Coupled Effects Between Damage and Permeability With A View To Discrete Modelling

CONCREEP 10 49
Coupled Effects between Damage and Permeability with a View to Discrete

Modelling
G. Pijaudier-Cabot; F. Khaddour; D. Grégoire; and L. Ecay
Laboratoire de Fluides Complexes et leurs Réservoirs, UMR 5150 UPPA – CNRS –

Total, Allée du Parc Montaury, F-64600 Anglet, France. E-mail:
Gilles.Pijaudier-Cabot@univ-Pau.fr
Abstract
The purpose of this work is to achieve a better understanding of the

relationship between mechanical damage and the transport properties of cementitious
materials. In the literature, analyses are usually restricted to intrinsic permeability of
the material and the evolution of the apparent permeability with respect to the
pressure gradient and to the nature of the fluid considered are left aside. A new model
capable to provide the apparent permeability of a porous material to gas, directly
from the pore size distribution and from the properties of the gas is discussed.
Comparisons with experimental data on mortar specimens show that the model can
reproduce the intrinsic permeability and its evolution when the material is subjected
to mechanical damage, provided the pore size distributions are available. For a given
pore size distribution, the evolution of the apparent permeability is also provided and
test data with several types of gases compare quite well with the model.
INTRODUCTION
In this work, we aim at a better understanding of the relationship between damage,

and the transport properties in cementitious materials, namely permeation of gas. This
is typically of utmost importance in the long-term assessment of tightness of vessels
and containment facilities.
Many authors have looked for predictive models of porous media permeability based
on the pore size distribution of the material. Pioneering work by Kozeny [1] related
the permeability to the porosity, to the tortuosity and to an average pore size and later
on Katz & Thompson [2] predicted the permeability from the electrical conductivity
of the porous material through a critical pores radius. There are also several studies in
which the pore size distribution was modeled, e.g. with a bimodal [3] or multimodal
distribution [4], and then entered into some capillary bundle model. Recently, models
for the calculation of the intrinsic permeability derived directly from the experimental
PSD, without any idealization have been also discussed [5]. Here, we shall follow this
approach which requires (i) to transform the pore size distribution into a hydraulic
network, and (ii) to solve the equations governing fluid flow once the hydraulic
network has been set.
© ASCE
CONCREEP 10 50
As far as the porous network in the material is concerned, we shall use here a
description that is consistent with the Pore Size Distribution (PSD) characterization
technique used for this study, mercury intrusion. The pore network is not defined in
the material space, e.g. with a distribution of pores connected to each other following
a topological distribution characterised e.g. by microtomography, but rather following
a stochastic hierarchical bundle model. The technique yields a pore network, which
mimics the porous space measured experimentally.
Usually, analyses are restricted to the intrinsic permeability of the material and to the
evolution of the apparent permeability, with respect to the pressure gradient. Here,
the model is capable to provide estimates of the apparent permeability directly from
the pore size distribution and from the properties of the fluid. The intrinsic
permeability and the evolution of the apparent permeability with mean pressure are
provided by combining Darcy’s law [6], which represents the flow in the porous
media at the macroscale and a combination of Poiseuille [7] and Knudsen laws [8],
which represent respectively viscous flow and diffusion flow at the microscale. For
typical gases, the contributions of laminar flow and of Knudsen diffusion are of the
same order for pores of diameter in the range of a few hundred of nanometers [9].
Comparisons with experimental data on mortar specimens show that the model can
reproduce the intrinsic permeability and its evolution when the material is subjected
to mechanical damage, provided the pore size distributions are available at each level
of damage. For a given pore size distribution, the evolution of the apparent
permeability is also obtained, for several gases, and compares quite well with
experiments. The model provides also the condition of permeation of mixture of two
gases. Knudsen effect being different for each species, selective permeation is
obtained and preliminary comparisons with experiments are provided.
STOCHASTIC HIERARCHICAL BUNDLE MODEL
We consider here that the microstructure of the porous material consists in cylindrical
pores of different lengths and diameters. Assuming that the porous space has been
characterized in term of PSD by mercury intrusion technique, we may link, for each
pore size, the pore length to the pore volume fraction.
A simple assembling technique consists in distributing all the pores of a given size (or
interval of size) in parallel and in a regular manner, assuming that all pores have an
identical length corresponding to the cubic root of the total porous volume times the
tortuosity. Predictions of permeability based on such parallel assembly of pores have
been discussed in [5], and it was observed that tuning parameters were required in
order to obtain relevant results in term of apparent and intrinsic permeability. Large
pores drive the permeability and since the model assumes that they cannot be
connected to smaller ones, their effect on permeability has to be moderated.
We thus developed a different approach where the different pores sizes are organized
randomly within the material and are randomly connected to each other, while
© ASCE
CONCREEP 10 51
satisfying the PSD description under the condition that a given pore may only be
connected to a smaller one.
Let Vt be the total porous volume, characterized by mercury intrusion, which may be
organized according to the PSD into a series of volumes ܸ௣௜ of mean diameter ݀௜ .
Hence, we discretize the PSD into a finite set of pore diameters, ݅ ൌ ͳǡʹǡ ǥ ݊. We
may now relate, for each pore size, the total pore length ‫ܮ‬௜ to the pore diameter ݀௜
and to the pore volume fraction ܸ௣௜ assuming that pores are cylindrical:
ͶǤ ܸ௣௜
‫ܮ‬௜ ൌ (1)
ߨǤ ݀௜ ଶ
This pore length is then cut into small segments of random length ¨Li (see Figure
1.a). Their assembly is performed from larger pores to smaller pores. Each pore
segment either creates a new assembling site (on which subsequent segment will be
connected) or is connected to an existing one of larger or equal size on an existing
assembling site (Figure 1.b). We then define the average pore length ‫ܮ‬௘ and a critical
length ‫ܮ‬௖ as
‫ܮ‬௘ ൌ యඥܸ௧ , ‫ܮ‬௖ ൌ ܶǤ ‫ܮ‬௘ T = tortuosity (2)
As soon as the sum of all pore segment lengths reaches the critical length on one
assembling site, further segments cannot be added anymore and the permeability of
the resulting capillary is estimated and added to the global permeability.
(a). Discretization of pores (b). Random assembly

Fig.1. Random network assembly of pores.
This random process is performed until the overall permeability does not change by
adding a new assembled capillary (with some specified tolerance). This criterion is
used to speed up the process: there is a huge amount of pore segments of very small
diameter but they do not contribute so much to the global permeability, which is
mostly driven by larger pores. As soon as this criterion is reached, the remaining
porosity is organized according to a parallel capillary bundle, each capillary being of
the same diameter. Their contribution is also taken into account in the global
permeability.
© ASCE
CONCREEP 10 52
Some adjustments need to be discussed with respect to the mercury intrusion

technique from which the PSD is obtained. During mercury intrusion, two different
phases are performed. Firstly, mercury is introduced at low pressure (from 0 MPa to
0.2 MPa) and large diameter pores are characterized. Secondly, mercury is introduced
at high pressure (from 0.2 MPa to 200 MPa) in order to drive mercury into smaller
pores. Unfortunately, the data obtained at low pressure are quite noisy and the
distribution of sizes for very large pores is not characterized accurately. Since large
pores drive the permeability, this may be very problematic in order to predict the
permeability accurately. Usually, the solution to circumvent this artifact consists in
removing the low-pressure information. Here we adopt another, less arbitrary
technique, motivated by the fact that upon damage, pores (or cracks) of large aperture
might contribute significantly to the permeability.
Figure 2.a presents a typical PSD response of a cementitious material (mortar). We

observe both the low-pressure signal and the high-pressure one (in the range of 100
micrometers). In order to keep the relevant low-pressure information and at the same
time to remove the noisy artifacts, we implement a specific filtering of the low-
pressure information: let us consider the pores of size corresponding to the low-
pressure measurements only. These pores are discretized randomly into segments of
equal size. We now count the number of segments corresponding to these pore sizes.
Figure 2.b shows this number for several random realizations. We observe that the
number of segments lies below 10000 for the material that is typically considered in
this study. Instead of removing these pores, we assume that each segment below this
threshold of 10000 is forced to create a new assembling site. Consequently pores of
such sizes cannot percolate throughout the material, unless the total number of
segments is over the threshold.
(a) (b)
Fig. 2. Method of eliminating of the noise at low pressure for an undamaged

sample: (a) pore size distribution corresponding to the low and high pressures;
(b) statistical study of the low frequency for 100 random realizations.
In the hierarchical assembly forming the porous network, the random function that
decides if the current pore segment should be connected to an existing site or will
create a new site is weighted by the following exponential function:
© ASCE
CONCREEP 10 53
‫ݔ‬െ‫ܤ‬
݂ሺ‫ݔ‬ሻ ൌ ሺെ ሻ (3)
‫ܣ‬
where x is the index of the current pore segment dealt with, ‫ ܣ‬and ‫ ܤ‬are parameters.
‫ ܤ‬controls the pore diameter from which new pore segments may be connected to
existing assembling sites. If the index x is lower than B, the current pore segment is
forced to create a new site. From the PSD in (Figure 2.a), the parameter B is chosen
in such a way it assures that the first 10000 low-pressure pore segments are forced to
create a new assembling site. Upon damage, if meaningful information appear in the
low-pressure part of the PSD, the number of pore segments characterized under low
pressure will increase and then they will be taken into account when x>B.
Parameter A controls the pore diameter from which new pore increments have almost
no chance to create a new assembling site. Therefore, it controls indirectly the time at
which the stochastic assembling of pore segments will start to be in parallel
deterministically. Since the transition between a random assembly and a deterministic
assembly is also driven by a criterion on the global permeability, the influence of
parameter A is not determinant if this last criterion is met first (which is the case in
our calculations).
FLUID FLOW IN THE PORE NETWORK
From a given hierarchical assembly of capillary pores, the objective is now to

estimate not only the intrinsic but also the apparent permeability. The apparent
permeability is obtained by combining Darcy’s law, which represents the flow at the
macro-scale, and Poiseuille and Knudsen laws, which represent the flow at the micro-
scale.
Poiseuille’s flow at the micro-scale

Let us consider a cylindrical pore of diameter݀௜ . The flow (i.e. volume of fluid per
unit time) passing through the pore of length ‫ܮ‬௖ is given by:
ͳ ߨ ݀௜ସ ሺܲଵଶ െ ܲଶଶ ሻ

ܳ௉௩ ൌ ൬ ൰ (4)
ʹ ͳʹͺ ߤ ‫ܮ‬௖ ܲଶ
where (ܲଵ ǡ ܲଶ ) are the upstream and downstream pressures respectively, ܳ௉௩ is the
volume of fluid estimated at the outlet pressure and ߤ is the fluid dynamic viscosity.
This classical expression relies on the assumption of laminar flow [7].
Knudsen flow at the micro-scale

In the Knudsen region, the mass transfer of a gas occurs due to free-molecule flow
[8]. This region is defined for each pore diameter from the Knudsen number ܰ௜௄௡ :
© ASCE
CONCREEP 10 54
ܰ௜௄௡ ൐ ͳ
ߣ ͳ ܴܶ (5)
ܰ௜௄௡ ൌ ൌ൬ ൰
݀௜ ݀௜ ߨ ξʹ ݀௝ଶ ܰ஺௩ ܲ௠
where (ߣ, ݀௝ ,ܲ௠ ) are the average free path, the molecule diameter and the average
pressure of the considered gas respectively, and (ܰ஺௩ ) is the Avogadro’s number.
When the Knudsen number is greater than one, the interaction between the walls of
the pore and the gas, at the molecular level, are dominant and the fluid flow is quite
different from Poiseuille’s flow. The volume of fluid per unit time passing in the pore
is:
ߨ ݀௜ଷ ͺ ܴ ܶ οܲ
௩
ܳ௄௡ ൌ ඨሺ ሻ (6)
ͳʹ ‫ܮ‬௖ ߨ ‫ܲ ܯ‬ଶ
whereοܲ ൌ ܲଵ െ ܲଶ is the difference between the upstream and downstream

pressures.
In the transition regime defined as (ͲǤͲͳ ൏ ܰ௜௄௡ ൏ ͳሻ the flow rate through a pore ሺ݅ሻ
due to a pressure gradient is given as the sum of Poiseuille flow (Eq. 4) and Knudsen
flow (Eq. 6):
Macroscopic transport properties

At the macro-scale, the flow rate is described by Darcy’s law [6]:
‫ܭ‬௔ οܲ ܲ௠
ܳ஽௩ ൌ ܵ (7)
ߤ ‫ܮ‬௘ ܲଶ
For hierarchical assembly of capillaries, the transport properties are obtained by

adding the contribution of each capillary pore. Comparison between Eq. (7) and the
equations at the micro-scale provides the expression of the apparent permeability‫ܭ‬௔ .
EXPERIMENTAL DATA AND COMPARISONS WITH THE MODEL
Experiments have been performed on mortars specimens with water/cement ratio of

0.7 and a maximum aggregate diameter of 3mm. Mortar hollow cylinders were cast
in aluminum molds (outer diameter: 24.5mm; inner diameter: 6.5mm; height: 40
mm). Top and bottom surfaces were rectified and the samples were cured in a
ventilated oven at 80°C at constant humidity during 48 hours.
The testing apparatus consists in a conventional electro-mechanical compressive

machine coupled with a nitrogen permeameter in order to measure in-situ the
permeability under loading [5]. The initial permeability is measured before applying a
given compressive load corresponding to a given level of damage (measured by the
variation of the Young’s modulus of the material upon unloading [9,11]).
Permeability is measured under load and after load. Several load levels are
© ASCE
CONCREEP 10 55
considered, corresponding to several levels of damage ranging from 0% to 12% (in

the pre-peak regime).
The apparent permeability has been measured by applying a pressure gradient and
measuring the radial gas flow through the sample (see e.g.[11]):
ܴ
ܳ௜ ߤ ܲ௜ ሺܴଶ ሻ
‫ܭ‬௔ ൌ ଵ (8)
ߨ݄ ሺܲ௜ଶ െ ଶ
ܲଶ ሻ
where ܳ௜ is the gas volumetric flow, (݄ǡ ܴଶ ǡ ܴଵ ) are the sample height, outer and inner
diameters respectively and ሺܲ௜ ǡ ܲଶ ) are the gas upstream and downstream pressure
respectively. Typically ܲଶ is the atmospheric pressure. In order to derive the intrinsic
permeability from experimental data, the approach of Klinkenberg [10] is applied:
ߚ (9)
‫ܭ‬௔ ൌ ‫ܭ‬௜௡ ሺͳ ൅ ሻ
ܲ௠
where ߚ is the coefficient of Klinkenberg, ‫ܭ‬௜௡ is the intrinsic permeability and ܲ௠ is
the mean pressure.
After complete unloading, the middle part of the sample (15mm over 40mm) is cut
out in order to characterize the PSD. The PSD measurements has been performed by
mercury intrusion (Micromeritics–AutoPore IV). Figure 3 shows the evolution of
pore size distribution of different samples subjected to different mechanical damage
(from 0 to 12%). Initially, a peak at 150nm is observed with no porosity higher than
1000nm. As damage increases, a slight shift of the PSD is observed between 1000nm
and 3000nm. Moreover new porosity is created around 10000nm. This new porosity
leads to the increase of the intrinsic permeability of two orders of magnitude.
Fig. 3. Evolution of pore size distribution upon damage.
Figure 4 present a comparison between the measured and predicted intrinsic

permeabilities and Klinkerberg coefficient evolutions upon damage from 0% to 12%.
It is shown that the predicted values are in good agreement with the experimental
© ASCE
CONCREEP 10 56
ones whatever the state of damage. Note that no tortuosity parameter has been
introduced in the model (tortuosity equals 1).
Fig. 4. Comparison between the measured and predicted intrinsic permeability

(left) and Klinkenberg coefficients (right).
Figure 5 shows a comparison between the apparent permeability to nitrogen and to

helium for an undamaged sample measured experimentally and computed according
to the same PSD measured experimentally. We observe again a good agreement. It is
important to notice that when performing the comparisons, no specific model
parameter is set to predict the apparent permeability with the proposed model upon a
change of the nature of the gas. Since the helium mean free path is greater than the
nitrogen one, the slip flow effect is higher for helium and then the slope of the
apparent permeability vs. mean pressure curve is also higher compared to the other
gases.
Fig. 5. Comparison between the model and the experimental results of apparent
permeabilities for N2 and He.
CONCLUSIONS AND EXTENSIONS
A theoretical model aimed at computing the apparent permeability of a porous

material to gas has been presented. The model uses the pore size distribution of the
material as an input. The porous network is generated according to a hierarchical
© ASCE
CONCREEP 10 57
scheme in which bundles of capillaries are assembled randomly. Within each

capillary pore, fluid flow is described by a combination of Poiseuille Flow and
Knudsen (diffusion) flow. The outcome at the macroscopic scale results into a
Klinkerberg-type of apparent permeability.
The model has been compared with experiments on mortar subjected to compression.
The theoretical model captures experimental trends very consistently. In addition, the
same model is also capable of predicting the permeability of the material to a
different gas, without any adjustment of the parameters.
This theoretical model can be extended to several cases, among which the
permeability to a mixture of gases is a first step. According to Poiseuille flow, the
separation of gases upon permeation through a porous material cannot occur. By
separation, we mean that the composition of the mixture measured upstream, before
entering the porous material, and downstream is not the same. Because Knudsen
effect depends on the molecular sizes of each species, separation will occur and will
depend on the relative amount of species. Hence, a deviation from the linear variation
of the downstream molar fractions with the change of composition of the mixture
upstream, obtained according to Poiseuille flow, is to be expected.
The experimental set-up has been modified in order to accommodate the case of a
mixture of two gases and the analysis of the composition of the mixture upstream and
downstream has been added. Figure 6 shows the case of He/CH4 mixture considered
as a test case. On the horizontal axis is plotted the molar fraction of He on the
upstream face of the specimen. On the vertical axis are plotted the molar fraction of
each gas downstream. Deviation from straight lines is due to Knudsen effect and
show that the porous material separates the gas components.
Fig. 6: Separation of He/CH4 mixture (predicted and measure experimentally)
Further developments are currently being considered, investigating the case of a

mixture of phases of the same fluid, for instance water and vapour where as opposed
to He/CH4, phase change may occur depending on the diameter of the pores and on
the temperature and pressure conditions.
© ASCE
CONCREEP 10 58
ACKNOWLEDGEMENTS
Financial supports from the European Research Council through the ERC advanced
grant Failflow (27769), and from the French Région Aquitaine through the grant
CEPAGE (20121105002) are gratefully acknowledged. F. Khaddour is grateful to the
Syrian Ministry of Higher Education for its support (13153/4/W).
REFERENCES
1. Kozeny, J. 1927. Über Kapillare Leitung Des Wassers Im Boden [About

capillary pipe of water in soil]. Wien, Akad. Wiss, 136: 271–306.
2. Katz, A. J. and A. H. Thompson. 1986. Quantitative prediction of permeability
in porous rock. Physical review B, 34: 8179–8181.
3. Aït-Mokhtar, A., O. Amiri, P. Dumargue and S. Sammartino. 2002. A new
model to calculate water permeability of cement materials from MIP results.
Magazine of Concrete Research, 14: 43–49.
4. Amiri, O., A. Aït-Mokhtar and M. Sarhani. 2005. Tri-dimensional modelling of
cementitious materials permeability from polymodal pore size distribution
obtained by mercury intrusion porosimetry tests. Magazine of Concrete
Research, 17: 39–45.
5. Khaddour, F., D. Grégoire and G. Pijaudier-cabot. 2012. Upscaling Permeation
Properties in Porous Materials from Pore Size Distributions. In Geomechanical
Issues in CO2 Storage Facilities, eds G. Pijaudier-Cabot and J.-M. Pereira, 44–
56.
6. Darcy, H. 1856. Les fontaines publiques de la ville de Dijon [The public
fountains of the city of Dijon]. Paris: Dalmont.
7. Poiseuille, J. 1840. Recherches expérimentales sur le mouvement des liquides
dans les tubes de très petits diamètres [Experimental research on the movement
of liquids in capillary of very small diameters]. Comptes rendus hebdomadaires
des séances de l’Académie des sciences, T11: 1041–1048.
8. Steckelmacher, W. 1986. Knudsen flow 75 years on: the current state of the art
for flow of rarefied gases in tubes and systems. Rep. Prog. Phys, 49: 1083–
1107.
9. F. Khaddour, D. Grégoire and G. Pijaudier-Cabot. 2015. Capillary bundle
model for the computation of the apparent permeability from pore size
distributions. European journal of environmental and civil engeneering, 9: 168–
183.
10. Klinkenberg, L. J. 1941. The permeability of porous media to liquids and to
gases. In Drilling and Production Practice, ed. American Petroleum Institute,
200–213.
11. Choinska, M., A. Khelidj, G. Chatzigeorgiou and G. Pijaudier-Cabot. 2007.
Effects and interactions of temperature and stress-level related damage on
permeability of concrete. Cement and Concrete Research, 37: 79–88.
© ASCE
CONCREEP 10 59
Recent Experimental Results about the Basic Creep of Concretes andD New
Approach to Model It
Pierre Rossi
Institut Français des Sciences et Technologies des Transports, de

l’Aménagement et des Réseaux (IFSTTAR), Department of Materials and
Structures, Cité Descartes, 14-20 bd Newton, 77447 Champs-sur-Marne cedex 2,
France. E-mail: pierre.rossi@ifsttar.fr
Abstract
This paper presents a summary of recent experimental researches about

basic creep of concretes. From the analysis of these researches, a proposal is
made concerning the dominant physical mechanism at the origin of this basic
creep. Based on this physical mechanism, a new way to model, by using finite
elements method, basic creep of concrete structures is proposed.
INTRODUCTION
The international literature on basic creep of concrete is considerable.

Despite this, there is no, today, an international consensus about the physical
mechanisms related to this mechanical behaviour.
It should be noted that several experimental studies have been conducted
over the last five years [1 to 6] which converge to the high probability of
existence of a dominant physical mechanism that is responsible for this basic
creep.
This dominant physical mechanism leads to the questioning of the way
usually chosen to model the creep behaviour of concrete structures.
The first part of this paper summarizes the experimental evidences
obtained in the frame of the studies evocated above and the physical mechanism
proposed in relation with these evidences.
The second part of the paper presents the basis of a new numerical model
which takes into account the physical mechanism proposed.
The third and last part presents the way to determine experimentally the
values of the mechanical parameters of the model.
© ASCE
CONCREEP 10 60
EXPERIMENTAL EVIDENCES RELATED TO THE BASIC CREEP OF

CONCRETE
The experimental studies, above referenced, led to the following

evidences:
- The basic creep strain is directly leaked to the creation of microcracks

during the creep loading.
- The basic creep in uniaxial tension is about three times less than the
basic creep in compression when tests are performed on specimens of
similar volume (usually used in the laboratories).
- The basic creep in uniaxial tension depends on the stressed volume of
material. The higher the volume, the greater the basic creep is. This
volume effect has to be related to the one existing when the static
tensile behaviour of concrete is concerned.
- The compressive basic creep is not volume effect dependent.
- The compressive basic creep can be considered as linear until a stress
level of about 50% of the compressive strength of the material.
- The tensile basic creep can be considered as linear until a stress level of
about 70% of the uniaxial tensile strength of the material.
- The residual strains observed after unloading are related to the initial
stresses existing in the material which are released during the creep
loading when micro-cracks are created.
All these experimental evidences leaded to a proposal concerning a

predominant physical mechanism at the origin of the basic creep of concrete.
This physical mechanism can be summarized as following:
Under the effect of creep loading, concrete undergoes a micro-cracking

process. This process results in an increase in the density of micro-
cracks over time. This evolution leads to the creation of an additional
self-drying shrinkage. In other words, the basic creep of concretes can
be considered as self-drying shrinkage under stresses.
These experimental evidences lead to these logical consequences: the

basic creep has no more to be modelled like in the past. It is thus necessary:
© ASCE
CONCREEP 10 61
- To stop considering that the total strain under creep loading can be
considered as the sum of the strain related to self-drying shrinkage and
the strain related to basic creep. This decomposition has no physical
meaning.
- To stop considering that the basic creep in tension is identical to the
basic creep in compression.
- To take into account the volume effects related to the basic creep in
tension.
A new way to model basic creep of concrete is then proposed in the frame
of finite elements method. This approach which is more physically relevant than
those existing today is presented below.
A NEW NUMERICAL MODEL TO DESCRIBE THE BASIC CREEP OF

CONCRETE
Taking into account what is proposed concerning the physical origin of the
basic creep of concrete, it is possible to identify the main assumptions of a
numerical model physically based. Thus, such a model should (not exhaustive
list):
- Be able to describe explicitly the cracking process during the creep

loading.
- Take into account the existence of initial stresses within the material.
- Take into account the volume effects related to the static tensile
behaviour of concrete.
- Explicitly connect the cracking process during the creep loading to the
self-drying shrinkage induced by this cracking process,
A numerical model consistent with the respect to the explicit modelling of

cracking process was developed since several years [8 to 10] at IFSTTAR. This
fully validated model is a probabilistic one. It serves as the basis of the model
proposed to describe creep of concrete.
Some recalls related to the probabilistic explicit cracking model are presented
below.
© ASCE
CONCREEP 10 62
The probabilistic explicit cracking model. The numerical model is based on

three physical evidences:
- Concrete is a heterogeneous material in which the heterogeneities can

be modeled through a random spatial distribution of mechanical
properties considered as dominant in the cracking process, namely the
Young's modulus and the tensile strength;
- Scale effects of concrete cracking (experimentally proved) are taken
into account by the fact that the mechanical properties of the material
depend on the size of the mesh elements chosen for the finite element
analysis.
- Cracking is explicitly treated through the creation of random kinematic
discontinuities, which provides access to quantitative information on
the cracking state (number of cracks, opening and spacing).
Numerically speaking, these cracks are represented by interface
elements.
The criterion for the crack creation is very simple: the interface element
opens when the normal tensile stress at the centre of the interface element
reaches a critical value (probabilized through a Weibull distribution function
[10]). This means that the rigidity of the interface element becomes equal to zero
(perfect brittle behaviour). As mentioned above, the value of this critical value
depends on the total volume of the two volumetric elements interfaced by the
considered interface element. The criterion of crack creation is schematically
presented in the Figure 1.
Remark. It is very important to introduce a random distribution of the
Young modulus to generate local tensile stresses (and so cracks) when the
compressive behaviour of a specimen is concerned. The random distribution of
the Young modulus depending of the size of the element mesh is a log-normal
one [11].
From the above model, it is necessary to introduce strains versus time
evolution in relation with the cracking process to describe basic creep of
concrete in the frame of the physical mechanism proposed in chapter II. So, an
extension of the probabilistic explicit cracking model is presented in the
following.
© ASCE
CONCREEP 10 63
Figure1. Criterion of crack creation (explicit probabilistic cracking model)
Strains versus time evolution generated by the cracking process. From the
cracking model presented above, it is important to take into account the self-
drying shrinkage generated by this cracking and so to model the strain versus
time induced by this mechanism. Two types of approaches are considered and
actually in development to model the effect of this self-drying shrinkage:
- a direct one which consists in introducing, in the volume elements

surrounding the interface elements considered as opened, an additional
negative strain which evolutes during time.
- a second one, indirect, which consists in decreasing during time the
value of the Young modulus of the volume elements surrounding the
interface elements considered as opened.
These two strategies have for consequence, in complement to the fact that
a global strain versus time is generated during the creep loading, to induce a
stress redistribution around the (static) cracks. This redistribution allows the
creation of new cracks during the creep loading and, by this way, allows the
evolution of this global strain.
© ASCE
CONCREEP 10 64
The cracking process obtained by the use of the probabilistic explicit

cracking model described above is mesh independent. To conserve this same
type of mesh independency with the model extensions presented above, the
additional strain versus time law or the Young modulus versus time law are
considered as dependent of the mesh size.
Remarks. The residual strain observed when a creep specimen is unloaded is

generated, in the proposed numerical model, by the fact that initial stresses are
introduced before any mechanical loading and released during the cracking
process.
The experimental evidence related to the fact that the basic creep in
tension is different of the basic creep in compression is taken into account by the
cracking model itself.
The strategy to quantify the additional strain versus time law or the Young
modulus versus time law in function of the mesh size is presented in chapter IV.
EXPERIMENTAL STRATEGY TO QUANTIFY THE MODEL

PARAMETERS OF THE BASIC CREEP OF CONCRETES
It is needed to define an experimental strategy for determining the input

data of the proposed model.
Considering what is said in paragraph II and the works it refers [1 to 6],
the inevitable material parameters to be considered in the experimental strategy
are:
- The ability of the concrete to micro-cracking (creating of greater or

lesser density of micro-cracks) under a given stress ratio.
- The water/binder ratio of the concrete. This ratio will determine on the
one hand the amount of movable water and vapour and, on the other
hand, will determine the potentiality of the material to create self-drying
shrinkage.
- The temperature within the material. This temperature can increase the
density of micro-cracks (the coefficients of thermal expansion between
the cement paste and the aggregates being different, they can generate
tensile stresses and therefore local micro-cracks), an increase of the
water viscosity, as well as changes in the water/vapour equilibrium.
These last two points can induce changes in the amplitude and the
kinetics of the additional self-drying shrinkage.
© ASCE
CONCREEP 10 65
Taking into account these material parameters the mechanical test to be

performed to quantify the mechanical parameters of the proposed model are:
- Compressive basic creep tests on specimens of any dimension (the basic

compressive creep being no sensitive to the volume effects). These tests
can be performed in a range of loading ratios comprised between 10%
and 85% (it has been shown in the past that this last loading ratio is
convenient to performed and control a creep test for studying all the
steps of the creep behaviour until the rupture of the specimen).
- Bending basic creep tests on specimens of different dimensions (to
perform uniaxial tensile creep tests on specimens of different dimension
being very difficult). The range of loading ratio will be identical to the
one of the compressive creep tests. From these bending basic creep
tests, the basic creep in uniaxial tension, integrating volume effects, can
be determined by using an inverse approach. As matter of fact, knowing
the basic creep in compression, it is easy to determine, by using a
numerical model (finite elements method), the basic creep in uniaxial
tension that correctly reproduces the basic creep in bending. To do this,
it is necessary to use the same numerical model as the one chosen to
analyse the basic creep of concrete structures.
Remarks.
- If, as mentioned above, the basic creep of a given concrete is mainly
linked to that of its matrix through the water/binder ratio, it is
necessary, for each type of test (compressive and bending creep tests),
to test various concrete with different water/binder ratios and different
volumes of binder.
- In a fairly comprehensive experimental study related to the compressive
basic creep of normal concrete [7], it was clearly shown that this basic
creep is a mechanical characteristic for which the dispersion is
relatively large. It therefore seems necessary to develop a probabilistic
approach to model the basic creep behaviour rather than a deterministic
one as usually done. To do this, it is imperative, for each type of basic
creep test above mentioned and for each type of concrete, to perform
enough tests to get enough experimental data for this probabilistic
approach.
© ASCE
CONCREEP 10 66
From these experimental tests, the mechanical parameters of the model are
obtain through the use of an inverse approach. It means that the tests have to be
simulate with the model. The parameters values of the cracking model being
known [8 to 11], the additional strain versus time law or the Young modulus
versus time law are obtained by fitting the tests results.
The dispersion related to the creep tests are considered as the consequence
of the dispersion related to the cracking process. So, for each type of test and for
a given meshing of the specimen, several simulations are performed. To get the
additional strain versus time law or the Young modulus versus time law in
function of the mesh size, different meshing of the specimens has to be
considered.
CONCLUSIONS
This paper proposes a new numerical model for analysing basic creep
behaviour of concrete structures. This model is based on experimental evidences
which led to the proposal concerning the predominant physical mechanism at the
origin of the basic creep of concretes. This paper proposes also an experimental
strategy to quantify the mechanical parameters of the numerical model proposed.
The originality of the paper is based on the overall consistency of the
approach (which integrates a proposal of the main physical mechanism of basic
creep of concrete, a numerical model taking into account this physical
mechanism and an experimental strategy to quantify the mechanical parameters
of the model).
The next paper will present quantitative comparisons between the
proposed model and experimental results (this work is in progress).
REFERENCES
Clément, J.L., Le Maou, F. (2001). "Experimental repeatability of creep and

shrinkage concrete tests - data statistical analysis and modelling."
Concreep 6, 6th international conference on creep, shrinkage and
durability mechanics of concrete and other quasi-brittle materials, MIT
Cambridge USA.
Rossi, P., Richer, S. (1987). "Numerical modelling of concrete cracking based
on a stochastic approach." Materials and Structures, vol. 20, pp. 334-337.
Rossi, P., Wu, X. (1992). "A probabilistic model for material behaviour analysis
and appraisement of the concrete structures." Magazine of Concrete
Research, vol. 44, n°161, pp. 271-280.
© ASCE
CONCREEP 10 67
Rossi, P., Ulm, F.J., Hachi, F. (1996) "Compressive behaviour of concrete:

physical mechanisms and modelling." ASCE Engineering Mechanics,
vol. 122 (11), pp-1038-1043.
Rossi, P., Tailhan, J.L., Le Maou, F., Gaillet, L., Martin, E. (2012). "Basic
Creep Behaviour of Concretes. Investigation of the physical mechanisms
by using acoustic emission." Cement and Concrete Research, Vol. 42,
pp.61-73.
Rossi, P., Tailhan, J.L., Le Maou, F. (2013). "Comparison between concrete
creeps in tension and in compression: influence of the concrete age at the
loading and of the drying conditions." Cement and Concrete Research,
51, 78-84.
Rossi, P., Tailhan, J.L., Le Maou, F., Martin, E (2013). "Creep strain versus
residual strain of a concrete loaded under various levels of compressive
stress." Cement and Concrete Research, 51, 32-37.
Rossi, P., Charron, J.P., Bastien-Masse, M..Tailhan, J.L, Le Maou, F.,
Ramanich, S. (2014). "Tensile basic creep versus compressive basic creep
at early ages: comparison between normal strength concrete and a very
high strength fibre reinforced concrete."Materials and Structures, vol. 47,
pp. 1773-1785
Rossi, P., Boulay, C., Tailhan, J.L., Martin, E (2014). "Macrocrack propagation
in a concrete specimen under sustained loading: study of the physical
mechanisms." Cement and Concrete Research, vol. 63, pp. 98-104.
Tailhan, J.L., Dal Pont, S., Rossi, P. (2010). "From local to global probabilistic
modelling of concrete cracking." Ann. Solid. Struct. Mech., vol. 1, pp.
103-115.
Tailhan, J.L., Boulay, C., Rossi, P., Le Maou, F., Martin E. (2012).
"Compressive,tensile and bending basic creep behaviours related to the
same concrete."Structural Concrete, 14, n°2, pp. 124-130.
© ASCE
CONCREEP 10 68
The Three-Fluids Model for Concrete and Beyond
B. A. Schrefler
Department of Civil, Environmental and Architectural Engineering, University of

Padua, Via Marzolo 9, 35131 Padova, Italy; and Department of Translational
Imaging and Department of Nanomedicine, Houston Methodist Research Institute,
6670 Bertner Ave., Houston, TX 77030.
E-mail: Bernhard.schrefler@dicea.unipd.it
Abstract
Recent advances in the application of the three fluids model for concrete will be
discussed where creep matters such as in repair problems. As known, this model
considers concrete as composed of a deformable solid matrix with pores filled by
dry air, capillary water and water vapor. The interaction of the constituents as well
as phase change is taken into account. It will also be shown that the same material
free balance equations can successfully be adopted for the simulation of tumor
growth and drug delivery. In this case the solid phase is the extracellular matrix
and the three fluids are formed by the healthy cells, the tumor cells and the
interstitial fluid which transports the chemical species (therapeutic agents). The
extracellular matrix is an elasto-visco-plastic porous solid of the Perzyna type.
Applications for both forms of the general model will be shown.
INTRODUCTION
In the last couple of decades we have developed two major three fluid flow
models in deforming porous solids. The first is our three fluids model for concrete
which has been developed from the late nineties on and extended to many
applications in concrete technology, such as concrete under high temperatures
(Gawin et al. 1999, 2006 c), leaching in isothermal and non-isothermal conditions
(Gawin et al. 2008 a,b, 2009), Alkali-Aggregate Reactions (Pesavento et al.
2012), young concrete and repair problems (Gawin et al, 2006 a,b, Sciumè et al.
2013 a), and recently also freezing/thawing (Koniorczyk et al. 2015). It considers
concrete as formed by a deformable solid matrix the pores of which are filled with
dry air, water vapour and liquid water. The interactions between all the
constituents are duly taken into account. The second one is our recently developed
tumor growth model which considers tumors as comprised of different
compartments: the extracellular matrix as deformable porous solid, the pores of
which are filled with healthy cells, tumor cells, both living and necrotic, and the
interstitial fluid which carries nutrients, therapeutic agents, and waste products
(Sciumè et al. 2012 a, 2013c, 2014 a-c, Bao et al. 2014). The
vasculature/neovasculature is a separate compartment with only mass exchange
with the other compartments. The cells are treated as adhesive fluids. Both models
are hence de facto three fluid flow models in a deforming porous medium. The
same set of material free balance equations apply to these apparently completely
different problems, governing the structure of the code for numerical solution.
© ASCE
CONCREEP 10 69
This insight allowed us to build the second model in very short time starting from
the concrete model even if the constitutive behavior and evolution laws differ
substantially. Both models are implemented in the code CAST3M (2003) of the
French Atomic Energy Commission (CEA) and are easily accessible. In the
following we show the balance equations for both cases to convince the reader of
the analogy, and a recent application for each of the two cases.
THREE-FLUIDS MODEL FOR CONCRETE AND APPLICATION
The balance equations are written by considering concrete as a multi-phase porous

material assumed to be in thermal-, hygral- and mechanical equilibrium state
locally, even if chemical reactions (hydration process) are progressing with a
certain rate. The local chemical equilibrium is reached only at the end of the
spontaneous hydration process and from that time local thermodynamic
equilibrium state is reached at a given point. Characteristic times of the local
thermal, hygral and mechanical processes are much smaller than that of hydration
reaction; hence, for these processes we use the equilibrium relations (with the
actual value of hydration degree being ‘frozen’) (Gawin et al. 2006 a). The solid
skeleton voids are filled partly by liquid water and partly by a gas phase. The
liquid phase consists of bound water, which is present in the whole range of
moisture content, and capillary water, which appears when the degree of water
saturation exceeds the upper limit of the hygroscopic region. The gas phase is a
mixture of dry air and water vapor (condensable gas constituent) and is assumed
to behave as an ideal gas. The model equations are obtained by means of the
Thermodynamically Constrained Averaging Theory TCAT (Gray and Miller,
2014, Gray et al., 2013) which is used to transform known micro scale relations to
mathematically and physically consistent macro scale relations. The balance
equations are here written already at macroscopic level for sake of brevity.
The mass balance equation for the solid phase s is
∂ (ε s ρ s )
( )
l → Hs
+ ∇ ⋅ ε s ρ s vs = M
∂t
The mass balance equation of liquid water l is
∂ (ε l ρ l )
( )
l → Hs l →Wg
+ ∇ ⋅ ε l ρ l vl = − M − M
∂t
The mass balance of water vapor Wg reads
(
∂ ε g ρ g ωWg ) +∇⋅ ε
( ) ( )
l →Wg
g
ρ gωWg v g + ∇ ⋅ ε g ρ gωWg uWg = M
∂t
and that of dry air Ag is
© ASCE
CONCREEP 10 70
(
∂ ε g ρ gω Ag ) +∇⋅ ε
∂t
( g
) ( )
ρ gω Ag v g + ∇ ⋅ ε g ρ gω Ag u Ag = 0
l → Hs l →Wg
where M is the chemically combined water, M the vaporized water, g
indicates the gaseous phase, mixture of dry air and water, İ the volume fraction, Ȧ
the mass fraction, v the phase velocity, u the relative velocity and ȡ the density.
Finally the linear momentum balance equation is
§ ∂t eff
s
∂p s ·
∇ ⋅¨ − 1¸ = 0
¨ ∂t ∂t ¸¹
©
s
where t eff is the effective stress and p s the solid pressure (Gawin et al. 2007).
These balance equations have to be supplemented with appropriate constitutive
equations and evolution laws for which the reader is referred to the relevant
papers in the References.
As typical application of the model we show the results of the numerical
simulation of cracking of a massive concrete beam. The analyzed test is a large
beam specimen built for ConCrack (2011): the international Benchmark for
Control of Cracking in Reinforced Concrete Structures (Figure 1). This
benchmark has been organized within the French national project CEOS
(Comportement et Evaluation des Ouvrages Speciaux vis-à-vis de la fissuration et
du retrait) dedicated to the analysis of the behavior of special large specimens
concerning cracking and shrinkage.
Figure 1. Experimental specimen (right) and finite elements mesh (left),

Reprinted from Simoni and Schrefler (2014), Copyright (2014), with
permission from Elsevier.
Here one of the experiments of the benchmark program and its numerical
simulation are presented. The test is divided in two phases. During the first 48
hours after casting, the structure is thermally isolated and protected from drying.
Then after two days the isolation and the formwork are removed and the structure
is conserved during two months in the open environment. During the two phases
of the test the longitudinal strains of the structure are globally restrained by two
metallic struts. After two months the structure is loaded by a static bending test.
© ASCE
CONCREEP 10 71
For more details on the analyzed test see the reference web page of the benchmark
(www.concrack.org). The mesh of concrete is made of 3D elements (Figure 1,
left). Truss elements rigidly linked with the concrete 3D mesh are used to model
the struts. The local damage model of Mazars (1986) has been used. This
approach allows predicting cracking in mode I (De Sa et al. 2008, and De Sa and
Benboudjema 2012) which is the case here, since cracking is induced by
shrinkage restraint. The model is regularized in tension with the introduction of a
characteristic length, lc, related to the size of each finite element (Rots, 1988, and
Cervera and Chiumenti, 2006) to overcome the possible mesh dependency. The
model for concrete at early age is that of Sciumè et al. (2013 a).
The boundary conditions are assumed to be of convective type for both heat and
mass exchange, see (Simoni and Schrefler, 2014) for details. For the material
parameters see Sciumè et al. (2012 b).
During the first and second phase the longitudinal displacements of the specimen
have been globally restrained by the two metallic struts. Therefore during
hydration the thermal extension of the concrete has been restrained and so in this
phase the concrete cross-section was compressed. After the hydration it is the
shrinkage of the beam that has been contrasted. The position of the displacement
measurement points are depicted in Figure 2 and the deformed configuration and
the damage at day 60 in Figure 3. The solution is not symmetric due to the non-
symmetric solar irradiation: this has been experimentally measured by the
benchmark organizers and taken into account in the modeling process.
Figure 2. Position of the displacement measurement points (and coordinates).

Reprinted from Simoni and Schrefler (2014), Copyright (2014), with
permission from Elsevier.
Figure 3. Deformed configuration (x 500) and damage after 60 days. Face

exposed to the sun (left) and face not exposed to the sun (right). Reprinted
from Simoni and Schrefler (2014), Copyright (2014), with permission from
Elsevier.
© ASCE
CONCREEP 10 72
The comparison between the experimentally and numerically obtained relative

displacements and axial force is shown in Figure 4. Note the complex
displacement evolution.
Figure 4. Relative displacement between the point C and D (left). Axial force
in the two metallic struts that restrain shrinkage (right). Reprinted from
Simoni and Schrefler (2014), Copyright (2014), with permission from
Elsevier.
After two months a static four point bending test has been carried out until
rupture. Compression has been applied by eight jacks and live controlled by a
pressure sensor with an independent data registration system. The load is applied
with increments of 50 kN and each loading step is kept during 20 min. The
comparison between the experimental and numerical crack pattern and the force
versus displacement diagram are shown in Figure 5. The comparison in Figures 4
and 5 with experimental results shows that the adopted approach is able to
simulate such complicated multi-field fracture phenomena in realistic engineering
structures. For more results concerning hydration and thermal field the reader is
referred to the website of ConCrack (2011) and the final report.
Figure 5. Comparison between the numerical and experimental crack pattern

(left). Force vs displacements during the bending test (right). Reprinted from
Simoni and Schrefler (2014), Copyright (2014), with permission from
Elsevier.
TUMOR GROWTH MODEL AND APPLICATION
From review papers of numerical models for tumor growth such as Roose et al.
(2007), Deisboeck et al. (2011) and Sciumè et al. (2013 c), it appears clearly that
in the realm of a continuum approach a vast majority of models describe the
malignant mass (TC), the host cells (HC) and the interstitial fluid (IF) as
homogeneous, viscous fluids and employ reaction-diffusion-advection equations
© ASCE
CONCREEP 10 73
for predicting the distribution and transport of nutrients. The Extracellular matrix
(ECM) is absent in almost all the models.
In our model the tumor is comprised of the following phases: i) tumor cells (TC),
which partition into living cells (LTC) and necrotic cells (NTC); ii) healthy cells
(HC); iii) extracellular matrix (ECM); and iv) interstitial fluid (IF); see Sciumè et
al. (2012 a, 2013c, 2014 a-c). The ECM is a porous solid, while all other phases
are fluids. The tumor cells may become necrotic upon exposure to low nutrient
concentrations or excessive mechanical pressure. The IF is a mixture of water and
biomolecules, such as nutrients, oxygen, therapeutic agents and waste products. In
the following mass and momentum balance equations, Į denotes a generic phase, t
the tumor cells (TC), h the healthy cells (HC), s the solid phase (ECM), and l the
interstitial fluid (IF).
The balance equations are listed below to show the analogy with the concrete
model. For the full development the reader is referred to Sciumè et al. (2013 b,
and 2014 a-c).
The mass balance equations of the solid s is
∂ (1 − ε )
+ ∇ ⋅ ª¬(1 − ε ) v s º¼ = 0
∂t
where vs is the solid phase velocity.
The mass balance equations of the tumor cell phase t, the host cell phase h and
interstitial fluid l are respectively
∂ (ε S t ) § k t k ts a · 1 l →t
+ ∇ ⋅ ( ε S t v s ) − ∇ ⋅ ¨ rel t 1 − t t ∇p t ¸ = M
∂t ¨ μ ∇p ¸ ρ growth
© + ¹
∂ (ε S )
h § k h k hs a ·
+ ∇ ⋅ ( ε S h v s ) − ∇ ⋅ ¨ rel h 1 − h h ∇p h ¸ = 0
∂t ¨ μ ∇p ¸
© + ¹
∂ (ε S )
l
§k kl ls
· 1 l →t
+ ∇ ⋅ ( ε S l v s ) − ∇ ⋅ ¨ rel l ∇pl ¸ = − M
∂t © μ ¹ ρ growth
Į
where aĮ is the adhesion (al =0), μ is the dynamic viscosity, k rel α
is the relative
permeability which takes care of the presence of the other two fluid phases
(Sciumè et al. 2014 a), k α s is the absolute permeability, pĮ is the pressure and ȡ is
l →t
the common density; M is the rate of growth term. The linear momentum
growth
balance equations of the fluid phases have here already been introduced to show
the difference between cells and IF.
Assuming that: i) there is no diffusion of either necrotic or living cells; ii) there is
no exchange of necrotic cells with other phases; the mass conservation equation
for the necrotic portion of the tumor cells phase reads
(
∂ ε t ρω N t ) + ∇ ⋅ ε ρω v − ε r
∂t
(
t Nt
) t t Nt
=0
© ASCE
CONCREEP 10 74
ε Nt ρ Nt
where ω N t = is the mass fraction of necrotic cells in the tumor cells phase
ε tρt
t, ε t r Nt the rate of generation of necrotic cells and vt the velocity of the tumor
cells phase.
The mass balance equation of the nutrient species in the interstitial fluid is
( )
∂ ε S lcg
( )
gl →t
− ∇ ⋅ ε S l Deffgl ∇c g = − M
∂t
where Deffgl is the effective diffusion coefficient depending on the available pore
gl →t
space, (Sciumè et al., 2014 a) and M is the nutrient consumption rate which
depends on the local nutrient availability. Finally the linear momentum balance
equation of the solid phase in rate form reads as
§ ∂t eff
s
∂p s ·
∇⋅¨ − 1¸ = 0
¨ ∂t ∂t ¸
© ¹
s
where t eff is the effective stress tensor in the solid and 1 is the unit tensor. The
interaction between solid and the fluids, inclusive of the cell populations, has been
accounted for through the effective stress principle (Gray and Schrefler, 2007).
Clearly these balance equations have to be completed with the appropriate
constitutive equations which can be found in Sciumè et al. (2013 b, 2014 a-c).
Figure 6. Skin structure and geometry of the modeled case. Reprinted from
Sciumè et al. (2014 c), Copyright (2014), with permission from IOP
publishing.
We show here as application the evolution of a melanoma during the avascular
stage. Cutaneous melanoma is the most dangerous form of skin cancer. It arises in
the melanocytes, the specialized melanin producing cells, which are scattered
along the epidermis-dermis border, Figure 6 . It is recalled that the outer structure
of the skin has a layered structure where three compartments can be evidenced:
the epidermis, an outer epithelium of stratified cells; the dermis, an intermediate
cushion of vascularized connective tissue; and the hypodermis, the lowermost
© ASCE
CONCREEP 10 75
layer made of loose tissue and adipose cells. The dermis is separated from the
epidermis by the basement membrane or basal laminae, a tough sheet of ECM, see
Figure 6. The following example deals with the evolution of a melanoma during
the avascular stage (Sciumè et al. 2014 c). Figure 7 shows the melanoma after two
weeks of growth and a qualitative comparison with a clinical observed case of
Chung et al. (2010).
Figure 7. Paraffin-embedded tissue sections of a melanocytic lesion, Chung et

al. (2010); b.m. stands for basement membrane (a); numerical results
(volume fraction of TCs) after two weeks of growth (b). Reprinted from
Sciumè et al. (2014 c), Copyright (2014), with permission from IOP
publishing.
CONCLUSION
Three fluid flow models in deforming porous media have several applications
ranging from geo-material mechanics such as concrete, to reservoir engineering
such as water, gas and oil flowing in a deforming formation, and to mechanics in
medicine such as tumor growth models. Two of such applications have been
shown. With a general framework in place it is relatively easy to pass from one
application to another in a different field. However a good aptitude to out of the
box thinking is needed as well as collaboration with experts of the fields
concerned.
Acknowledgment: the author wishes to thank Paolo Decuzzi, Mauro Ferrari,
Dariusz Gawin, William G. Gray, Francesco Pesavento, Raffaella Santagiuliana,
Giuseppe Sciumè, and Luciano Simoni for their precious collaboration.
REFERENCES
Bao, G., Bazilevs, Y., Chung, J-H., Decuzzi, P., Espinosa, H. D., Ferrari, M., Gao,
H., Hossain, S. S., T. J. R. Hughes, T. J. R., R. D. Kamm, R. D., Liu, W.
K., Marsden, A., and Schrefler, B. (2014). “USNCTAM perspectives on
mechanics in medicine.” J Royal Soc Interface, 11, 20140301.
© ASCE
CONCREEP 10 76
Cast3M (2003) FE code developed by the French Atomic Agency (CEA), France.
wwwcast3m.cea.fr.
Cervera, M., Chiumenti, M. (2006). “Mesh objective tensile cracking via a local
continuum damage model and a crack tracking technique.” Comp.
Methods Appl. Mech. Engng., 196, 304–320.
Chung, L. S., Man, Y-G. and Lupton, G. P. (2010). “WT-1 expression in a
spectrum of melanocytic lesions: Implication for differential diagnosis.”
Journal of Cancer, 1, 120.
ConCrack. (2011).” International benchmark for control of cracking in R.C.
Structures.” <www.concrack.org>.
De Sa, C., Benboudjema, F., Thiery, M., and Sicard, J. (2008). “Analysis of
microcracking induced by differential drying shrinkage.” Cement and
Concrete Composites, 30, 947–956.
De Sa, C., and Benboudjema, F. (2012). “Modeling of concrete nonlinear
mechanical behavior at high temperatures with different damage-based
approaches.” Materials and Structures, 44, 1411–1429.
Deisboeck, T. S., Wang. Z., Macklin, P., and Cristini, V. (2011) “Multiscale
cancer modeling” Annual Rev. Biomed. Eng. 13.
De Schutter, G., and Taerwe, L. (1997). “Fracture energy of concrete at early
ages.” Materials andStructures, 30, 67–71.
Gawin, D., Majorana, C.E., and Schrefler, B.A. (1999). “Numerical analysis
of hygro-thermic behaviour and damage of concrete at high
temperature.” Mech. Cohesive-Frictional Materials, 4, 37-76.
Gawin, D., Pesavento, F., and Schrefler, B.A. (2006 a). “Hygro-thermo-
chemo-mechanical modelling of concrete at early ages and beyond.
Part I: Hydration and hygro-thermal phenomena.” Int. J. Num.
Methods Eng., 67, 299-331.
Gawin, D., Pesavento, F., and Schrefler, B.A. (2006 b). “Hygro-thermo-
chemo-mechanical modelling of concrete at early ages and beyond.
Part II: Shrinkage and creep of concrete.” Int. J. Num. Methods
Eng.,67, 332-363.
Gawin, D., Pesavento, F., and Schrefler, B. A. (2006 c). “Towards
prediction of the thermal spalling risk through a multi-phase porous
media model of concrete.” Comput. Methods Appl. Mech. Engrg.,
195, 5707-5729.
Gawin, D., Pesavento, F., and Schrefler, B. A. (2007). “Modelling creep and
shrinkage of concrete by means of effective stress.” Materials and
Structures, 40, 579-591.
Gawin, D., Pesavento, F., and Schrefler, B. A. (2008 a). “Modeling of
cementitious materials exposed to isothermal calcium leaching, with
considering process kinetics and advective water flow. Part 1: Theoretical
model.” Solids and Structures, 45, 6221-6240.
Gawin, D., Pesavento, F., and Schrefler, B. A. (2008 b). “Modeling of
cementitious materials exposed to isothermal calcium leaching, with
considering process kinetics and advective water flow. Part 2: Numerical
solution.” Solids and Structures, 45, 6241-6286.
© ASCE
CONCREEP 10 77
Gawin, D., Pesavento, F., and Schrefler, B. A. (2009). “Modeling deterioration of

cementitious materials exposed to calcium leaching in non-isothermal
conditions.” Comput. Methods Appl. Mech. Engrg., 198 3051-3083.
Gray, W. G., Miller, C. T., and Schrefler, B. A. (2013). “ Averaging theory for
description of environmental problems: What have we learned?” Adv.
Water Res., 51, 123-38.
Gray, W. G., and Miller, C. T. (2014). Introduction to the Thermodynamically
Constrained Averaging Theory for Porous Medium Systems. Springer.
Koniorczyk, M., Gawin, D., and Schrefler B. A. (2015). Modelling evolution of
frost damage in fully saturated porous materials, submitted.
Mazars, J. (1986). “A description of micro and macroscale damage of concrete
structures.” Engineering Fracture Mechanics, 25, 729–737.
Pesavento, F., Gawin, D., Wyrzykowski, M., Schrefler, B. A., and Simoni, L.
(2012). “Modeling alkali-silica reaction in non-isothermal, partially
saturated cement based materials.” Comp-Meth. Appl. Mech. Engrg., 225-
228, 95-115.
Roose, T., Chapman, S. J. and Maini, P. K. (2007) “Mathematical models of
avascular tumor growth.” Siam Review, 49(2), 179-208.
Rots, J. G. (1988). “Computational modeling of concrete fracture.” PhD Thesis,
Delft University of Technology.
Sciumè, G., Shelton, S. E., Gray, W. G., Miller, C. T., Hussain, F., Ferrari, M.,
Decuzzi, P., and Schrefler, B.A. (2012 a). “Tumor growth modeling from
the perspective of multiphase porous media mechanics.” Molecular and
Cellular Biomechanics, 9(3), 193-212.
Sciume`, G., Schrefler, B. A., and Pesavento, F. (2012 b). “Thermo-hygro-chemo-
mechanical modelling of the behaviour of a massive beam with restrained
shrinkage.” CONCRACK 3, International workshop on crack control of
mass concrete and related issues concerning early-age concrete structures,
RILEM, JCI, Paris, France.
Sciume`, G., Benboudjema, F., De Sa, C., Pesavento, F., Berthaud, Y., and
Schrefler, B. (2013 a). “A multiphysics model for concrete at early age
applied to repairs problems.” Engineering Structures, 57, 374–387.
Sciumè, G, Shelton, S. E., Gray, W. G., Miller, C. T., Hussain, F., Ferrari. M.,
Decuzzi, P. and Schrefler, B. A. (2013 b). “A multiphase model for three-
dimensional tumor growth.” New Journal of Physics, 15(1), 015005.
Sciumè, G., Gray, W. G., Ferrari, M., Decuzzi, P., and Schrefler, B.A. (2013 c).
“On computational modeling in tumor growth.” Archives Comp. Methods
Eng., 20, 327-52.
Sciumè, G., Gray, W. G., Hussain, F., Ferrari, M., Decuzzi, P., and Schrefler B.
A. (2014 a). “Three phase flow dynamics in tumor growth.” Comp. Mech.,
53, 465-484.
Sciumè, G., Ferrari, M., and Schrefler, B. A. (2014 b). “Saturation–pressure
relationships for two-and three-phase flow analogies for soft matter.”
Mech. Res. Com., 62, 132-137.
Sciumè, G., Santagiuliana, R., Ferrari, M., Decuzzi, P., and Schrefler, B. A. (2014
c). “A tumor growth model with deformable ECM.” Phys. Biol., 11,
065004.
Simoni, L., and Schrefler, B. A. (2014). “Multi Field Simulation of
Fracture.” Chapter 4 in Advances in Applied Mechanics, Bordas, S.,
P., A., ed. Vol 47, AAMS, Academic Press, UK, 367-519.
© ASCE
i i
CONCREEP 10 “Template” — 2015/3/27 — 13:09 78

i i
A Few Analogies Between the Creep of Cement and of Other Materials
M. Vandamme
Université Paris-Est, Laboratoire Navier (École des Ponts ParisTech, IFSTTAR,

CNRS), 6-8 Av. B. Pascal, 77420 Champs-sur-Marne, France; PH (+33) 1 64 15 37
04; FAX (+33) 1 64 15 37 41; email: matthieu.vandamme@enpc.fr
ABSTRACT
Creep is not observed for cementitious materials only: most materials creep,
in particular materials very relevant for civil engineering applications, namely clay
or wood. Although each of those materials possesses its own specific features, some
analogies between their respective creep behaviors can be drawn. From these analogies,
we can try to obtain meaningful pieces of information on the physical processes at stake
during creep. After having introduced the topic, I present a few of those analogies and
discuss briefly what can be learned from them.
INTRODUCTION
Cementitious materials creep, i.e., slowly deform under the application of a
sustained load. Most materials also do, some of which are relevant for civil engineering
applications, e.g., clay, wood, or rock.
In terms of microstructure, cement, clay, and wood, exhibit several common
features (for a description of the multiscale structure of each material, see Constan-
tinides and Ulm (2004), Bobko and Ulm (2008), and Rafsanjani et al. (2012), respec-
tively): 1) they are heterogeneous, 2) they are multiscale, 3) they are porous, 4) their
porosity spans several orders of magnitude down to micropores (i.e., to pores with a
diameter smaller than 2 nm). However, a feature that significantly distinguishes cement
from clay or wood is that, in cement, the microstructure continuously evolves over time
because of hydration, while, in absence of any external solicitation, the microstructure
of clay or wood does not evolve with time.
At the scale of the nanometric building blocks of which they are made, cemen-
titious materials and clay-based materials exhibit strikingly similar features. Indeed,
C-S-H (i.e., calcium silicate hydrate) is made of Ca-O sheets covalently bonded with
silicate chains, which form stacks of parallel solid layers separated by a few layers of
water molecules. Clays are phyllosilicates made of stacks of layers combining silicate
tetahedra with aluminum octahedra: Again, those parallel solid layers are separated by
a few layers of water. Therefore, both C-S-H and clay are microporous layered sys-
tems at the nanometric scale. In contrast, at this same scale, wood is very different, as
the main layer (namely S2) of the cell wall is composed of stiff cellulose microfibrils
i i
© ASCE
i i
i i

i i
surrounded by a soft amorphous polymeric matrix made of hemicelluloses reinforced

with lignin.
Cement, clay, and wood all creep, and their creep properties have been studied
in depth. In the next sections, we underline some analogies in their creep behavior.
QUALITATIVE FEATURES OF CREEP

Cementitious materials are known to creep in a logarithmic manner with respect
to time in the long term. Such are clay-based materials, for which this logarithmic
feature is so intrinsic that it actually defines how creep experiments on those materials
are analyzed: A creep test on clay makes it possible to identify its so-called coefficient
of secondary compression (noted Cα ), defined as Cα = de/d(log(t), where de is a
small increment of the void ratio and is proportional to the volume strain classically
used in solid mechanics. In fact, this logarithmic feature of long-term creep is not
limited to clays only, but is observed for all soils, including sands. In soils, the non-
logarithmic short-term creep is known to be due to movements of water and dissipation
of pore pressure. In contrast, as is the case in cement science, what causes long-term
logarithmic creep is still under debate, even for sands. For those latter, whether such
feature is due to the behavior of individual contacts between grains or, in contrast,
whether it is the result of a collective behavior (e.g., some propose that creep of sand
is due to the progressive breakage of grains (Leung et al., 1996)) is unclear. In any
case, if the reason for a logarithmic creep is the same in cement and in clay, then this
observation teaches us that the origin of logarithmic creep is not due to a chemical
specificity of C-S-H (as the chemical structures of clay and C-S-H are different), and is
not due to hydration per se (as clays are usually not subjected to chemical processes).
An interesting observation is that, for clays, creep in tension is also a logarith-
mic function of time after a transient period. To reach this conclusion, we performed
creep experiments on self-standing clay films manufactured by deposition and evapo-
ration of montmorillonite (Carrier, 2013). The manufactured films (which were a few
dozens of microns in thickness) were solicited in tension and we measured the result-
ing strains by Digital Image Correlation: For the various films tested, creep became a
logarithmic function of time after a few seconds. This observation makes it possible
to discard the idea that logarithmic creep in clays could be explained by free-volume
dynamics as, under tension, the free volume (and thus the creep rate) should increase
with time.
Another qualitative feature of both scientific and practical interests is the effect
of the age of loading on the creep of cementitious materials. Even in absence of any
chemical process, i.e., in the hypothetical case where the material would be fully hy-
drated, one can wonder whether the age of loading should still have an impact. Accord-
ing to the microprestress theory (Bažant et al., 1997), which can successfully explain
why long-term creep is a logarithmic function of time, even if the hydration degree has
reached an asymptotic value, the age at loading still plays a role, as the microprestress
i i
© ASCE
i i
i i

i i
decreases over time. Therefore, according to the microprestress theory, a concrete that
would be fully hydrated should creep differently if creep testing is performed 1 year af-
ter pouring or 10 years after pouring. To the best of my knowledge, for clays, the age of
the tested sample is never taken into account, even though soil scientists not only work
with geologically old intact samples, but also with recompacted samples obtained by
consolidating a clay slurry directly in the lab.
SENSITIVITY OF MECHANICAL PROPERTIES TO WATER

The creep properties of cement, clay, and wood, depend on the hydric state:
As could be thought intuitively, the more humid the material, the more it creeps. Such
trend is observed for wood (Hunt, 1999), we observed it for C-S-H by microindentation
(Zhang, 2014) and also for our self-standing clay films solicited in tension (Carrier,
2013).
Actually, on those materials, the hydric state is known to affect not only the
creep, but also the short-term mechanical properties, namely swelling and elastic prop-
erties, in an amount that varies according to the material. Regarding the impact on those
properties, one can wonder what the most relevant parameter is to quantify the hydric
state. For wood, in the range of hydric states for which an effect is observed, dimen-
sional variations depend somewhat linearly on the moisture content (i.e., on the water
content), and the effect on the stiffness is also expressed as a function of the moisture
content (Glass & Zelinka, 2010). For the clay films we manufactured, we observed that
both their swelling and stiffness depended on the hydric state: however, although we
observed an hysteresis when plotting swelling and stiffness versus relative humidity,
the hysteresis fully disappeared when plotting those data versus water content (Carrier,
2013; Carrier et al., 2013). As is the case for wood, swelling depended roughly linearly
on water content. Those observations suggest that the main physical parameter to con-
sider when focusing on the effect of the hydric state should be the water content rather
than the relative humidity (although expressing constitutive laws in terms of relative
humidity also has its advantages). It is likely that such suggestion can be extended to
creep properties and cementitious materials.
THE PICKETT EFFECT

The Pickett effect, also called drying creep, is the fact that the creep of a sam-
ple submitted to a varying relative humidity is more important than what one would
expect from a mere addition of the effects of creep at constant relative humidity and of
drying shrinkage (Pickett, G., 1942). Interestingly, this effect is also observed in wood
(Armstrong and Kingston, 1960), under the name of mechano-sorptive creep. Actually,
creep enhancement due to moisture variations is quite general, as it is also observed in
keratin and synthetic or semi-synthetic polymers (Vlahinić et al., 2012). The fact that
Pickett-type effects are observed on so different materials shows that the Pickett effect
in cement is not due to a chemical specificity of C-S-H or to the fact that C-S-H is lay-
i i
© ASCE
i i
i i

i i
ered. A likely candidate to creep enhancement would be the presence of micropores,

which are present in all materials mentioned by Vlahinić et al. (2012).
Although models of drying creep can describe the hydric state through the in-
ternal relative humidity (Bažant & Yunping, 1994), based on the discussion in the pre-
vious section, one can wonder what the hydric driving force for the Pickett effect is.
As an alternative to variations of relative humidity, this driving force could be varia-
tions of local water content, or flow of water. Note that those latter two alternatives
are not equivalent: local variations of water content always require a flow of water, but
flow can occur without variations of relative humidity and/or water content. Such is
the case for a system with different faces exposed to different relative humidities: in
a steady state, one observes a flow of water but no variations of relative humidity or
water content. Actually, such experiment was performed on wood (Armstrong, 1972):
different relative humidities were imposed inside and outside the wood core and, in the
steady state, no creep enhancement was observed. For the variety of materials on which
Pickett-type effects are observed, Vlahinić et al. (2012) propose the driving force to be
the local flow of water from and to the microporosity.
CREEP MODELING
For the modeling of the creep behavior of clays, one approach which has been
particularly successful is the isotache approach (Šuklje, 1957; Leroueil, 2006), which
postulates that the strain rate ε̇ at a given time is only a function of the state of the
material (characterized by its actual strain ε) and of the stress σ to which the material
is submitted, i.e.:
ε̇ = R(ε, σ) (1)
where R is a function. For clays, such approach has been verified experimentally and
proves to be very useful: Indeed, the creep of clays is known to be nonlinear with
respect to the applied stress, so that, in contrast to cementitious materials, linear vis-
coelasticity (and thus the principle of superposition) is not relevant.
Interestingly, an equivalent approach has been proposed for cementitious mate-
rials, which is called the incremental model or “equivalent time method” (Acker et al.,
1989) and starts by postulating the exact same relation as Eq. (1). An appealing feature
of those 2 approaches for clay or cement-based materials is that the strain rate does not
depend per se on the full loading history, but only depends on the actual state of the
material (i.e., actual stress and strains), which, at least from a physical point of view,
makes sense. Note that linear viscoelasticity and isotache approach are not antagonistic
models: the isotache approach can also be used for linear viscoelastic materials.
Given the success of the isotache approach in the modeling of creep for clay
materials, revisiting the “equivalent time method” to the light of the isotache approach
could be of interest for the cement science community. In the same spirit, one could
imagine modeling the creep of concrete by starting from an isotache-type model of the
i i
© ASCE
i i
i i

i i
creep of C-S-H and upscaling it to the scale of concrete through micromechanics by

intermixing C-S-H with pores and several classes of elastic inclusions.
CONCLUSIONS
Although each type of materials possesses its own unique features, here we have
shown that some phenomenology observed on cementitious materials is not specific to
them but can be observed for clay or for wood, e.g., a logarithmic long-term creep, a
sensitivity of the mechanical properties to water, or the Pickett effect. Assuming that,
when the same phenomenology is observed for various types of materials, the under-
lying physical mechanisms are the same, we infer that: 1) The logarithmic feature of
long-term creep of cement is not due to the hydration process per se or to any chemical
specificity of C-S-H, since such feature is also observed with clay and 2) the Pickett
effect is not due to the layered structure of the C-S-H or to its specific chemistry, since
it is observed over a wide variety of materials.
When aiming at modeling the effect of water on mechanical properties, making
use of the water content as the parameter to describe the hydric state of the material
may have some advantages. Also, we observe that the “equivalent time method” for
cementitious materials is equivalent to the isotache approach for clays, which has been
proposed more than 50 years ago and is still of great value today for the soil science
community.
ACKNOWLEDGMENTS
I thank P. Acker for his constructive feedback on the manuscript.
REFERENCES
Acker, P., Lau, M. Y., & Collet, F. (1989). “Comportement différé du béton: validation
expérimentale de la méthode du temps équivalent.” Bulletin de Liaison Des
Laboratoires Des Ponts et Chaussées, 163, 31-39.
Armstrong, L. D. (1972). “Deformation of wood in compression during moisture move-
ment.” Wood Science.
Armstrong, L. D., & Kingston, R. S. T. (1960). “Effect of moisture changes on creep
in wood.” Nature, 185, 862-863.
Bažant, Z. P., Hauggaard, A. B., Baweja, S., & Ulm, F.-J. (1997). “Microprestress-
solidification theory for concrete creep. I: Aging and drying effects.” Journal of
Engineering Mechanics, 123(11), 1188-1194.
Bažant, Z. P., & Yunping, X. I. (1994). “Drying creep of concrete: constitutive model
and new experiments separating its mechanisms.” Materials and Structures,
27(1), 3-14.
Bobko, C., & Ulm, F.-J. (2008). “The nano-mechanical morphology of shale.” Mechan-
ics of Materials, 40(4-5), 318-337.
i i
© ASCE
i i
i i

i i
Constantinides, G., & Ulm, F.-J. (2004). “The effect of two types of C-S-H on the elas-
ticity of cement-based materials: Results from nanoindentation and microme-
chanical modeling.” Cement and Concrete Research, 34(1), 67-80.
Carrier, B. (2013). Effect of water on the short- and long-term mechanical properties of
swelling clays: experiments on self-standing films and molecular simulations,
Ph.D. thesis, Université Paris-Est.
Carrier, B., Wang, L., Vandamme, M., Pellenq, R. J.-M., Bornert, M., Tanguy, A., &
Van Damme, H. (2013). “ESEM study of the humidity-induced swelling of clay
film.” Langmuir: The ACS Journal of Surfaces and Colloids, 29(41), 12823-33.
Glass, S. V., & Zelinka, S. L. (2010). “Chapter 4: Moisture Relations and Physical
Properties of Wood.” In Wood Handbook - Wood as an Engineering Material.
General technical report FPL: GTR-190.
Hunt, D. G. (1999). “A unified approach to creep of wood.” Proceedings of the Royal
Society A: Mathematical, Physical and Engineering Sciences, 455(1991), 4077-
4095.
Leroueil, S. (2006). “The isotache approach. Where are we 50 years after its devel-
opment by Professor Šuklje? (2006 Prof . Šukljes Memorial Lecture).” In Pro-
ceedings of the XIII Danube-European Conference on Geotechnical Engineer-
ing. Ljubljana, Slovenia.
Leung, C. F., Lee, F. H., & Yet, N. S. (1996). “The role of particle breakage in pile
creep in sand.” Canadian Geotechnical Journal, 33(6), 888-898.
Šuklje, L. (1957). “The analysis of the consolidation process by the isotache method.”
In Proceedings of the 4th International Conference on Soil Mechanics and
Foundation Engineering. Vol. 1. London, England.
Vlahinić, I., Thomas, J. J., Jennings, H. M., & Andrade, J. E. (2012). “Transient creep
effects and the lubricating power of water in materials ranging from paper to
concrete and Kevlar.” Journal of the Mechanics and Physics of Solids, 60(7),
1350-1362.
Zhang, Q. (2014). Creep properties of cementitious materials: effect of water and mi-
crostructure. An approach by microindentation, Ph.D. thesis, Université Paris-
Est.
i i
© ASCE
i i
CONCREEP 10 84
Useful Fundamentals of Shrinkage and Creep of Concrete
Folker H. Wittmann1
1
Aedificat Institute Freiburg, D-79100 Freiburg, Schlierbergstr. 80; and Swiss Federal
Institute of Technology Zurich (ETHZ). E-mail: wittmann@aedificat.de
Abstract
In this contribution selected fundamental properties of concrete are presented

and discussed. Basic mechanisms of shrinkage are first described in some detail. With
increasing age of concrete different mechanisms of shrinkage become activated.
During the first hours after mixing dissolution of cement is the main shrinkage
mechanism. Depending on the humidity of the surrounding air capillary pressure in
the pore water may cause serious damage in the first hours. After this initial loss of
water, the remaining aqueous solution is retreated in the free spaces between fine
aggregates and first hydration products. The role of disjoining pressure is pointed out
in detail. In this context it becomes clear that the influence of capillary pressure on
shrinkage of hardened concrete can be neglected. In equilibrium with RH lower than
50 % the dominating mechanism is the increasing surface energy of the drying gel
particles. Shrinkage of high strength concrete has a totally different time evolution as
the time-dependent moisture distribution in aging high strength concrete is
completely different. If high strength concrete is exposed to an environment with RH
higher than 80% there will be no shrinkage at all but swelling. Creep of concrete can
be described in a realistic way by means of rate theory. Elements of rate theory are
briefly described in the last section. Prediction of creep can be significantly
improved if these useful fundamentals are taken into consideration.
INTRODUCTION
There are few properties of concrete, which have been presented and
discussed in such a contradictory way as creep and shrinkage. For quite some time
creep was characterized by a simple creep number that means the elastic deformation
was multiplied by a factor of 2 to 3 to obtain a so-called total deformation. Shrinkage
was given as a basic value, which depends on the environmental humidity and a
factor, which takes the dimensions of the structural element into consideration. In the
meantime the complex nature of creep and shrinkage was widely realized and more
sophisticated models for prediction of creep and shrinkage in real structures were
developed. In this way material properties can be described more precisely and
hopefully some disasters which happened in the past due to underestimation of
shrinkage and creep can be avoided in the future. There are fundamentals, which are
of interest for scientists exclusively and there are fundamentals, which may be helpful
for practitioners.
© ASCE Rev. 11/2014

CONCREEP 10 85
In the meantime several codes for prediction of creep and shrinkage were
published and they are permanently further checked and refined. Model B3, based on
several precursors, was first published as a RILEM Draft Recommendation in 1995
(RILEM Draft Recommendation, 1995). Statistical aspects and the sensitivity as well
as the theoretical basis were further outlined in two follow-up papers (Bazant, Z. P.,
and Baweja, S., 1995(1)) and (Bazant, Z. P., and Baweja, S., 1995(2)). Another code
was prepared by ACI committee 209 (ACI committee 209, 1992). And a third
alternative is the CEB-FIP Model Code (CEB-FIP Model Code, 1990). In fact there
exist still more models all with different advantages and shortcomings. All codes are
critically checked, compared, and improved on the basis of ever growing databases
(see for example: Akthem Al-Manaseer, and Armando Prado, 2015). At the moment
the most comprehensive database is probably the NU-ITI database (Bazant, Z.P., and
G.-H. Li, 2007).
Scatter plots of measured versus predicted values of shrinkage of concrete are
shown for instance in Bazant Z. P., and Baveja S., 1995(1). Prediction models predict
the average value reasonably well. The average value as measured on many different
types of concrete is not necessarily of primary interest. The type and amount of
cement, the water-cement ratio, and the type of aggregates is not taken into
consideration adequately. If we consider for instance in the scatter plots of the models
a predicted shrinkage value of 0.4·10-3 the measured values may vary between 0.2
and 0.6·10-3. Because of this enormous scatter the value for long term prediction is
limited in practice. If, however, some characteristic properties of the material, which
can be deduced from fundamental mechanisms of creep and shrinkage of concrete,
are taken into consideration in addition, this will be helpful to obtain more realistic
results by applying predictive models. It has already been suggested by several
authors that short-time tests followed by extrapolation will be necessary to reduce the
enormous scatter of data obtained from the different prediction models (see for
example: Bazant, Z. P., and Li, G.-H., 2008). Results of short-time tests, however,
should be evaluated carefully and understood on the basis of real mechanisms of
creep and shrinkage. Then and then only the prediction will be based on a more solid
basis and predictions will become more reliable.
In this contribution dominant mechanisms of shrinkage and creep shall be
briefly presented. On the basis of real mechanisms it will be possible to predict the
influence of a number of parameters on creep and shrinkage. This may help to make
predictions of model codes even more reliable. Because of obvious limitations of
time and space a selected number of examples can be presented only.
SHRINKAGE
Dissolution and capillary shrinkage
Fresh concrete behaves in the first few hours like a mixture of fine and coarse
particles mixed with water. During this early phase the volume of the mix is reduced
by evaporation and in particular by dissolution of cement in water. Both evaporation
and the dissolution shrinkage lead to an early volume reduction of the fresh mix,
which in most cases leads to a vertical movement of the top surface (Beltzung F., and
© ASCE Rev. 11/2014

CONCREEP 10 86
Wittmann, F. H., 2001). Dissolution shrinkage may be of particular interest for the
technology of steel pipes filled with concrete for instance. But it can also create
problems when the fresh concrete is placed in complex formwork.
If evaporation of water continues the water film on the surface will be
replaced by menisci between the particles near the surface. A capillary pressure pc
will be created in the water under these menisci:
ଶఙ௖௢௦ఏ
‫݌‬௖ ൌ (1)
௥
ı in equ.(1) stands for the surface energy, ș for the wetting angle and r
represents the average radius of the menisci. The capillary shrinkage of fresh concrete
under the drying surface is shown in Fig. 1. It can be seen that initially the capillary
shrinkage follows the evolution of the capillary pressure but after few hours capillary
pressure collapses and capillary shrinkage approaches a final value. At this moment
the aqueous system in the mix becomes unstable and the remaining water moves into
the spaces between the particles (Wittmann, F. H., 1976). By the way this sudden
transition can be visually observed easily on the surface of concrete.
Figure 1. Evolution of capillary pressure p in fresh concrete and resulting

capillary shrinkage İ as function of age of concrete (Wittmann F. H., 1976).
It is a wide spread heresy that capillary pressure is a major mechanism of

shrinkage of hardened concrete (Wittmann, F. H., 2009). Wide cracks can be formed
by capillary pressure in the first hours after compacting fresh concrete, but these
cracks can be easily avoided by keeping the surface of fresh concrete wet (Slowik et
al. 2009). But after few hours capillary shrinkage becomes negligible.
Figure 2. Water layer between two spherical particles
In Figure 2 two spherical particles are shown which are connected by a thin
water layer. We may consider fresh concrete to be built up by many of these elements
forming a three-dimensional network. The free water layer between the solid particles
© ASCE Rev. 11/2014

CONCREEP 10 87
can be characterized by two Radii R1 and R2 as shown in Fig. 2. From this geometry
follows a capillary pressure Pc in the water:
ଵ ଵ
ܲ௖ ൌ ߪܿ‫ߠݏ݋‬ሺ ൅ ሻ (2)
ோభ ோమ
If water evaporates from this system both radii R1 and R2 decrease and as a
consequence the capillary pressure Pc increases. As capillary pressure increases, the
surface of contact between the liquid and the solid particles decreases. It can be
shown that under these conditions the attractive force Fa between the two particles
remains practically constant (Schubert H., 1972). The green strength of the young
concrete also remains constant. That clearly shows that other mechanisms than
capillary pressure must be at the origin of drying shrinkage. We will discuss most
important mechanisms in the following section.
Drying Shrinkage, influence of disjoining pressure
Let us first have a look at results of an experiment to observe the action of

disjoining pressure. In thin water layers between two particles disjoining pressure is
created. The theoretical background of disjoining pressure has been studied in great
detail (see for instance Churaev, N. V., and Derjaguin, B. V., 1985). Splittgerber has
visualized the action of disjoining pressure (Splittgerber, H., 1974, and Splittgerber,
H., 1976). The experimental set-up of Splittgerber is shown schematically in Fig. 3
(left). A thin quartz plate is kept at constant distance from a polished quartz block at
the right end, while the left end is free. In the dry state the left end is in direct contact
with the quartz block and from the bending of the plate van der Waals attractive force
can be determined. When the relative humidity of the surrounding air is increased
above 55 % the thin glass plate is separated from the support by a thin water layer
which squeezes in between y capillary condensation. The increasing distance with
increasing RH is shown in Fig. 3 (right).
Figure 3. (left) Thin glass plate separated from the quartz support by disjoining
pressure of water; (right) Distance from the free end of the thin glass plate from
the support as function of RH. The adsorption and desorption branches are
shown.
In Fig. 3 (right) a pronounced hysteresis between the adsorption and

desorption branches can be observed. This explains essentially the origin of the
hysteresis of sorption isotherms as measured on hardened cement paste or concrete.
Based on this information we have to subdivide the influence of RH on properties of
cement-based materials into two regions: (I) Below 40 % RH adsorbed water reduces
the surface energy of gel particles. This leads to expansion of all particles. (II) At
higher RH the disjoining pressure modifies the nano-structure of the material by
separating existing gaps. This change of the system is also reflected in the change of
strength and elastic modulus of the material. Strength of dry concrete decreases with
© ASCE Rev. 11/2014

CONCREEP 10 88
increasing RH until a value of about 50 % is reached, because surface energy

decreases. At higher RH strength decreases even more due to the action of disjoining
pressure.
In Fig. 4 hygral length change is plotted as function of RH. The observed
length change can be subdivided into two parts. (I) At low RH adsorbed water
reduces surface energy of the gel particles. All particles expand until RH reaches a
value around 50%. (II) At higher RH additional length change is due to the action of
disjoining pressure (dashed zone in Fig. 4). In systems where capillary pressure is at
the origin of shrinkage and swelling, shrinkage and not swelling is observed when
capillary condensation takes place.
Figure 4. Hygral length change of hardened cement paste as function of RH.
It is well-known that disjoining pressure can be modified by the type and

concentration of ions dissolved in the pore water. If disjoining pressure increases
shrinkage will increase. In Fig. 5 the total shrinkage of a given type of concrete is
plotted as function of the alkali content in the pore water. The final shrinkage
increases approximately linearly with increasing alkali content. This observation
cannot be explained by the action of capillary pressure. Taking the influence of
disjoining pressure into consideration shrinkage can be modified and reduced within
limits by choosing cement, which contains a small amount of alkali only. This will
also help to use shrinkage prediction in a more realistic way of course.
Figure 5. Final shrinkage of concrete as function of the alkali content

(Ŷ Na2O and Ɣ K2O), (F. Beltzung et al., 2001).
© ASCE Rev. 11/2014

CONCREEP 10 89
A striking example of the role of disjoining pressure in the complex process of

shrinkage can be observed in the so-called Chichibu GRC Cement. This cement is
composed in such a way that the portlandite, which is formed during hydration of C3S
and C2S is progressively consumed by secondary pozzolanic reactions and formation
of ettringite. Concrete and mortar fabricated with this special cement has extremely
low shrinkage. If, however, Ca(OH)2 is added to the mix, the usual shrinkage
deformation is observed. Surface tension of pore water is not changed by Chichibu
GRC Cement, nor is it changed by the addition of Ca(OH)2. Shrinkage can be
changed within wide borders by the influence of Ca ions on disjoining pressure.
Shrinkage of high performance concrete
So far we have considered dissolution shrinkage, capillary shrinkage and

drying shrinkage. In looking into the different mechanisms we could find ways to
modify under given circumstances the prediction of model codes and come to more
realistic predictions. With the advent of high strength concrete another important
aspect became apparent. In high strength concrete a considerable amount of the
mixing water is consumed by hydration of cement. In normal concrete the relative
humidity in the pore space may be reduced to 95 % at most. Under these conditions
an influence on shrinkage exists but it remains modest. With high strength concrete,
however, RH in the pore space may be reduced to values below 80 % at an early age.
Under these conditions self-desiccation and autogenous shrinkage can be observed. In
Fig. 6 results of Miyazawa S. and Tazawa E. ,2001, are shown. It can be seen that
shrinkage of high strength concrete differs fundamentally from shrinkage of normal
concrete. Due to hydration of cement a considerable amount of mixing water is
consumed after seven days already. This rapid self-desiccation leads to considerable
shrinkage at very early age. After seven days high strength concrete has reached
shrinkage deformation, which can be observed on normal concrete after several years
only.
Figure 6. Self-desiccation and autogenous shrinkage of high strength concrete,

(S. Miyazawa and E. Tazawa, 2001)
© ASCE Rev. 11/2014

CONCREEP 10 90
If this type of concrete is exposed to an environment with RH = 80 % no

shrinkage can be observed at all, but in contrast the material absorbs humidity from
the environment and swells. It will swell even more if placed in an environment with
RH of 90 %. Obviously in this case predictions of most model codes are far from the
observed deformations. To understand the fundamental differences of shrinkage of
normal and high strength concrete we have to look into the moisture distributions as
function of time. In Fig. 7 the distributions of relative humidity of normal strength
and high strength concrete are shown for the first five years after casting (Alvaredo,
A. M., and Wittmann, F. H., 1995). In this case the concrete is supposed to be
exposed to an environment with an average relative humidity of 50 %. The concrete
sample is supposed to have a thickness of 500 mm. The quick self-desiccation of high
strength concrete leads to a decrease of RH to approximately 80 % over the entire
cross section after two weeks and finally a value of slightly more than 70 % is
reached after six months. The RH will not decrease further in the center because the
hydration of cement is practically stopped at this low relative humidity. The
remaining pore water is not available for further hydration of cement. The slow loss
of water by drying continues of course for many years and finally equilibrium with 50
% RH will be reached. Based on the time-dependent moisture distributions shown in
Fig. 7 shrinkage of specimens made with the two types of concrete can be
numerically determined. Results are shown for the first 5 years in Fig.8 (left). It is
obvious that the time-dependence of shrinkage is totally different. From these results
follows that it is not realistic to assume a common shrinkage law for concrete with
significantly different water-cement ratio.
If the drying specimens of normal and high strength concrete are fixed at both
ends a time-dependent restraint reaction can be observed. Results of numerical
predictions are shown in Fig.8 (right). The normal strength concrete will fail under
tension after 500 days of drying. The high strength concrete, however, will fail after
13 days already because of the quick endogenous drying process.
Figure 7. Moisture distribution in normal and high strength concrete (curves

plotted with circles) in the first five years
© ASCE Rev. 11/2014

CONCREEP 10 91
Figure 8. (left) Evolution of shrinkage of normal (n.c.) and high performance

concrete (h.p.c.) as determined along the centre of the specimens; (right)
Evolution of restraint reaction of fully restraint members of normal (n.c.) and of
high performance concrete (h.p.c.)
CREEP
Creep of different materials such as metals, ceramics, and polymers can be

described in a realistic way by rate theory. This very general approach can also be
applied to cement-based materials (Wittmann, F. H., 1977, Klug, P. and Wittmann, F.
H., 1974, Wittmann, F. H. and Lukas J., 1974). On this basis creep strain İ as function
of time t can be expressed by the following equation:
ொ ௏
ߝ ൌ ܽ‫ ݐ‬௡ ቀെ ቁ ቀ ߪቁ (3)
ோ் ோ்
In Eq.(3) Q stands for the activation energy and V for the activation volume of
the material. The activation energy of concrete is in most cases approximately 4.5
kcal/Mol. The activation volume varies with water-cement ratio and with humidity
content. An average value of 10-17 mm3 has been determined (Klug, P. and Wittmann,
F. H., 1974)
Rel. Humidity
Figure 9. a0 and n of Eq. (4) as function of relative humidity
© ASCE Rev. 11/2014

CONCREEP 10 92
If creep at constant temperature and at moderate applied stress is to be

described, Eq.(3) can be simplified:
ߝ ൌ ܽ଴ ‫ ݐ‬௡ ߪ (4)
Under constant moisture content creep follows a simple power law. But the
moisture content has a significant influence on the activation volume. The lower the
moisture content, the lower will be the creep strain. Values for ܽ଴ and n of Eq. (4) are
plotted in Fig. 9. As expected, creep of saturated specimens is at maximum. In high
strength concrete the equilibrium humidity in the mass may be reduced after one
week already to values around 75 %. According to Fig. 9 creep of high strength
concrete will then be reduced to one third of the value at saturation.
CONCLUSIONS
Shrinkage of porous materials such as concrete is a complex process based on

a number of different mechanisms. The most important mechanisms at an early stage
are dissolution shrinkage and capillary shrinkage. In the hardened state of concrete
shrinkage is essentially due to loss of water in the pore space. Water can get lost by
drying to the environment and by internal drying by continuing hydration (self-
desiccation). Disjoining pressure in concrete depends strongly on the amount and
type of ions dissolved in the pore water. When all these mechanisms are taken into
consideration prediction of shrinkage by model codes can be significantly improved.
For high strength concrete and for concrete produced with blended cements existing
model codes are applicable with certain limits only. When obvious fundamental
properties are taken into consideration, model codes can be modified for given
applications and for special types of concrete they can be made more realistic.
REFERENCES
ACI committee 209, “Prediction of creep, shrinkage and temperature effects in

concrete structures”, ACI 209 R-92, American Concrete Institute, Detroit.
Al-Manaseer A., and Prado A., (2015) “Statistical comparison of creep and shrinkage
prediction models using RILEM and NU-ITI Databases, ACI Material Journal
112(1) 125-135
Alvaredo, A. M., and Wittmann, F. H.,(1995) “Shrinkage and cracking of normal and
high performance concrete”, in High Performance Concrete: Material
Properties and Design, Wittmann F. H. and Schwesinger P., editors,
Aedificaatio Publishers, Freiburg, 91-110
Bazant, Z. P., and Baweja, S., (1995(1)) “Justification and Refinements of Model B3
for concrete creep and shrinkage, 1. Statistics and sensitivity” Materials and
Structures 28, 415-430.
Bazant, Z. P. and Baweja, S., (1995(2) “Justification and refinements of model B3 for
concrete creep and shrinkage, 2. Updating and theoretical basis” Materials and
Structures 28, 488-495.
Bazant, Z. P., and Li, G.-H., (2007) “NU Database of Laboratory Creep and
Shrinkage Data” available for download from
www.civil.northwestern.edu/people/Bazant.html.
© ASCE Rev. 11/2014

CONCREEP 10 93
Bazant, Z. P., and Li, G.-H., (2008) “Unbiased statistical comparison of creep and
shrinkage prediction models” ACI Materials Journal 105(6) 610-621.
Beltzung, F., and Wittmann, F. H., (2001) “Early chemical shrinkage due to
dissolution and hydration of cement” Materials and Structures 34, 279-283
Beltzung, F., Wittmann, F. H., and Holzer, L.,(2001) „Influence of composition of
pore solution on drying shrinkage“ in Creep, Shrinkage and Durability
Mechanics of Concrete and other Quasi-brittle Materials, edited by Ulm F.-J.,
Bazant Z. P. and Wittmann F. H., Elsevier Science Ltd., 39-48
Beltzung, F., and Wittmann, F. H., (2005) “Role of disjoining pressure in cement
based materials” Cement and Concrete Research 35, 2364-2370.
CEB-FIP Model Code, Design Code, (1990), Thomas Telford, London,
Churaev, N. V., and Derjaguin, B. V., (1985) “Inclusion of structural forces in the
theory of stability of colloids and films” J. Coll. Interf. Sci. 103(2), 542-553
Klug, P. and Wittmann, F.H., (1974) “Activation energy and activation volume of
creep of hardened cement paste”, Mat. Sci. and Eng. 15, 63-66
Miyazawa, S., and Tazawa, E., (2001) “Prediction model for shrinkage of concrete
including autogenous shrinkage”, in Creep, Shrinkage and Durability
Mechanics of Concrete and other Quasi-brittle Materials, edited by Ulm F.-J.,
Bazant Z. P. and Wittmann F. H., Elsevier Science Ltd., 735-740
RILEM Draft Recommendation, Projets de Recommendation de la RILEM, (1995)
107-GCS Guidelines for the Formulation of Creep and Shrinkage Prediction
Models, Creep and Shrinkage Prediction Model for Analysis and Design of
Concrete Structures – Model B3, Materials and Structures 28, 357-365.
Schubert, H.,(1972) “Untersuchungen zur Ermittlung von Kapillardruck und
Zugfestigkeit von feuchten Haufwerken aus körnigen Stoffen“, Dissertation,
Karlsruhe University of Technology.
Schubert, H., (1992) “Kapillarität in porösen Feststoffsystemen“, Springer Verlag.
Slowik, V., Neumann, A., Dorow, J., and Schmidt, M., (2009) “ Early age cracking
and its influence on durability of concrete structures” in Creep, Shrinkage and
Durabiity Mechanics of Concrete and Concrete Strucutres, Tanabe T., Sakata,
K., Mihashi, H., Sato, R., Maekawa, K., and Nakamura, H., editors, CRC
Press, 471-477.
Splittgerber, H., (1974) “Einfluss adsorbierter Wasserfilme auf die van der Waals
Kraft zwischen Quarzglasoberflächen“, Surface Science 41, 504-514
Splittgerber, H., (1976) “Spaltdruck zwischen Festköpern und Auswirkung auf
Probleme in der Technik“, Cement and Concrete Research 6, 29-36.
Wittmann, F. H., and Lukas, J., (1074) “The application of rate theory to time-
dependent deformation of concrete “, Mag. of Concr. Res. 26, 191-197.
Wittmann, F. H., (1976) “On the action of capillary pressure in fresh concrete”
Cement and Concrete Research 6, 49-56
Wittmann, F. H., (1977) “Grundlagen eines Modells zur Beschreibung
charakteristischer Eigenschaften des Betons“ , Deutscher Ausschuss für
Stahlbeton, Schriftenreihe Heft 290, 43-101
Wittmann, F. H., (2009) “Heresies on shrinkage and creep mechanisms” in Creep,
Shrinkage and Durability Mechanics of Concrete and Concrete Structures,
Tanaba et al., editors, Taylor and Francis Group, London, 3-9.
© ASCE Rev. 11/2014

Coupled Effects Between Damage and Permeability With A View To Discrete Modelling

Uploaded by

Document Informationclick to expand document informationcreep on concrete

Document Informationclick to expand document information

Copyright:

Available Formats

Coupled Effects Between Damage and Permeability With A View To Discrete Modelling

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

Coupled Effects Between Damage and Permeability With A View To Discrete Modelling

Uploaded by

Copyright:

Available Formats

CONCREEP 10 49

Coupled Effects between Damage and Permeability with a View to Discrete

G. Pijaudier-Cabot; F. Khaddour; D. Grégoire; and L. Ecay

Laboratoire de Fluides Complexes et leurs Réservoirs, UMR 5150 UPPA – CNRS –

The purpose of this work is to achieve a better understanding of the

In this work, we aim at a better understanding of the relationship between damage,

STOCHASTIC HIERARCHICAL BUNDLE MODEL

‫ܮ‬௘ ൌ యඥܸ௧ , ‫ܮ‬௖ ൌ ܶǤ ‫ܮ‬௘ T = tortuosity (2)

(a). Discretization of pores (b). Random assembly

Some adjustments need to be discussed with respect to the mercury intrusion

Figure 2.a presents a typical PSD response of a cementitious material (mortar). We

Fig. 2. Method of eliminating of the noise at low pressure for an undamaged

FLUID FLOW IN THE PORE NETWORK

From a given hierarchical assembly of capillary pores, the objective is now to

Poiseuille’s flow at the micro-scale

ͳ ߨ ݀௜ସ ሺܲଵଶ െ ܲଶଶ ሻ

Knudsen flow at the micro-scale

whereοܲ ൌ ܲଵ െ ܲଶ is the difference between the upstream and downstream

Macroscopic transport properties

For hierarchical assembly of capillaries, the transport properties are obtained by

EXPERIMENTAL DATA AND COMPARISONS WITH THE MODEL

Experiments have been performed on mortars specimens with water/cement ratio of

The testing apparatus consists in a conventional electro-mechanical compressive

considered, corresponding to several levels of damage ranging from 0% to 12% (in

Fig. 3. Evolution of pore size distribution upon damage.

Figure 4 present a comparison between the measured and predicted intrinsic

Fig. 4. Comparison between the measured and predicted intrinsic permeability

Figure 5 shows a comparison between the apparent permeability to nitrogen and to

CONCLUSIONS AND EXTENSIONS

A theoretical model aimed at computing the apparent permeability of a porous

scheme in which bundles of capillaries are assembled randomly. Within each

Fig. 6: Separation of He/CH4 mixture (predicted and measure experimentally)

Further developments are currently being considered, investigating the case of a

1. Kozeny, J. 1927. Über Kapillare Leitung Des Wassers Im Boden [About

Institut Français des Sciences et Technologies des Transports, de

This paper presents a summary of recent experimental researches about

The international literature on basic creep of concrete is considerable.

EXPERIMENTAL EVIDENCES RELATED TO THE BASIC CREEP OF

The experimental studies, above referenced, led to the following

- The basic creep strain is directly leaked to the creation of microcracks

All these experimental evidences leaded to a proposal concerning a

Under the effect of creep loading, concrete undergoes a micro-cracking

These experimental evidences lead to these logical consequences: the

A NEW NUMERICAL MODEL TO DESCRIBE THE BASIC CREEP OF

- Be able to describe explicitly the cracking process during the creep

A numerical model consistent with the respect to the explicit modelling of

The probabilistic explicit cracking model. The numerical model is based on

- Concrete is a heterogeneous material in which the heterogeneities can

Figure1. Criterion of crack creation (explicit probabilistic cracking model)

- a direct one which consists in introducing, in the volume elements

The cracking process obtained by the use of the probabilistic explicit

Remarks. The residual strain observed when a creep specimen is unloaded is

EXPERIMENTAL STRATEGY TO QUANTIFY THE MODEL

It is needed to define an experimental strategy for determining the input

- The ability of the concrete to micro-cracking (creating of greater or

Taking into account these material parameters the mechanical test to be

- Compressive basic creep tests on specimens of any dimension (the basic