- 10 OUTLOOK
ity. The foundation response can be very sensitive to soil layering
and size of the caisson and skirt, which cannot be considered by In this article a general overview of the geotechnical design of suc-
the macro-elements. Furthermore, also other aspects than capacity tion caissons for OWTs has been provided. The different phases
and serviceability may be design driving as detailed in the corre- of the design were detailed, and the relevant aspects were outlined.
sponding sections in this contribution. The general design of suction caissons is reasonably well under-
Until today, the geotechnical sizing is uncoupled from the struc- stood and many authors have proposed numerous methodologies
tural analysis and it is not expected that this may change in the near for specific geotechnical calculations, such as capacity, installa-
future without compromising an optimization of the caisson geom- tion and stiffness analysis. Only some few were mentioned in this
etry. article; mainly those which are familiar to the author from personal
61
experience. The reader is encouraged to get himself an overview Dove, D., Roberts, D., Evans, D., Tappin, D., Lee, J., Long, D.,
of the numerous methods proposed in the literature. The article Mellett, C. & Callard, S. (2016), Refining glacial statigraphy
at hand may serve as a guideline to evaluate the suitability of a in the southern north sea - new bathymetric model brings re-
method for the particular design aspect. newed value to legacy seismic, in ‘Near Surface Geoscience
Due to the increased interest in suction caissons for OWTs, 2016 - Second Applied Shallow Marine Geophysics Confer-
a number of researchers and practitioners are currently working ence’, EAGE.
to continuously advance the knowledge. Several of the currently Foglia, A., Govoni, L., Gottardi, G. & Ibsen, L. (2014), Investi-
designed and installed OWTs are equipped with comprehensive gations on macro-element modelling of bucket foundations for
health monitoring systems, which will provide further inside into offshore wind turbines, DCE Technical Memorandum 48, Aal-
the short- and long-term behavior of suction caissons. borg: Department of Civil Engineering, Aalborg University.
Forsberg, C., Lunne, T., Vanneste, M., James, L., Tjelta, T., Bar-
ACKNOWLEDGMENT wise, A. & Duffy, C. (2017), Synthetic cpts from intelligent
ground models based on the integration of geology, geotechnics
The authors gratefully acknowledge the support by his colleagues and geophysics as a tool for conceptual foundation design and
at NGI in the numerous projects which provided the basis for this soil investigation planning, in ‘Offshore Site Investigation and
article. Thanks goes to Knut H. Andersen and Youhu Zhang for Geotechnics: Smarter Solutions for Future Offshore Develop-
the review of the manuscript. ments (SUT-OSIG)’.
Gazetas, G. (1991), Foundation vibrations, in H.-Y. Fang,
11 REFERENCES ed., ‘Foundation Engineering Handbook’, Springer Sci-
ence+Business Media.
Aas, P., Saue, M. & Aarsnes, J. (2009), Design predictions and Houlsby, G. & Byrne, B. (2005a), ‘Design procedures for instal-
measurements during installation of suction anchors with and lation of suction caissons in clay and other materials’, Proceed-
without water-flow system to help installation through layered ings of the Institution of Civil Engineers - Geotechnical Engi-
soil profiles, in ‘Offshore Technology Conference, 40. Hous- neering 158(GE2), 75–82.
ton’, number OTC-20294-MS. Houlsby, G. & Byrne, B. (2005b), ‘Design procedures for installa-
Andersen, K. (2015), Cyclic soil parameters for off- tion of suction caissons in sand’, Proceedings of the Institution
shore foundation design : The third issmge mcclel- of Civil Engineers - Geotechnical Engineering 158(GE3), 135–
land lecture, in V. Meyer, ed., ‘Frontiers in Offshore 144.
Geotechnics III, Oslo, Norway’, pp. 5–82. Revised IEC (2009), Wind turbines - part 3: Design requirements for off-
version in: http://www.issmge.org/committees/technical- shore wind turbines, Standard IEC 61400-3:2009, International
committees/applications/offshore and click on Additional Electrotechnical Commission.
Information. Johansson, P., Aas, P. & Hansen, S. (2003), Field model tests for
Andersen, K., Jostad, H. & Dyvik, R. (2008), ‘Penetration resis- a novel suction anchor application, in ‘6. International Sympo-
tance of offshore skirted foundations and anchors in dense sand’, sium on Field Measurements in Geomechanics, Oslo, Norway’,
Journal of geotechnological and geoenvironmental engineering pp. 145–153.
134(1), 106–116. Jostad, H. & Andersen, K. (2015), Calculation of undrained hold-
Andersen, K., Puech, A. & Jardine, R. (2013), Cyclic resistant ing capacity of suction anchors in clays, in V. Meyer, ed., ‘Fron-
geotechnical design and parameter selection for offshore engi- tiers in Offshore Geotechnics III, Oslo, Norway’, pp. 263–268.
neering and other applications, in ‘ISSMGE - TC 209 Workshop Jostad, H., Andersen, K., Khoa, H. & Colliat, J. (2015a), Interpre-
- Design for cyclic loading: Piles and other foundations - Pro- tation of centrifuge tests of suction anchors in reconstituted soft
ceedings of TC 209 Workshop, 18th ICMGE, Paris’. clay, in V. Meyer, ed., ‘Frontiers in Offshore Geotechnics III,
BSH (2015), Minimum requirements concerning the constructive Oslo, Norway’, pp. 269–276.
design of offshore structures within the Exclusive Economic Jostad, H., Grimstad, G., Andersen, K., Saue, M., Shin, Y. & You,
Zone (EEZ), number 7005, 2 edn, BSH. D. (2014), ‘A fe procedure for foundation design of offshore
Byrne, B. (2000), Investigations of suction caissons in dense sand, structures – applied to study a potential owt monopile foun-
PhD thesis, Department of Engineering Science, The University dation in the korean western sea’, Geotechnical Engineering
of Oxford. 45(4), 63–72.
Colliat, J., Dendani, H. & Schroeder, K. (2007), Installation of Jostad, H., Grimstad, G., Andersen, K. & Sivasithamparam, N.
suction piles at deepwater sites in angola, in ‘Offshore Site In- (2015b), A fe procedure for calculation of cyclic behaviour
vestigation and Geotechnics: Confronting New Challenges and of offshore foundations under partly drained conditions, in
Sharing Knowledge (SUT-OSIG)’. V. Meyer, ed., ‘Frontiers in Offshore Geotechnics III, Oslo, Nor-
Cotter, O. (2009), The installation of suction caisson foundations way’, pp. 153–172.
for offshore renewable energy systems, PhD thesis, University Kaynia, A. (2017), Earthquake response of offshore wind turbines,
of Oxford, Magdalen College. in ‘International Conference on Performance-based Design in
Cotterill, C., Phillips, E., James, L., Forsberg, C., Tjelta, T., Carter, Earthquake Geotechnical Engineering (PBD-III)’.
G. & Dove, D. (2017), ‘The evolution of the dogger bank, north Kelly, R., Houlsby, G. & Byrne, B. (2006), ‘A comparison of field
sea: A complex history of terrestrial, glacial and marine environ- and laboratory tests of caisson foundations in sand and clay’,
mental change’, Quaternary Science Reviews (171), 136–153. Géotechnique 56(9), 617–626.
Cudmani, R. & Sturm, H. (2006), An investigation of the tip re- Krathe, V. & Kaynia, A. (2016), ‘Implementation of a non-linear
sistance in granular and soft soils during static, alternating and foundation model for soil-structure interaction analysis of off-
dynamic penetration, in ‘Int. Sym. on vibratory pile driving and shore wind turbines in fast’, Wind Energy 20(4), 695–712.
deep soil compaction, TRANSVIB’. Langford, T., Solhjell, E., Hampson, K. & Hondebrink, L. (2012),
DNV-GL (2016), Support structures for wind turbines, Standard Geotechnical design and installation of suction anchors for the
DNVGL-ST-0126. skarv fpso, offshore norway, in ‘Offshore Site Investigation and
62
Geotechnics: Integrated Geotechnologies - Present and Future Sturm, H. (2011), ‘Geotechnical performance of a novel grav-
(SUT-OSIG)’, pp. 613–620. ity base type shallow foundation for offshore wind turbines’,
Leinenkugel, H. (1976), Deformations- und Festigkeitsverhal- Geotechnik 34(2), 85–96.
ten bindiger Erdstoffe. Experimentelle Ergebnisse und ihre Sturm, H. & Mirdamadi, A. (2017), Reliability based stiffness
physikalische Deutung, PhD thesis, University of Karlsruhe. analysis for application during installation of suction cais-
Nguyen-Sy, L. (2005), The theoretical modelling of circular shal- sons, in ‘Proceedings of the 36th International Conference
low foundations for offshore wind turbines, PhD thesis, Univer- on Ocean, Offshore and Arctic Engineering, OMAE’, number
sity of Oxford, Magdalen College. OMAE2017-62043.
Nguyen-Sy, L. & Houlsby, G. (2005), The theoretical modelling Sturm, H., Nadim, F. & Page, A. (2015), A safety concept for
of a suction caisson foundation using hyperplasticity theory, in penetration analyses of suction caissons in sand, in V. Meyer,
‘Frontiers in Offshore Geotechnics I, Perth, Austrlia’, pp. 417– ed., ‘Frontiers in Offshore Geotechnics III, Oslo, Norway’,
423. pp. 1393–1398.
Niemunis, A., Wichtmann, T. & Triantafyllidis, T. (2005), ‘A high- Svanø, G., Eiksund, G., Kavli, A., Langø, H., Karunakaran, D.
cycle accumulation model for sand’, Computers and Geotech- & Tjelta, T. (1997), Soil-structure interaction of the draupner e
nics 32(4), 245–263. bucket foundation during storm conditions, in ‘8. International
Norén-Cosgriff, K., Jostad, H. & Madshus, C. (2015), Idealized Conference on the Behaviour of Offshore Structures, Delft’,
load composition for determination of cyclic undrained degra- Vol. 1, pp. 163–176.
dation of soils, in V. Meyer, ed., ‘Frontiers in Offshore Geotech- Wichtmann, T., Niemunis, A. & Triantafyllidis, T. (2010), Appli-
nics III, Oslo, Norway’, pp. 1097–1102. cation of a high-cycle accumulation model for the prediction
of permanent deformations of the foundations of offshore wind
Page, A. M., Schafhirt, S., Eiksund, G., Skau, K., Jostad, H. &
power plants, in ‘Frontiers in Offshore Geotechnics II, Perth,
Sturm, H. (2016), Alternative numerical pile foundation mod-
Austrlia’.
els for integrated analyses of monopile-based offshore wind tur-
bines, in ‘26. International Offshore and Polar Engineering Con-
ference (ISOPE)’, pp. 111–119.
Salciarini, D., Bienen, B. & Tamagnini, C. (2011), A hy-
poplastic macroelement for shallow foundations subject to six-
dimensional loading paths, in ‘International Symposium on
Computational Geomechanics (ComGeo II), Cavtat-Dubrovnik,
Croatia’, pp. 721–733.
Salciarini, D. & Tamagnini, C. (2009), ‘A hypoplastic macroele-
ment model for shallow foundations under monotonic and cyclic
loads’, Acta Geotechnica 4(3), 163–176.
Saue, M., Aas, P., Andersen, K. & E., S. (2017), Installation of
suction anchors in layered soils, in ‘Offshore Site Investigation
and Geotechnics: Smarter Solutions for Future Offshore Devel-
opments (SUT-OSIG)’.
Schonberg, A., Harte, M., Aghakouchak, A., Brown, C., Andrade,
M. & Liingaard, M. (2017), Suction bucket jackets for offshore
wind turbines: applications from in situ observations, in ‘In Pro-
ceedings of TC209 Workshop (Foundation design of offshore
wind structures): 19th International Conference on Soil Me-
chanics and Geotechnical Engineering, Seoul, South Korea’.
Senders, M. & Randolph, M. (2009), ‘Cpt-based method for the
installation of suction caissons in sand’, Journal of geotechno-
logical and geoenvironmental engineering 135(1), 14–25.
Skau, K., Chen, Y. & Jostad, H. (2017), ‘A numerical study of ca-
pacity and stiffness of circular skirted foundations in clay sub-
jected to combined static and cyclic general loading’, Géotech-
nique .
Solhjell, E., Blaker, Ø., Knudsen, S. & Rahim, A. (2014), Geotech-
nical design and installation of suction anchors for the goliat
fpso, offshore norway, in ‘Offshore Technology Conference
Asia, Kuala Lumpur, Malaysia’, number OTC-24989-MS.
Sparrevik, P. (2002), Suction pile technology and installation in
deep waters, in ‘34. Offshore Technology Conference (OTC),
Houston, US’, number OTC 14241.
Sparrevik, P. & Strout, J. (2015), Novel monitoring solutions solv-
ing geotechnical problems and offshore installation challenges,
in V. Meyer, ed., ‘Frontiers in Offshore Geotechnics III, Oslo,
Norway’.
Sturm, H. (2009), ‘Numerical investigation of the stabilisation be-
haviour of shallow foundations under alternate loading’, Acta
Geotechnica 4(4), 283–292.
63

64
Suction bucket jackets for offshore wind turbines: applications from in situ observations
Suction caisson jackets pour des éoliennes en mer: applications des observations in situ
Avi Shonberg, Michael Harte, Amin Aghakouchak, Cameron S. D. Brown, Miguel Pacheco Andrade and
Morten A. Liingaard
DONG Energy, [email protected]
ABSTRACT: Offshore wind foundations of the future face many challenges. Due to the growing demand for renewable energy, future
projects will be built further offshore in deeper water with larger wind turbine generators (WTGs) leading to increased foundation
loads. Furthermore, in some regions, foundations must be installed without exceeding strict regulatory requirements on underwater
noise. Thus, traditional foundation solutions are being pushed to their limits and there is a need for innovative foundation concepts.
The suction bucket jacket (SBJ) is one such foundation solution which addresses these future challenges. In 2014, DONG Energy
installed a highly instrumented SBJ at the Borkum Riffgrund 1 wind farm to support a Siemens 4 MW WTG. This is the first wind
turbine to be founded on a SBJ. Using monitoring data obtained between September 2014 and January 2016, this paper outlines some
key findings. The paper focuses on the in-place response of the SBJ to loading including long term behaviour of the structure, vertical
stiffness response, load transfer along the bucket skirt and the variation of pore water pressures.
RÉSUMÉ: Les fondations des futures éoliennes en mer font face aux nombreux défis. Par suite de la croissance des énergies
renouvelables, les futurs parcs éoliens en mer seront construits plus loin de la côte dans des plus grandes profondeurs d’eau et les
éoliennes seront plus grandes avec des charges augmentées sur les fondations. En plus, dans certains régions du monde il y a des
limitations sur les nuisances sonores sous-marin qui sont générés pendant l’installation des fondations. Les solutions des fondations
traditionnelles sont à leurs limites ce qui augmente la demande des idées innovantes. La technologie “suction caissons” réponds à ces
futurs défis. DONG Energy a déjà installé une fondation type "suction caissons" au parc éolien Borkum Riffgrund 1 avec une éolienne
de 4MW de capacité fabriqué par Siemens. C'est la première éolienne en mer qui est construit sur une fondation du type "suction
caissons". Des mesures ont été fait depuis l'installation de la fondation en 2014 jusqu’à janvier 2016. Les résultats de l'analyse de ces
données sont présentés dans cet article. Cet article mettra l'accent sur la réaction in-situ de la fondation type "suction caissons" au
chargement y compris le comportement à long terme de la structure, ainsi que la réaction au rigidité verticale, le transfert de charge le
long de la caisson et la variation de la pression interstitielle.
KEYWORDS: suction bucket jacket, shallow foundation, in place performance, case study
INTRODUCTION Overall, this paper uses the in situ observations from a world-
first full scale monitoring system to better understand the
A suction bucket jacket (SBJ) foundation has been installed at behaviour of the SBJ under environmental loading conditions.
the Borkum Riffgrund 1 (BKR01) offshore wind farm to support
a Siemens 4 MW wind turbine generator (WTG). The innovative
foundation concept (shown in Figure 1) was installed by DONG BACKGROUND
Energy in 2014 in collaboration with the Carbon Trust Offshore
Wind Accelerator (OWA) program in order to addresses future The 312 MW BKR01 offshore windfarm is located in the
challenges associated with offshore wind farm developments. German sector of the North Sea, approximately 38 km north of
Whilst the SBJ is not a new concept, the BKR01 SBJ is the first the island of Borkum (Figure 2). The wind farm contains 78
wind turbine to be founded on a SBJ structure. WTGs and covers an area of approximately 36 km2.
The BKR01 SBJ has been instrumented with a
comprehensive measurement system to monitor the structural
and geotechnical behaviour during the lifetime of the structure.
This paper outlines some key findings from analysis of the
monitoring data collected from 2014 to 2016. The paper analyses
the in-place response of the SBJ subjected to environmental wind
and wave loading acting on the WTG. This includes analyses
over a full year of monitoring data and much smaller subsets of
data where appropriate.
This paper investigates the long term behaviour of the
structure and the average vertical stiffness response of the
different suction bucket components. Importantly, the stiffness
response is also investigated as a function of loading direction
and loading level. Furthermore, the load transfer along the bucket
skirt using strain gauge measurements is compared with hand
calculations using conventional methods. Finally, the pore water
pressure response to different loading frequencies are analysed
to estimate the boundaries between drained and undrained
suction bucket behaviour.
Figure 1. Illustration of the BKR01 SBJ
65
Proceedings of TC 209 Workshop - 19th ICSMGE, Seoul, 20 September 2017
MONITORING SYSTEM
The BKR01 SBJ is extensively instrumented with strain gauges,

accelerometers, inclinometers, pressure sensors and temperature
sensors. The monitoring system, designed by the Norwegian
Geotechnical Institute (NGI), was installed to record the
structural and geotechnical response of the SBJ during the
offshore environmental loading experienced during the lifetime
of the structure. As described by Sturm (2017), very few
examples of full-scale monitoring systems for offshore structures
similar to the BKR01 SBJ are found in the literature, thus the
monitoring system installed at BKR01 provides a significant new
dataset for observations of this nature. The monitoring system, as
Figure 2. Location of the BKR01 wind farm also described by Sparrevik & Strout (2015) and Ehrmann et al.
(2016), included measurement points on the jacket, at the
One of the foundations installed at BKR01 is a three-legged connection between the jacket and the buckets and on each of the
SBJ foundation as illustrated in Figure 1. The layout and bucket three suction buckets.
naming convention is shown in Figure 3. At BKR01, the Table 1, Figure 3, Figure 4 and Figure 5 provide an overview
remaining WTGs are supported by monopile foundations. The of the sensors installed on each bucket relevant to the
SBJ foundation has a soft-stiff design and supports a Siemens geotechnical analyses. Bucket forces were measured in 6 degrees
4MW WTG. The BKR01 SBJ was successfully installed using a of freedom (DOF) at each bucket via a bespoke ‘clean section’
suction assisted operation in the summer of 2014. connection (further discussed in Section 3.1) between the bucket
Prior to the BKR01 project, the SBJ foundation concept had and jacket leg. For redundancy and efficiency, the monitoring
been used for other North Sea foundation structures, such as the system was different for each bucket. Bucket BC is the most
Sleipner T (Tjelta, 1994; Bye et al., 1995; Tjelta, 1995) and heavily instrumented bucket.
Draupner E structures (Erbrich & Tjelta, 1999) but never as a
foundation for a WTG. The BKR01 SBJ was commissioned to Table 1. Bucket monitoring system measurements and number of sensors
assess the economic and environmental benefits associated with Measurement or sensor Bucket BC Bucket AB Bucket AC
jacket structures and suction assisted installation for WTG
foundations. Bucket forces in 6 DOF Yes Yes Yes
The soil conditions at the site are typical of German North Bucket lid accelerations 1 3 2 1
Sea conditions with the near surface consisting of medium dense 2
Bucket lid inclinations 3x1D 3x1D 1x2D
to dense sands. A layer of more silty sand is present at between
1
approximately 3 m and 6 m below seafloor (bsf). Skirt strains Yes No Yes
The suction buckets installed at BKR01 have a diameter (D) Excess pore pressures 3 Yes Yes No
of 8 m and a skirt length (L) of 8 m, giving a length to diameter 1
See Figure 4 and Figure 5
(L/D) ratio of 1. The skirt length adopted included an allowance 2
Where a 1D inclinometer measures inclinations about one axis only and a 2D
for the expected soil heave during installation and a minimum inclinometer measures inclination about two axes.
3
volume for grout post installation. The grout injection, Pore pressure sensors are located under the bucket lid and at 0.5 m from the skirt
tip (inside and outside). The sensors measure the differential pore pressures between
undertaken after installation of the SBJ, ensured full contact their level and the top (outside) of the bucket where a reference sensor was located.
between the underside of the bucket lids and the soil.
3.1 Clean section
A unique feature of the BKR01 SBJ monitoring system was the

inclusion of a ‘clean section’ at the connection between the
bucket and jacket leg to measure the force vector in 6DOF at this
critical location. The bespoke ‘clean section’, designed by
Ramboll, consisted of a ring stiffener to homogenise the forces
in the section and an array of 12 strain gauges distributed around
the inner circumference of the section which allowed for
correlation between measured strains and the 6DOF section force
vector. The ‘clean section’ was calibrated using known forces
during the installation process.
3.2 Monitoring system operation
The monitoring system was operational during the installation of

the SBJ, during placement of the tower and rotor/nacelle
assembly (RNA) onto the SBJ and during regular periods
between September 2014 and January 2016. The time periods for
which data is available from 2016 are shown in
Figure 6 (red zones indicate periods of some sensor functionality
loss, and the relevant sensor data has been excluded from the
analyses). Monitoring continues to date, although the results
from the aforementioned period are the only measurements
considered in this study.
Figure 3. SBJ layout and monitoring system setup
66
conditions, the horizontal to vertical load ratio at the mudline is

approximately 0.1.
The dead load from the entire structure (including the SBJ,
tower and RNA) have been deducted from all forces presented in
this paper. Therefore, the forces presented are relative to the dead
load of the entire structure. As shown in Figure 16, negative
values represent compressive live loads and positive values
represent tensile live loads. Absolute tensile forces have not been
observed during the monitoring program.
Table 2. BKR01 SBJ soil profile
Top of Relative Fines Peak
layer Description density content friction
(m bsf) (%) (%) angle (°)
Figure 4. Position of accelerometers on the bucket lids SAND, very dense,

0 100 5 - 10 47
medium to coarse grained.
SAND, medium dense,
2.9 fine to medium grained, 70 5 - 20 40
some layers of silty sand.
SAND, very dense,
5.6 90 5 - 15 45
medium to fine grained.
SAND, dense to very
dense, fine to medium
15.1 70 5 - 10 40
grained, some layers of
silty sand to sandy silt.
IN PLACE BEHAVIOUR
5.1 Inclination
The inclination of the BKR01 SBJ structure is measured by a bi-

directional inclinometer located on the transition piece (TP)
Figure 5. Position of strain gauges on the bucket skirts above the water surface. The resultant inclination of the structure
from Q4 2014 to Q1 2016 for standstill periods only (i.e. when
the wind turbine is not operational) is shown in Figure 7. The 10
minute averages of inclination are shown in the light colour and
the weekly average tilts are shown with the darker line. After first
Figure 6. Monitoring system operations power, the 10 minute average inclination is effectively removed
by the filtering as fewer standstill periods occur.
As expected, a minor change in inclination was observed
SITE CHARACTERISTICS immediately after the WTG installation, with the inclination from
vertical reducing. Since installation of the WTG, the inclination
4.1 Soil conditions from vertical of the BKR01 SBJ has generally stayed below
0.01°. Figure 7 shows that the resultant SBJ inclination has not
An extensive site investigation and laboratory testing program varied significantly with time. This indicates that environmental
was carried out specifically for the BKR01 SBJ, which included loads have little impact on the overall tilt of the structure in the
a CPT at each suction bucket location (outside the footprint of long term and that the structure is stable.
each individual bucket) and a central borehole from which
samples were retrieved.
The soil conditions at the site are typical of German North
Sea conditions with the near surface generally consisting of
medium dense to dense sands. A layer of more silty sand was
identified by the CPTs between approximately 3 m and 6 m
below the surface. Table 2 outlines the identified layers and
interpreted soil parameters. The water depth at the SBJ location
is 24.4m (relative to LAT). Scour protection was installed at the
SBJ location which included a pre-installed ‘filter layer’ and
post-installed rock armour layer.
4.2 Loading
Figure 7. Inclination measurements through time, for standstill
Operational wind loads at BKR01 are generally omnidirectional conditions only
but have a tendency towards emanating from headings between
When the 10 minute average inclination at the TP is plotted
200° and 300°. Storm wind loads and wave loads also tend to
against wind speed (Figure 8), it can clearly be observed that the
emanate from headings between 200° and 300°. During these
inclination varies with wind speed but returns to a small value at
conditions, higher compressive loads (referred to as
low wind speeds where the thrust-induced inclination is small.
‘compressive live loads’) are observed in Bucket AB and Bucket
The maximum inclination correlates with the maximum bucket
AC and lower compressive loads (referred to as ‘tensile live
force associated with the rated wind speed of the WTG
loads’) are observed in Bucket BC. Under normal operational
(approximately 12 m/s). The large scatter in inclinations at the
67
maximum wind speed is due to wind direction variation. The Both the Doherty et al. (2005) and Suryasentana et al. (2017)
cluster of data points at low inclinations do not follow the general methods are calibrated from extensive numerical modelling. A
trend and these relate to standstill or no power conditions. key assumption of both methods is that the lid is a rigid element,
although variations in skirt stiffness are taken into account by the
former. However, the response of the BKR01 SBJ bucket shows
that understanding the effect of the lid stiffness is vital to
correctly predicting the overall bucket response.
6.1 BKR01 SBJ overall suction bucket response

The suction bucket’s response to vertical loading is calculated
using the bucket specific vertical force vector and the bucket
specific vertical displacement, calculated by double integration
of the accelerations measured on the top of the buckets (az in
Figure 11). The bucket stiffness can only be estimated for the
dynamic part of the load as the accelerometers cannot measure
long term displacement relative to a reference point. Therefore,
the measured vertical stiffness of the suction bucket is a measure
of the suction bucket’s response to dynamic vertical loading,
often referred to as the FLS stiffness (Figure 10).
The suction bucket response to the dynamic portion of the
Figure 8. Normalised resultant TP inclination as function of wind speed load is predominantly undrained, as described in Section 9.
Therefore, the estimated vertical stiffness of the suction buckets
5.2 Excess pore pressures
implicitly includes a contribution from the resistant force
Figure 9 shows the 10 minute average of the excess pore generated by the excess pore water pressure beneath the bucket
pressures measured for Bucket BC during the monitoring period. lid.
Up until the time of first power, the excess pore pressures do not
vary from the long-term average indicating that the response is
essentially drained. After first power, the excess pore pressures
measured under the suction bucket lid show increased variability,
indicating some generation of excess pore pressure associated
with an undrained response. This trend is not observed at the skirt
tip, indicating that excess pore pressure is not generated, or is at
least not maintained, in the region of the skirt tip.
Figure 10. Idealised load-time history and load-deflection response
6.2 Stiffness components and calculation methodology
The measured bucket displacement is a combination of three

motions: bucket plunge, bucket rotation and lid deflection
(Figure 11). These motions are independently calculated and are
then used to determine the deflection of the bucket lid and the
Figure 9. Excess pore water pressure with time for Bucket BC a) under deflection of the entire bucket for the vertical stiffness
the lid, b) inside skirt tip and c) outside skirt tip calculations. The rotational stiffness is not considered in this
study.
Fz Fz
SUCTION BUCKET RESPONSE TO LOADING
az,center az,center
A key geotechnical input to the design of a multi-footed jacket az,0 az,180
structure is the response of the foundations. Similar to the
assumptions commonly adopted for spudcan analyses (Houlsby, dp az,0 az,180
2014) and as described by Dekker (2014) with specific reference
dℓ
to suction bucket foundations, a structural analysis of the jacket
structure often requires the geotechnical engineer to provide a
6X6 stiffness matrix for use as the boundary condition in the a
jacket model. Numerous authors have proposed methods for b
estimating the 6DOF suction bucket response to loading
(Doherty et al., 2005; Suryasentana et al., 2017) using a limited Figure 11. a) Bucket plunge (dp) and b) bucket lid deflection (dl)
number of site specific inputs such as soil stiffness and suction
bucket dimensions. Due to the ‘push-pull’ nature of the loading The following vertical stiffness components can be estimated
on the SBJ, the vertical stiffness will typically dominate the using the resolved force and the bucket displacement:
response in terms of soil structure interaction.  The soil-skirt stiffness (using the bucket plunge)
68
 The lid stiffness (using the lid deflection) 6.3.1 Effect of load direction on bucket vertical load
 The total bucket stiffness (using the total deflection) The measured mean vertical load (Fz) at the ‘clean section’ for a
series of nacelle positions and wind speeds is shown in Figure
6.3 Stiffness observations and results 13. There is a strong relationship between wind direction
(indicated as a heading in bold above each sub plot) and the
Figure 12 shows the 10 minute average normalised vertical
forces in the individual buckets. For example, for winds
stiffness for each stiffness component of Bucket BC after first
emanating from heading 232.5° to 262.5°, Bucket BC
power of the WTG. Results are presented from Bucket BC as
experiences a lower compressive load whilst Bucket AB and
only Bucket BC was equipped with the sensors to resolve all
Bucket AC experiences a higher compressive load.
displacement components.
For some intermediate loading directions, the load appears to
During periods of low excitation, the signal to noise ratio
act around an axis perpendicular to the loading direction. For
decreases and the stiffness cannot be estimated. Therefore,
example, with winds emanating from heading 262.5° to 292.5°,
periods at low wind speeds or standstill were filtered out and not
very little force is observed in Bucket AC whilst ‘push-pull’
analysed. The plot is presented in terms of probability density
forces are observed to be acting on Bucket AB and Bucket BC
which provides a relative indication of the number of occurrences
respectively.
of each normalised vertical stiffness value.
Figure 12. Total, lid and soil-skirt stiffnesses measured at Bucket BC
As expected, Figure 12 indicates that the suction bucket lid

exhibits a relatively consistent linear stiffness response to
loading which is independent of load level or load direction, as
the normalised lid stiffnesses fall within a narrow band of values.
The mean normalised vertical lid stiffness, kv/kv,ref, is Figure 13. Mean vertical load on the three suction buckets as a function
approximately 0.56. of wind speed and yaw direction
Conversely, the soil-skirt stiffness exhibits significant
variability during the assessed monitoring period, with
6.3.2 Effect of mean load level on bucket stiffness
normalised vertical stiffnesses ranging from approximately 0.38
to approximately 0.88. This indicates that the soil-skirt stiffness A key observation from the BKR01 SBJ response to loading is
is not constant as it is dependent on the load level, frequency that soil-skirt stiffness is dependent on the mean load level acting
and/or load direction, which is further described in Section 6.3.1 on the bucket. In Figure 14, the soil-skirt stiffness values binned
and Section 6.3.2. The mean normalised vertical soil-skirt by wind direction are shown for Bucket BC. This shows that
stiffness is approximately 0.66. The mean stiffness of the suction Bucket BC behaves stiffer when subject to higher compression
bucket is greater than the mean stiffness of the lid potentially due loads (winds generally from the north east with a heading of
to the large soil mass being mobilised during loading. between -7.5° and 82.5°), as indicated by the blue series’, and
During design, the suction bucket’s structural elements less stiff when subject to lower compression loads (winds
(bucket lid and soil-skirt response) are often idealised as a series generally from the south west with a heading of between 202.5°
of springs. When calculating the total bucket vertical stiffness, it and 262.5°), as indicated by the yellow-orange series’.
is often assumed that the different components of the bucket can This effect is studied by focusing on two wind direction bins
be combined as system of spring in series, such that: which give the maximum and minimum compression loads on
Bucket BC. Examining these two directional bins, the effect of
= (1) static (mean) compressive loads on the bucket becomes more
apparent. Orientation of these wind direction bins are shown in
Calculating the total stiffness using Eq. 1, with the mean Figure 15 with the maximum compression load bin shown in red
stiffness values shown in Figure 12, a total normalised vertical (wind from heading 37.5° to 97.5°) and the minimum
stiffness of approximately 0.31 is found. This correlates well compression load bin shown in blue (wind from heading 217.5°
with the observed total bucket vertical stiffness shown in Figure to 277.5°).
12. Therefore, observations from the BKR01 SBJ confirm the
assumption that in terms of vertical stiffness, the suction bucket
lid and soil response can be treated as pair of vertical springs in
series.
69
Where C is a material constant, f(e) is the void ratio

normalisation function, Pr is the reference pressure and α varies
from 0.5 to 0.6 at small strains and increases to 1 as higher shear
strain levels are reached. Using the relationship in Eq. 2 with the
methods proposed by Doherty et al. (2005) or Suryasentana et al.
(2017) a prediction of the increase in bucket stiffness as a
function of increase in vertical effective stress can be made.
However, the increase in vertical effective stress beneath the
suction bucket is unknown as the bucket skirt and the pore water
(undrained behaviour) will take a portion of the load. By using
the measured total bucket stiffness as a function of change in load
level (Figure 16), the increase in vertical effective stress beneath
the bucket as a proportion of the applied load (B) can be back-
calculated to provide an estimate of the load distribution between
lid and skirt.
Figure 14. Bucket BC stiffness as a function of nacelle position and Assuming a linear relationship between soil shear stiffness
wind speed and foundation stiffness (Suryasentana et al., 2017), the vertical
effective stress under the foundation ( ′ ) after the application of
environmental loads (Fz) can be calculated by assuming the
initial vertical effective stress ( ´ , the foundation stiffness prior
to applying the load (K1) and the new foundation stiffness value
(K2). This relationship is summarised by Eq 3.
′ (3a)
′
′ (3b)
′
′ ′ (3c)

Where A is the area of the suction bucket and B is the proportion
of the applied load which is observed beneath the suction bucket.
For the BKR01 SBJ, the initial stiffness (K1) and initial
vertical effective stress ( ´ are taken from the point where the
Figure 15. Two sectors aligned with Bucket BC
maximum tensile live load is observed (approximately 3500 kN
on Figure 16b). It is assumed at this point that the vertical
The resulting soil-skirt stiffness in these two load direction
effective stress is equal to the in situ vertical effective stress prior
bins is shown in Figure 16a. When Bucket BC is experiencing
compressive live loads (higher compression loads, shown in red), to installation of the SBJ such that ′ .
the normalised vertical stiffness is notably higher than when the
bucket is experiencing tensile live loads (lower compression
loads, shown in blue). This confirms that load level (which is a
function of wind direction) has an effect on vertical stiffness.
Figure 16b shows two clear regions of distinct behaviour.
When Bucket BC is experiencing compressive live loads, the
foundation stiffness stays relatively constant. When the bucket is
experiencing tensile live loads, the foundation stiffness tends to
decrease linearly with decreasing compressive loads. This is
particularly noticeable for the upper bound envelope of the data
cloud where less scatter is observed.
These variations in stiffness may be a result of changes in the
vertical effective stress regime under different loading
conditions. An increase in the mean load level due to
environmental loads (predominantly wind loading) will result in
an increase of the mean vertical effective stress beneath the
foundation. As the BKR01 SBJ is founded on predominantly
sandy strata, this will increase the vertical capacity of the suction
bucket as well as its stiffness response.
Wroth & Houlsby (1985) proposed a power function
relationship between the shear stiffness and effective mean
stress, p´. Aghakouchak (2015), in a set of recent laboratory tests
on quartz sand, showed that increases in vertical effective stress
caused an increase in the shear stiffness of sand. The relationship
between effective stress and shear stiffness can be presented as:
´
(2) Figure 16. Vertical soil-skirt stiffness in two directional sectors of
Bucket BC a) versus wind speed and b) versus vertical load
70
By using Eq. 3, the directional dependent stiffness 6.4 Variation of stiffness between buckets
observations from Bucket BC presented in Figure 16b for tensile
Figure 18 presents a comparison between the measured
live loads and taking K2 at the structure’s dead load level (0 kN
normalised vertical bucket stiffnesses for each bucket in terms of
on Figure 16b), the increase in effective stress in the soil mass
number of occurrences (probability density). It is clear that
back calculated from the stiffness variation is between 10% -
Bucket AB and Bucket BC have a similar mean normalised
20% of the applied load. In other words, is between about 0.1
vertical stiffness which is lower than the mean normalised
and 0.2. The remainder of the increased load may be
vertical stiffness of Bucket AC. This provides further evidence
dissipated/resisted by the soil-skirt friction resistance or pore
that the load level has a direct influence on the suction bucket
pressures.
vertical stiffness response with lower compressive loads leading
Figure 16b shows that the trend of increasing stiffness with
to lower vertical stiffness. As Bucket AB and Bucket BC
increasing load does not apply for compressive live loads, where
experience lower compressive forces under the prevailing wind
it is observed that the stiffness is relatively constant and tends to
direction (wind emanating from heading 200° to 300°) as shown
reduce after a threshold value of approximately -2500 kN.
in Figure 13, it would be expected that these buckets would
exhibit lower vertical stiffnesses, as observed in Figure 18.
Figure 17a shows the suction bucket load-displacement response
Conversely, Bucket AC experiences higher compressive forces
which is calculated for each data point by dividing the mean
under these loading conditions and would therefore be expected
applied load by the measured secant stiffness values. It has been
to exhibit a higher vertical stiffness response, which is also
assumed that the highest tensile live load level corresponds to an
observed in Figure 18. In general, the mean normalised vertical
absolute load of 0 kN. At high tensile live loads, the effect of soil
stiffness of Bucket AC is approximately 10% greater than the
stiffness on the load-displacement curve becomes apparent as the
mean normalised vertical stiffnesses of Bucket AB and Bucket
tangent slope of the curve reduces with reducing load. At
BC. This is consistent with the observations from Figure 16a,
compressive live load levels which are more than the deadweight
where it is shown that the vertical stiffness is approximately
of the structure, an initial linear load-displacement response is
10% – 20% higher for a bucket under higher compressive load
observed. At higher displacements, a non-linear response is
levels.
observed. The overall shape of the load-displacement curve is
Notably, a second normalised vertical stiffness ‘peak’ is
similar to that observed from model tests reported by Byrne &
observed for Bucket BC (normalised vertical stiffness of
Houlsby (2002). Of particular note, the ‘banana shape’ of the
approximately 0.67) which may reflect the bucket behaviour
load-displacement curve as the load is reduced towards 0 kN net
under higher compressive loads (when the winds emanate from
load is replicated in the BKR01 SBJ data. This is shown clearly
a more easterly heading).
in
Figure 17b where an indicative linear fit (in red) shows the data
following a non-linear response as live tensile loads increase.
Increasing
compressive
live loads
Increasing
tensile live
loads
Figure 18. Soil-skirt vertical stiffness for each bucket
LOAD TRANSFER THROUGH THE SKIRT
The strain gauges installed along the bucket skirt (see Figure 5)
can be used to measure forces in the skirt and therefore allow for
the calculation of the shaft resistance over a small section of skirt
length. The strain gauge readings can be used to estimate the
shaft resistance during loading and the distribution of the load
between the skirt and the lid during loading.
7.1 Unit shaft resistance

The shaft resistance was assessed by comparing the forces in the
Figure 17. a) Load-displacement response of Bucket BC b) load- clean section with the change in force between the two strain
displacement response at maximum tensile live loads in which response gauge levels. By taking the 10 minute average skirt force for each
becomes non-linear
strain gauge for all time points, the drained response of the
bucket skirt can be assessed.
Figure 19 and Figure 20 show examples of the measured
vertical force at the clean section against the calculated force in
71
the strain gauges for Bucket BC and Bucket AC. A mean of the approximately 16 kPa although this is considered unrealistically
data cloud is calculated for different load levels (blue line) and a high based on the general trend observed.
best fit line (red dashed line) was fitted to this dataset. A slope
of 1 (black dashed line) would indicate that the force measured
at the clean section was equal to the force calculated at the strain
gauge. A linear fit was generally observed up to load levels
approaching the maximum compression load at which point the
gradient is observed to reduce and exhibits nonlinear behaviour.
The ‘slope’ of the best fit line for each of the 9 strain gauges
has been calculated and plotted against depth below lid level on
Figure 21. Each data point shown on Figure 21 therefore
represents the average proportion of load transferred from the
clean section to each strain gauge. For example, for the strain
gauge at 1 m below lid level on Bucket AC (0°) shown in Figure
20, an average slope of 0.88 was calculated, indicating that on
average, the force measured at the strain gauge was 88% of the
force measured at the clean section. As expected, Figure 21
shows that the force transferred to the strain gauges generally
reduces with depth as force is transferred to the soil via friction
between the soil and the skirt wall (shaft resistance).
Figure 20. Bucket AC upper strain gauge response
By comparing the calculated slopes (load transferred to the
strain gauges compared to the clean section) at the two different
strain gauge levels (Figure 21), the unit shaft resistance of the 0
skirt between the two strain gauge levels can be calculated. Using Lower bound
the upper bound envelope from Figure 21 as an example, the -1000
envelope
upper strain gauge level slope is 0.95 and the lower strain gauge
Distance Below Bucket Lid [mm]
level slope is 0.78. On average, this indicates that at the upper -2000
Upper bound
strain gauge level, 95% of the force measured in the clean section Mean
-3000 envelope
is measured at the strain gauge, but this value reduces to 78% at
the lower strain gauge level. The force dissipated between the -4000
two strain gauge levels is therefore equal to the difference
between the two levels, in this case, 17% of the force measured -5000
in the clean section. Bucket BC 0°
-6000 Bucket BC 90°
Bucket AC 0°
-7000
-8000
0 0.2 0.4 0.6 0.8 1 1.2
slope [-]
Figure 21. Slopes of skirt force versus vertical input force for all skirt
strain gauges
For reference, the mean slope for the upper strain gauge level
(at approximately 1m below the lid level) is 0.79 and the mean
slope for the lower strain gauge level (at approximately 2 m
below the lid level) is 0.64.
Table 3. Back calculated drained unit shaft resistances between strain
gauge levels
Measurement value Unit shaft resistance (kPa)
Lower envelope 5.9
Upper envelope 6.2
Figure 19. Bucket BC lower strain gauge response
Mean 6.3
As the surface area of the bucket skirt between the two strain
gauge levels is approximately 69 m2, a unit shaft resistance can
be calculated by dividing the load dissipated between the two 7.2 Comparison of measured values against calculated values
strain gauge levels by the relevant bucket surface area. Assuming
a vertical force of 2500 kN measured at the clean section, which Assuming the fully drained conditions and the shaft capacity
approximately corresponds to the predicted drained resistance of calculation method described by Tomlinson & Woodward
the suction bucket skirt, Table 3 provides an estimate of the unit (2008) for cohesionless soils, the unit shaft resistance for the
shaft resistance based on the lower, mean and upper envelope of BKR01 conditions can be estimated. Both sets of strain gauges
the measurements of forces at the strain gauges. Based on the are located within the shallowest soil unit identified (Table 2).
back calculated unit shaft resistance values using the strain gauge Although this unit has a relatively high peak friction angle and
measurements (Table 3), the unit shaft resistance is relative density, the unit shaft resistance has been calculated
approximately 6 kPa. In the extreme case, taking readings from using standard design approaches which takes the interface
the Bucket BC 0° upper strain gauge and the Bucket AC 0° lower friction angle into account. Assuming an interface friction angle
strain gauge, the unit shaft resistance is found to be of 29° for sand with a mean particle size of 0.2 mm (Jardine et
72
al, 2005), a K0 value of 0.8 (API, 2007) and an even distribution

of shaft resistance between the inside and outside of the skirt, a
drained unit shaft resistance of 6.9 kPa could be expected over
the depth range being investigated. This is similar to that
measured at BKR01 and indicates that drained unit shaft
resistance values are likely to be appropriate on average. As the
strain gauge measurements are taken close to the ground surface,
OCR effects were not explicitly taken into account, although
these have been used to justify a K0 value considerably higher
than the Jaky (1944) formulation.
7.3 Load distribution estimate using strain gauge

measurements
Whilst the strain gauge measurements in Figure 21 show some
degree of scatter, these provide a useful estimate of the general
load distribution between the skirt and bearing under the suction
bucket lid. Although extrapolation of the trends below the lower
strain gauges is considered inappropriate, extrapolation of the
data to the bucket lid is considered acceptable as the strain gauges
are located relatively close to the lid. By extrapolating the upper
and lower bound envelopes to the lid level, the data in Figure 21
indicates that the load transferred from the clean section to the
skirt walls may be between approximately 80% (lower envelope)
and 100% (upper envelope), the latter implying that the load from
the structure is completely transferred to the skirt.
These values are generalised across the entire load spectrum Figure 22. Observations from an average operational day
and can therefore only be considered approximate and load
independent. As the strain gauge measurements are based on 10 8.2 Observations from a stormy period
minute average values, individual load cycles are not considered.
Similar patterns were observed for a ‘stormy’ day. During
electricity production, the stiffness is relatively consistent and
during periods of low production the stiffness becomes more
COMPARISON BETWEEN A ‘CALM’ PERIOD AND scattered and typically appears to decrease. The stiffness
A STORMY PERIOD recovers to its original value when load is reapplied (during
operation) and no stiffness degradation is observed. In addition,
Observations for a ‘calm’ period with wind speeds at or below no long term excess pore water pressures were generated during
the rated wind speed of the WTG ( the storm. This result was expected as the long-term SBJ
Figure 22) are compared to observations from a stormy period structural frequency has shown no measurable deviation since
(Figure 23) to assess if there is any variation in behaviour during first power.
these two periods. In
Figure 22 and Figure 23, the 10 minute average wind speed, wave
height, pore pressure beneath the bucket lid and vertical bucket
stiffness are plotted for several days with periods of operation
and non-operation. The green shading indicates times of high
electricity production and the unshaded areas indicate times of
low electricity production.
8.1 Observations from a ‘calm’ period

During electricity production on a standard operational day
where wind speeds at or below the rated wind speed of the WTG,
the stiffness is relatively consistent. The stiffness becomes more
scattered and appears to reduce during periods of low production,
which is likely due to the lower loads leading to a higher signal
to noise ratio. The stiffness recovers to approximately its original
value when load is re-applied (during operation) and no stiffness
degradation is observed.
Figure 23. Observations from a stormy day
73
The correlation between vertical load and excess pore water pressure will be generated and the vertical input force from the
pressure beneath the bucket lid is examined for a period during jacket leg will be taken by the soil-bucket interaction whereby
the peak of the storm (Figure 24). The dynamic component of the the bucket is said to be responding in a drained manner.
vertical pressure (vertical load minus mean load divided by Conversely, if the loading on the bucket lid is sufficiently fast,
bucket area) on the buckets is dominated by wave loads during the water beneath the bucket lid will not have sufficient time to
the peak of the storm. The pore water pressure within the bucket drain. As the volume of water is constant and assumed
(excess pore water pressure measured beneath the lid and at the incompressible, excess pore pressure will be generated creating
inside of the skirt tip) takes a significant portion of the vertical a pressure difference across the bucket lid leading to undrained
pressure with a minimal phase shift. For fully undrained conditions and significant additional capacity.
conditions, the excess pore water pressure response beneath the A key assumption of suction bucket design for non-cohesive
bucket lid and at the skirt tip (inside the bucket) should exactly materials is that undrained conditions exist if the loading is
match the vertical pressure applied to the bucket. However, sufficiently fast. For pull out capacity, this has been extensively
Figure 24 shows that this is not always the case suggesting that investigated and shown to be significant (Tjelta, 1994; Houlsby
some vertical pressure is unaccounted for, which as described in et al., 2005; Achmus & Thieken, 2014; Thieken et al. 2014).
Section 7, may be due to the lid transferring some load to the In this section, the force transferred from the jacket leg
bucket skirt or a number of other mechanisms. The excess pore (through the clean section) to the pore pressure inside the suction
water pressure is not completely be transferred between the lid bucket is investigated for the entire data set to determine under
and skirt tip levels. This might be due to the interaction between what loading conditions the bucket behaves in a drained or
the bucket grout and skirt or due to the presence of the silt layer undrained manner. The pore pressure beneath the lid acts on the
(described in Section 4) preventing the direct transfer of the bucket lid to resist the vertical input force from the jacket leg,
excess pore water pressure down the skirt. however it is difficult to distinguish the portion of load acting
In addition, observations from Figure 24 indicate that the through the lid due to the rigidity of the lid.
frequency and duration of the load has an effect on the excess
pore water pressure response. For example, the response for the
peak load at 29 s and 58 s can be compared. For the peak at 29 s, 9.1 Typical examples of undrained responses
which applies a relatively sinusoidal force over approximately An example of bucket input force (Fz) compared to the excess
5 s, the response of the excess porewater pressure is directly pore water pressure (converted back to a force, denoted FWP)
linked to the increase in vertical force giving an almost a fully beneath the lid during a period of well correlated short term
undrained response. For the peak at 58s, where the peak wave undrained behaviour is shown in Figure 25. The water beneath
load is applied over approximately 1s and is preceded by a time the lid takes a significant portion of the high frequency input load
of chaotic force application, the response of the excess porewater with a minimal phase shift. The mean load and the mean excess
pressure is not linked during application of the peak load. This pore pressure are also approximately equal (near 0 kN). The force
indicates that the frequency and duration of the load clearly not resisted by the pore pressure is assumed to be distributed to
influences the excess pore water pressure response. other resistance mechanisms such as skirt friction, skirt tip
resistance, lid plug contact and damping.
Figure 24. Time history of the vertical pressure and excess pore water
pressure (denoted PWP) beneath the lid and at the inside of the skirt tip
To calculate the average amount of vertical pressure carried

by the pore pressure for a select 10 minute period at the peak of
the storm on 30 November, the area under the curve for each
component was calculated using numerical integration. Over the Figure 25. Example of bucket responding to loading in an undrained
10 minute period, the pore pressure beneath the lid was estimated manner
to carry 64% of the vertical pressure and the pore pressure at the
skirt tip was estimated to carry 42% of the vertical pressure. Examining the time history in Figure 25, the cyclic loading
These percentages would be considerably higher if the data was frequency acting upon the bucket is approximately 0.3 Hz which
filtered to only include loads with a longer time period, such as is close to the 1P frequency of the structure
those associated with large wave loads as observed at 29s in (Harte et al., 2012), indicating that the loads presented in Figure
Figure 24. 25 are dominated by the wind loads acting on the RNA (hence
the zero mean) and not the wave loads which would be expected
to have a lower frequency.
DRAINAGE RESPONSE OF THE SUCTION BUCKET Figure 26 shows another example of bucket input force
If the variation in loading is sufficiently slow, the water beneath compared to the excess pore water pressure beneath the lid
the bucket lid will have sufficient time to drain and the pressure during a period with a non-zero mean input load. In Figure 26,
inside and outside the bucket will equalise. Thus, no excess pore the excess pore pressure reacts to each individual cycle in an
undrained manner (the excess pore water pressure response and
74
the load response are aligned but at different absolute

magnitudes) and the long term (steady state conditions) mean is 1
∙ ∙ (5)
zero. However, some excess pore water pressure is generated τ
between 150 s and 250 s as the mean load shifts from a Where τ is a time coefficient and FWP is the resistance force due
normalised mean load (Fz/Fz,ref) of 0.5 to -0.5 (also normalised to to the generation of excess pore water pressure. The time scale
the mean dead load). Steady state pore pressure conditions are on which the bucket’s FWP transitions between drained and
only re-established after approximately 100s. Thus, the drainage undrained response can be gauged by plotting the transfer
period for this example is approximately 1.5 minutes. function between the input load and the bucket’s FWP as follows:
∙
1 (6)
The spectrum of the input load and the bucket’s FWP is

estimated using a windowed fast Fourier transform (FFT) with a
Hanning weighting function (Welch, 1967). The windows are
taken with a 50% overlap and the resulting FFTs are averaged.
9.2.2 Drainage analysis for one day’s data set

The data for a single day is given in
Figure 28 and Figure 29. From
Figure 28, it can be seen that the spectra of the vertical input load
and the pore pressure resistance force , are similar at
high frequencies but diverge as the frequency drops. At short
periods, lower than about 2 minutes, the excess pore water
pressure resists a consistent proportion of the input load, showing
the undrained response of the bucket. While at longer periods,
Figure 26. Example of bucket responding to loading in a drained
manner
the ratio between the spectra decays.
The transition between drained and undrained response is
9.2 Drainage analysis highlighted in grey. At very long periods, greater than about
20 minutes, the pressure force does not resist much of the input
In order to estimate the suction bucket drainage response at load, showing the drained response of the bucket. For periods
different times during the monitored period, a more detailed around 5 s to 15 s (highlighted with blue shading), there appears
analysis of the drainage conditions was required, which can be to be a peak in the transfer function. As seen in the wave height
applied across the entire data set. To do this, the input force and spectrum (denoted ‘eta’ in
excess pore water pressure (beneath the bucket lid) were Figure 28), this region contains wave energy. As the waves pass
converted from the time domain to the frequency domain and over the bucket, the static head pressure outside the bucket
their coherence examined. changes and the coherence between input load and excess pore
water pressure decreases. This indicates that energy is being
9.2.1 Drainage analysis methodology added to the excess pore water pressure response in this
For the undrained case, where there is assumed to be no flow in frequency range. The increased coherences between the excess
or out of the bucket, the pore pressure resistance force (FWP) will pore water pressure and the wave height also indicates that
take some proportion (K) of the bucket input load (Fz) as follows: energy is being added to the excess pore water pressure spectrum.
Results from this frequency are considered unrealistic.
, ∙ (4)
Therefore, when trying to fit a transfer function between the input
Where the first-time derivative is denoted by the dot. For the force and the excess pore water pressure, this portion of the
bucket to drain, water within the bucket must flow through the transfer function should be excluded. This is shown in Figure 29
seabed into or out of the bucket, as idealised in Figure 27. where it is clear that the fitted function does not take this
frequency range into account.
Figure 27. Model of flow around a bucket during loading
The proportion of load (K, also denoted as ‘Gain’ in Section

9.3) and drainage time affect the rate of resistance. This
resistance is due to the generation of excess pore water pressure
which can be combined in the following differential equation:
75
available data to quantify time scales at which the transition

between the drained and undrained response occurs. A histogram
of the gain (K) and time constant (Eq. 5) for all the available data
for Buckets AB and BC is given in
Figure 30.
Figure 30. Model parameter fit results distribution
Figure 30 shows that the Buckets AB and BC exhibit different

drainage characteristics, with the drainage at each bucket
expected to be affected by the site specific soil conditions at each
bucket and the bucket loading history. On average, Bucket AB
experiences higher compressive loads than Bucket BC which
may explain the higher gain observed. During undrained loading
the gain for Buckets AB and BC varies from approximately 0.45-
0.65, thus on average the pore pressure is carrying between 45-
Figure 28. Spectrum of input load and excess pore water pressure 65% of the load transferred through the bucket clean section. A
beneath the lid for the analysed day similar result was found for the November 30th 10 minute storm
time history assessed in Section 8.2.
Examining the entire data set for both buckets, the
determined average thresholds between the drained and
undrained response are given in Table 4.
Table 4. Thresholds between drained and undrained regions
Drained Partially Drained Undrained
T> 46 min >T> 2.5 min >T
The undrained region is defined as the regions for which the

transfer function is greater than 90% of the gain (K) and
conversely the drained region is defined as the region for which
the transfer function is less than 10% of the gain. The partially
drained region is taken as the region where the transfer function
is between the two thresholds. These results show that the
buckets are undrained at relatively low frequencies compared to
what is typically assumed in design of suction buckets in non-
cohesive strata. This is most likely due to the site specific soil
conditions, in particular, the silt layer described in Section 4.
CONCLUSION
Figure 29. Best fit transfer function for the analysed day
The BKR01 SBJ, installed in 2014 in collaboration with Carbon
Trust OWA program, is the first SBJ to support a WTG. An
The resulting model parameters for this period were K=0.516 extensive monitoring system installed on the SBJ has provided
and τ=45 s which is shown to fit the transfer function well an insight into the behaviour of the suction buckets during
(Figure 29). However, it should be noted that these values are different loading conditions.
only for one day of data. The vertical stiffness response of the buckets was shown to
be dependent on the wind direction and the load level. The
assumption that the average total bucket vertical stiffness
9.3 Drainage analysis for the entire data set response can be represented by a system of springs in series was
confirmed, verifying an important design assumption. It was also
The analysis described in Section 9.2 was repeated for all the
76
shown that the vertical stiffness response is influenced by mean

load level with a distinct vertical stiffness increase of between 10 Ehrmann, A., Penner, N., Gebhardt, C.G. and Rolfes, R., 2016. Offshore
and 20% when the mean load was increased. The suction bucket Support Structures with Suction Buckets: Parameter Fitting of a
Simplified Foundation Model. In The 26th International Ocean and
response showed a typical non-linear response at low Polar Engineering Conference. International Society of Offshore
compressive load levels (high tensile live loads) confirming and Polar Engineers. Rhodes, Greece, June 26–July 2 2016.
previous research in this area. Erbrich, C.T. and Tjelta, T.I., 1999. Installation of bucket foundations
The strain gauge readings showed that the measured unit and suction caissons in sand-geotechnical performance. Offshore
shaft resistance was within the expected range when compared Technology Conference. Houston, Texas, USA, May 3-6.
with traditional shaft capacity hand calculations. The readings Harte, M., Basu, B. and Nielsen, S.R., 2012. Dynamic analysis of wind
also indicated that the load transferred from the clean section to turbines including soil-structure interaction. Engineering Structures,
the skirt walls may be between approximately 80% and 100%, 45, pp.509-518.
Houlsby, G.T., Kelly, R.B. and Byrne, B.W., 2005. The tensile capacity
for drained loading conditions. of suction caissons in sand under rapid loading. In Proceedings of
After comparing the results from a ‘calm’ period (wind the international symposium on frontiers in offshore geomechanics,
speeds at or below the rated wind speed of the WTG) with a Perth (pp. 405-410).
‘stormy’ period, an investigation of the pore pressure response to Houlsby, G. T, 2014. “Interactions in Offshore Foundation Design.” 54th
loading was presented. Transforming the data to the frequency Rankine Lecture. Imperial College London, UK. 19 March 2014.
domain, it was found that on average, loading frequencies faster Jaky, J., 1944. The coefficient of earth pressure at rest. Journal of the
than 2.5 mins led to undrained behaviour, loading frequencies Society of Hungarian Architects and Engineers, 78(22), pp.355-358.
slower than 46 mins were fully drained and loading frequencies Jardine, R., Chow, F., Overy, R. and Standing, J., 2005. ICP design
methods for driven piles in sands and clays. London: Thomas
in between were partially drained. It was also found that the Telford.
drainage characteristics varied from Bucket AB to Bucket BC Sparrevik, P. & Strout, J. M., 2015. Novel monitoring solutions solving
most likely due to the site specific loading conditions and ground geotechnical problems and offshore installation challenges.
conditions at each bucket. In Proceedings of the international symposium on frontiers in
Overall, analysis of the data from the comprehensive offshore geomechanics, Oslo (pp. 319-324).
monitoring system installed on the BKR01 SBJ has provided a Suryasentana S.K., Byrne B.W., Burd H.J., Shonberg A., 2017.
significant insight to the behaviour of suction bucket jackets Simplified model for the stiffness of suction caisson foundations
subjected to wind and wave loading in the North Sea. The in-situ under 6 dof loading. Offshore Site Investigation & Geotechnics
(OSIG) International Conference, London, UK, Sept 12-14.
observations from a world-first full scale monitoring system are Sturm, H. 2017. Design aspects of suction Caissons for offshore wind
likely to inform future offshore windfarm developments where turbine foundations. In Proceedings of TC209 Workshop
SBJs are utilised as the foundation solution. (Foundation design of offshore wind structures): 19th International
Conference on Soil Mechanics and Geotechnical Engineering,
Seoul, South Korea, 20 September.
Thieken, K., Achmus, M. and Schröder, C., 2014. On the behavior of
ACKNOWLEDGEMENTS suction buckets in sand under tensile loads. Computers and
Geotechnics, 60, pp.88-100.
The authors would like to acknowledge the many contributions Tjelta, T.L., 1994. Geotechnical aspects of bucket foundations replacing
made by their colleagues at DONG Energy especially from Mrs piles for the Europipe 16/11-E jacket. Offshore Technology
Annelies Vanstraelen and Dr. J Schupp. The authors would also Conference. Houston, Texas, USA, May 2-5.
like to acknowledge the contributions made by Mr Stephen Tjelta, T.I., 1995. Geotechnical experience from the installation of the
Suryasentana from the University of Oxford. Europipe jacket with bucket foundations. Offshore Technology
The BKR01 SBJ project was undertaken in collaboration Conference. Houston, Texas, USA, May 1-4.
with the Carbon Trust Offshore Wind Accelerator program and Tomlinson, M.J. and Woodward, J. 2008. Pile design & construction
practice. 5th ed. London: Taylor and Francis.
DONG Energy acknowledges the contributions made to the
Welch, P., 1967. The use of fast Fourier transform for the estimation of
project by the entire Carbon Trust team. power spectra: a method based on time averaging over short,
modified periodograms. IEEE Transactions on audio and
electroacoustics, 15(2), pp.70-73.
REFERENCES Wroth, C.P. and Houlsby, G.T., 1985. Soil mechanics-property
characterization and analysis procedures. Proceedings of the 11th
Achmus, M., and Thieken, K., 2014. Numerical Simulation of the Tensile
International Conference on Soil Mechanics and Foundation
Resistance of Suction Buckets in Sand. Journal of Ocean and Wind
Engineering. San Francisco, USA. Vol. 1.
Energy, 1(4), pp. 231–239.
American Petroleum Institute, 2007. API Recommended Practice for
Planning, Designing and Constructing Fixed Offshore Platforms –
Working Stress Design, API RP 2A WSD. API Publishing Services,
Washington DC, USA.
Aghakouchak, A., 2015. Advanced laboratory studies to explore the axial
cyclic behaviour of driven piles. PhD thesis, Imperial College
London, London, UK.
Bye, A., Erbrich, C., Rognlien, B. and Tjelta, T.I., 1995. Geotechnical
design of bucket foundations. Offshore Technology Conference.
Houston, Texas, USA, May 1-4.
Byrne, B.W. and Houlsby, G.T., 2002. Experimental investigations of
response of suction caissons to transient vertical loading. Journal of
Geotechnical and Geoenvironmental Engineering, 128(11), pp.926-
939.
Dekker, M.J., 2014. The Modelling of Suction Caisson Foundations for
Multi-Footed Structures. Norwegian University of Science and
Technology.
Doherty, J.P., Houlsby, G.T. and Deeks, A.J., 2005. Stiffness of flexible
caisson foundations embedded in nonhomogeneous elastic
soil. Journal of geotechnical and geoenvironmental
engineering, 131(12), pp.1498-1508.
77

Proceedings of TC209 Workshop - Www-Version

Uploaded by

Copyright:

Available Formats

Proceedings of TC209 Workshop - Www-Version

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

Proceedings of TC209 Workshop - Www-Version

Uploaded by

Copyright:

Available Formats

What are some of the key foundation design challenges for offshore wind structures?

What are some of the key foundation design challenges for offshore wind structures?

What factors are important for marine site characterization to support foundation design?

What factors are important for marine site characterization to support foundation design?

ICSMGE Workshop Proceedings_Cover_crop_marks.

pdf 1 2017-09-11 15:45:04

ISSMGE Technical Committee TC 209

Proceedings of TC 209 workshop

Foundation Design of Offshore Wind Structures

Edited by: Yunsup Shin

ISSMGE Technical Committee TC 209

Proceedings of TC 209 workshop

Edited by: Yunsup Shin

Geotechnical perspective on offshore wind plan, strategy, projects and research in 13

Design aspects for monopile foundations 35

Design Aspects of Suction Caissons for Offshore Wind Turbine Foundations 45

Yun Wook Choo

Jin Young Kim & Changho Choi

1 INTRODUCTION country, interest in offshore wind power has been increasing,

Figure 4. Mean wind speed measured at HeMOSU-1 met mast tower

Figure 5. Annual wind speed for 30 years estimated by MCP method

Figure 2. Main wind direction distribution (modified from www.

Fine sand with silt Fine sand with silt

Fine sand with silt

Weathered rock/2 m Soft rock

Figure 6. Boring log at Nos. 7 and 8 of southwest sea offshore wind

The Jeju Tamra offshore wind Farm located on the

Southwest Korean Offshore Jacket,

Jeju Island Korea Southern Under

Jeju Island KEPCO

Jeju Island Halla Wind Energy Under

Busan Korea Southern Under

SK Construction Under Figure 10. Jeju Tamra offshore wind farm.

3.2 Jeju Woljeong

Figure 11. Jeju Tamra offshore wind farm monitoring center.

As of July 2017, the first jacket substructure is under

Figure 12. Gunsan suction foundation test site.

Figure 15. Location of the Southwest Sea wind farm

Table 2. Construction plan of Southwest Sea wind farm.

Figure 13. Panoramic view of tripod suction foundation. Capacity

Substructure Jacket, Suction NA NA

Figure 14. Verticality control using suction pump.

3.5 Southwest Sea wind farm

4.2 Jacket substructure

4.4 Monopod concrete suction foundation

1 INTRODUCTION. • An evolutionary ground model to characterise site

Figure 2. The Geowing gradiometer; consisting of five fixed

2.3 Metocean Data

2.4 Environmental study

conditions around the piles). This information is mostly

Figure 8. Example of subsurface features as identified during desk

4.3 Geophysical Investigation

Figure 9. Example interpretation of geological soil units from a

Once the detailed geophysical survey is complete the results

4.4 Reconnaissance Geotechnical Investigation

In areas of high current, anchored vessels are a good

Figure 13. The SEADEVILTM hybrid vessel-seafloor drilling (Looijen

Figure 14. Example measured and interpreted CPT data

high quality coring can be challenging as there is a large