Engineering Structures 33 (2011) 34423453

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Thermohygromechanical modelling of self-induced stresses during the service

life of RC structures
Miguel Azenha a, , Carlos Sousa b,1 , Rui Faria b,2 , Afonso Neves b,3
a

ISISE Institute for Sustainability and Innovation in Structural Engineering, University of Minho, School of Engineering, Civil Engineering Department, Azurm Campus, 4800-058
Guimares, Portugal
b

LABEST Laboratory for the Concrete Technology and Structural Behaviour, Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal

article

info

Article history:
Received 26 November 2010
Received in revised form
21 April 2011
Accepted 1 July 2011
Available online 2 August 2011
Keywords:
Cement hydration
Service life conditions
Differential shrinkage
Cracking
Numerical simulation

abstract
Current practices of structural design in reinforced concrete (RC) structures usually account for stresses
caused by phenomena such as heat of hydration and drying shrinkage in a quite simplified manner. The
present paper aims to evaluate the consequences of explicitly considering self-induced stresses, which
actually vary significantly within structural cross-sections, combined with stresses caused by external
loads. The used numerical framework involves the explicit calculation of the temperature field in concrete,
with proper account for the heat of hydration of cement. Simultaneously, the moisture field in concrete is
computed in order to ascertain the relative humidity changes in the pore structure caused by drying, and
the inherent shrinkage strains. Stress calculations are made with due consideration of the evolution of
mechanical properties of concrete as a function of the equivalent age, as well as relevant phenomena like
creep, concrete cracking and influence of reinforcement. Two separate groups of numerical applications
are presented, checking influence of the self-induced stresses: a unrestrained concrete prism usually used
for shrinkage measurement, and concrete slabs subjected to external loads. Particularly for the second set
of applications, the obtained results (with explicit consideration of the differential effects of self-induced
stresses) are compared, in terms of cracking loads and crack propagation, to those that would be obtained
by using the simplified design approach based on considering uniform shrinkage fields in concrete. It is
found that the behaviour of both formulations is quite similar after crack stabilization, but may be quite
distinct in the crack propagation phase.
2011 Elsevier Ltd. All rights reserved.

1. Introduction
The existing methodologies for practical analysis and design of
RC structures comprise several simplifying assumptions, namely
concerning the stress state induced by heat of hydration and
drying shrinkage. Actually, due to the exothermic nature of
cement hydration reactions, concrete members endure nonuniform internal temperature distributions, usually with hotter
regions in the core and cooler ones close to the external surfaces.
Total or partial restrictions to the volumetric deformations
associated to these thermal fields induce stresses in concrete,
whose importance is usually disregarded by designers, most
of the time without a quantitative notion of their magnitude.

Corresponding author. Tel.: +351 93 840 4554; fax: +351 253 510 217.
E-mail addresses: miguel.azenha@civil.uminho.pt, miguel.azenha@gmail.com
(M. Azenha), cfsousa@fe.up.pt (C. Sousa), rfaria@fe.up.pt (R. Faria),
asneves@fe.up.pt (A. Neves).
1 Tel.: +351 22 508 1914.
2 Tel.: +351 22 508 1950.
3 Tel.: +351 22 508 1879.
0141-0296/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.07.008

Their quantification is, however, possible with recourse to

thermomechanical methodologies that allow the computation
of temperatures induced by heat liberation, and using them as
thermal loads in the numerical simulation of stress fields.
Regarding drying shrinkage, and according to code provisions [1,2], its structural effects are taken into account by using
reference shrinkage strains (based on the geometry of the concrete
member, environmental relative humidity and concrete grade
class), which thereafter are applied uniformly to the structure
in order to ascertain the corresponding stresses. This is clearly a
simplification for practical calculations, since drying occurs nonuniformly in concrete, which generates stresses in the presence of
restraints (which are external, caused by reinforcement or due to
cross-sectional effects). The quantification of the differential drying shrinkage calls for the necessity of knowing the moisture fields
within concrete, as well as the relation between moisture losses
and the corresponding volumetric variations. Since most generally
RC structures suffer effects of both heat of hydration and drying
shrinkage, it is important to have numerical simulation tools that
allow the computation of thermal, moisture and mechanical fields
for the computation of stress states during the service life.

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

Several authors conducted research in the field of numerical

simulation of self-induced stresses during the service life of structures. Kwak et al. [3], Kwak and Ha [4] studied non-structural
cracking in RC walls using an approach similar to the one adopted
in this paper, with explicit consideration of non-uniform temperature and moisture distributions, aiming the evaluation of
post-cracking behaviour. Bernard and Brhwiler [5] and Martinola
et al. [6] have explicitly calculated moisture fields and the corresponding drying shrinkage in the evaluation of debonding potential between concrete layers cast at different ages. Grasley et al. [7]
monitored internal RH in concrete specimens and evaluated the
self-induced stresses in both linearly restrained and free specimens. Oh and Cha [8], Oh and Choi [9] established a numerical
framework for analysis of thermal, moisture and mechanical fields,
applying it to the study of concrete decks of composite bridges.
Chen and Mahadevan [10] used a thermohygromechanical
framework for the cracking analysis of a 1 1 0.5 m3 concrete
block using a smeared approach. Granger et al. [11] used a hygromechanical approach to evaluate skin micro-cracking of concrete with recourse to contact analyses. Gawin et al. [12] devised
a numerical framework for thermohygro analysis in which concrete stresses due to creep and shrinkage were calculated using the
effective stress concept.
In regard to predicting the service life performance of ordinary
RC structures with advanced numerical techniques, none of
the above mentioned works focused however on addressing
comparisons with results obtained with simplified approaches
commonly adopted in design (based on codes such as Model Code
1990 [2]), to evaluate how RC structures are designed nowadays.
This is precisely the main purpose of the present paper,
commencing with a general description of the developed thermalhygro finite element (FE) framework, which is thereafter used
with the mechanical part of the FE code DIANA [13].
Two sets of applications are considered subsequently. The
first one concerns to the analysis of a concrete drying shrinkage prism, to check relevance of the self-induced stresses due
to thermalhygro phenomena. The second example focuses on
the behaviour of RC slab-like members, where influence of nonuniformity of the self-induced stresses due to heat of hydration
and drying shrinkage are evaluated with the members performing under different conditions: (i) under imposed longitudinal deformations, or (ii) under imposed curvatures applied at distinct
ages. Conclusions are extracted in regard to the influences of selfinduced stresses on the cracking loads, as well as on the performances of the RC members during the crack propagation phases.

2.1. Thermal submodel

Computation of the transient temperature fields in concrete is
made through the implementation in a FE code of the heat balance
equation
k ( T ) + Q = c T

(1)

where k is the thermal conductivity, c is the volumetric specific

heat and T is the temperature. Q is the volumetric heat generation
rate due to cement hydration, formulated as an Arrhenius type
law [14]
Q = Af () e

Ea
RT

cement hydration), R = 8.314 J mol1 K1 is the Boltzmanns

constant and f () is a normalized function for heat.
The formulation of boundary conditions
qT = h (Tb Te )

(2)

where A is a rate constant, Ea is the apparent activation energy, is

the degree of heat development (ratio between the heat Q released
up to time t and the total heat Qfinal released upon completion of

(3)

where qT is the heat flux per unit area, Tb is the boundary surface
temperature and Te is the environmental temperature, comprises
a mixed convectionradiation boundary transfer coefficient h [15].
The calculated temperatures are used as inputs for the mechanical model, where thermal strains of concrete are computed accounting for the material coefficient of thermal expansion (T ).
Even though this coefficient is known to be variable during the first
hours after casting [16,17], the applications envisaged in this paper
do not involve significant temperature changes at early ages, and
thus it was decided to use a constant value for T .
Also, the evolution of temperatures at each point in the structure allows computing the equivalent age teq (useful for updating
the age-dependent mechanical properties), based on the following
Arrhenius-type equation
t

teq =

ERa

1
T1
T ( )
ref

(4)

where Tref stands for the reference temperature (usually 20 C).

Specific details regarding the numerical implementation of this
formulation using the FE Method can be found elsewhere [18].
2.2. Moisture submodel
Modelling of the moisture states inside porous materials such as
concrete requires the selection of an appropriate driving potential.
Several approaches exist, using the pore pressure, the pore relative
humidity (H), the global water content, or even separating the
vapour water from the liquid water. For the applications discussed
herein the averaged pore relative humidity H has been selected
as the driving potential, and accordingly the moisture state in the
pore structure is expressed according to equation [19]

= (DH H ) + H S
H

(5)

S
where 0 H 1, DH is the moisture diffusion coefficient and H
relates to the H drop in the pore structure due to self-desiccation.
As the applications addressed in this paper pertain to ordinary
concrete, with relatively high water-to-cement ratios, the effect of
S has been neglected. Coefficient DH was considered to depend on
H
H as defined in MC90 [2]
1 H

DH = D1 H +

2. Thermohygromechanical model

3443

1 + [(1 H ) / (1 HC )]n

with H =

D0
D1

(6)

where D1 and D0 are the values of DH for H = 1 and H = 0,

respectively, HC is the pore relative humidity at which DH = 0.5D1
and exponent n is a material property.
Moisture fluxes qH across the boundaries are modelled through
the equation
qH = hH (Hb He )

(7)

where hH is the moisture boundary coefficient, Hb is the boundary

pore relative humidity and He stands for the environmental relative
humidity. The computed moisture field H is transformed into a
field of unrestrained potential shrinkage sh,pot according to [3]

sh,pot = sh, 0.97 1.895 (H 0.2)3

(8)

where sh, is the ultimate drying shrinkage strain, that is, upon
complete drying at infinite time (valid in the scope of this paper, as
self-desiccation due to autogenous shrinkage is being neglected).
It should be highlighted that sh,pot does not necessarily represent
the real shrinkage strain at a given point; in fact, it is the shrinkage

3444

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

Fig. 1. Concrete stressstrain diagrams normal to the crack: (a) outside the
effective area; (b) inside the effective area.

strain that would potentially occur at this point if no restraint to

deformation was present. The sh,pot field is to be applied in the
mechanical model as an imposed strain, and the real shrinkage
strain can then be calculated with due account to the present
restraints.
Details about the implementation of the moisture model can be
found in [18].
2.3. Mechanical submodel
The mechanical analyses were performed using the DIANA FE
software, adopting as inputs from the thermohygro models the
equivalent ages teq and the thermal and unrestrained potential
shrinkage strains.
Basic creep of concrete was accounted for through the use
of the Double Power Law (DPL), which has a reasonably good
performance on both early age and long term time spans [20]
J (t , t ) =

1
E0 ( t )

1
E0 (t )

(t )m (t t )n

(9)

where J (t , t ) is the compliance function at time t for a load

applied at instant t , E0 (t ) is the asymptotic elastic modulus, and
1 , m and n are material parameters. Even though the DPL was
devised for basic creep estimation of bulk concrete members, it is
deemed feasible to the approach of this numerical framework, in
accordance to the strategy also adopted in Ref. [7].
Cracking was reproduced via a smeared crack approach, with
the total strain tensor being decomposed into an elastic component e and a crack strain cr that is, = e +cr [21]. A stressstrain
tension model with softening, based on the Crack Band Theory
(CBT) developed by Bazant and Oh [22], was employed to describe
the crack formation and crack growth in plain concrete. The evolution of the stress normal to the crack, n , and the corresponding
normal strain n is depicted in Fig. 1(a), and it involves the concrete tensile strength fct and the fracture energy GF . Consistently
with the CBT the specific fracture energy GF /h is adopted, where
the characteristic length h of the current FE is dependent on its
size, to ensure objective results with respect to the refinement of
FE mesh adopted in the discretization.
The stiffness increase due to bond between concrete and
reinforcement (tension-stiffening effect) was accounted for by
modifying the n n relation for the effective area of concrete
surrounding the reinforcing bars, Ac ,ef , according to MC90 [2]. In
Fig. 1(b) concrete softening inside Ac ,ef is reproduced by a multilinear diagram defined by GF and the average crack spacing ls , as
recommended by Feenstra and de Borst [23]; accordingly, n and
n are assumed as averaged over ls . During the stabilized cracking
phase the average tensile stress carried by the concrete between
cracks takes a constant value of t fct , as suggested in [2325].
Bearing in mind the applications intended for the present paper,
and in accordance to the recommendations of MC90 [2], t was
taken as 0.4. The horizontal branch of the n n diagram of Fig. 1(b)
ends when the average strain n reaches yc , the latter being
quantified as [23]:

yc = sy t

fct
Es s

(10)

In this equation Es is the Youngs modulus of steel, s is the

reinforcement ratio with respect to Ac ,ef , and sy is the yielding
strain of steel. The tension-stiffening effect vanishes when the
average strain reaches sy .
An elasticperfectly plastic diagram was adopted for the
reinforcing steel, with a yielding stress fsy = Es sy .
Given the fact that for the present paper the structural
behaviour is to be considered solely for service life conditions, a
linear elastic behaviour was assumed for concrete in compression.
2.4. Coupling between the submodels
The physical phenomena involved in the three fields under
study (thermal, hygro and mechanical) interact with each other.
However, in the macro-level approach of this paper several
simplifications are implicit, and the purpose of this section is to
provide further information on the degree of coupling between the
three submodels just presented.
It is important to remark that the thermal model provides
some further information than strictly the nodal FE temperatures,
namely in what concerns the state of the chemical hydration
reactions, since equivalent ages and a degree of heat development
are inherently calculated also.
Thermal and moisture phenomena occur at rather dissimilar
time scales, because the thermal diffusivity of concrete is in the
order of k/ c = 106 m2 s1 , whilst the counterpart for moisture flow is around 108 1010 m2 s1 [26]. Moisture fluxes
within concrete transfer heat energy, as well as the evaporation/
condensation of water involved in moisture transport. However,
the mentioned differences in diffusivities imply that variations in
temperature due to thermal fluxes occur almost instantly within
concrete, by comparison with the thermal variations due to moisture fluxes. This provides grounds for the simplification assumed in
this work, which assumes that the thermal field is not influenced
by the moisture one.
Moisture equilibrium in the pore structure of concrete and diffusion properties are, however, strongly dependent on temperature. But in this work the diffusion parameters were derived
for an environmental temperature of 20 C, and the applications
to be presented later concern only to moisture movements under isothermal conditions at this temperature. Therefore, the used
model does not account for this dependency of the moisture field
with regards to the thermal one. The simplification of considering constant environmental temperature conditions holds valid for
the first application envisaged in this paper, as drying shrinkage
prisms are usually maintained inside climatic chambers under constant T and H. However, for the second application this simplification has implications on the accuracy of predictions of temperature
fields at early ages, and also on the effect of variable T on moisture
diffusion. Nonetheless, it was decided to do so to allow focusing
on non-thermal related issues, without having to cancel effects of
temperature.
Moisture state of concrete during early ages plays a fundamental role in cement hydration: it is known that under H < 80% hydration reactions almost stop, and that for H between 80% and 97%
they are strongly hindered [27,28]. Therefore, upon moisture depression at the concrete surfaces exposed to evaporation cement
hydration is locally retarded, a relevant issue that several authors
have put forward in numerical models [20,2931]. Yet, this situation is disregarded in the applications of the present paper, since
exposure to drying is prevented until the age of 28 days. If other
kinds of applications were envisaged, namely involving formwork
removal at earlier ages (7 days or less), or comprising high performance concrete with significant self-desiccation, the formulation
presented in this paper would have to be adapted, to account for
the degree of hydration dependency on the available water.

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

3445

Fig. 2. (a) Material parameters for heat generation; (b) E-modulus evolution.

The mechanical field is influenced by the thermal submodel not

only due to the volumetric changes induced by the temperature
variations, but also because the evolution of the mechanical
properties of concrete depend on the equivalent age field; see
Eq. (4). Also, the shrinkage strains computed in the moisture model
are important data for the mechanical model; see Eq. (8).
Conversely, the mechanical field may influence the thermal and
moisture fields in several ways:
Volumetric variations of concrete may cause it to separate from
boundaries, changing boundary coefficients for both thermal
and moisture transports. However, this interaction is usually
considered negligible and thus disregarded.
Cracking may affect transport phenomena in concrete, but
due to the usually small crack widths it is feasible to assume
that they only marginally affect heat transport. In the case
of moisture cracking, it does in fact increase the moisture
transport coefficients [3234], but this effect can be considered
small enough as to be disregarded without significantly
affecting the quality of numerical predictions.
As a result of what has been stated, the approach of onedirectional coupling of the involved phenomena is justified: the
heat and moisture fields are initially calculated, and the resulting
data is used as input to the mechanical field whose calculations are
performed separately.

For the moisture model a moist curing during 28 days is

considered in all the calculations in this paper, and the diffusion
parameters reported by Kim and Lee [19] for a similar situation
are adopted: D1 = 6.17 1010 m2 s1 , H = 0.047 and
HC = 0.8. Also based on the previous reference, the moisture
transfer boundary coefficient is null during the first 28 days (moist
curing conditions), and switches to hH = 5.81 109 m s1
thenceforward. sh in Eq. (8) is considered as 500 106 , (feasible
value of final drying shrinkage upon total drying, according to
several codes [1,2]), and for the initial moisture conditions H =
100% was considered. After the initial moist curing conditions
at 20 C, environmental conditions are constant throughout the
analyses, with T = 20 C and H = 50% after the age of 28 days.
For the mechanical analyses the following concrete properties
are adopted: fct = 2.6 MPa, GF = 0.085 N/mm, T = 105 and
a Poissons ratio = 0.2. These values correspond to a normal
strength concrete of the class C25, with a maximum aggregate
size of 20 mm, according to the Model Code 1990 [2]. Evolution
of concrete E-modulus as a function of the equivalent age at
the reference temperature of 20 C is represented in Fig. 2(b)
(E-modulus considered constant for ages above 28 days). Creep
parameters for the DPL are the ones obtained experimentally by
Atrushi [36]: 1 = 1.88, m = 0.2 and n = 0.125. Reinforcement
is characterized by a Youngs modulus Es = 200 GPa, a yield stress
fsy = 500 MPa and a coefficient of thermal expansion equal to the
one of concrete.

3. Applications
3.2. Drying shrinkage prism
3.1. Overview/material properties
The applications to be presented regard firstly to an unrestrained drying shrinkage prism, and then to a slab under the combined effects of self-induced stresses (heat of hydration and drying
shrinkage) and externally applied uniformly distributed loads.
Material properties in all the examples concern solely to normal
strength concrete, which is still the most widely used in current
construction.
In terms of heat of hydration generation, the thermal output of
a dosage of 250 kg m3 of cement type I 52.5R is represented in
Fig. 2(a), which also indicates the applicable material parameters
A, Ea and Qfinal . Other assumed relevant thermal properties of
concrete are k = 2.6 W m1 K1 and c = 2400 kJ m3 K1 .
The boundary transfer coefficient corresponding to a direct contact
between concrete and the environment is expressed by the
standard value h = 10 W m2 K1 , and the initial temperature of
concrete is 20 C. The values used for thermal properties k, c and
h were adopted in accordance to the conditions of a previous case
study of a wind tower foundation [35], where ordinary concrete
was used in outdoor conditions, with an average wind speed under
5 km/h. The effects of solar radiation and night cooling have been
neglected in this study.

3.2.1. Geometry and numerical model

The structural performance of a drying shrinkage prism with
dimensions 200 200 600 mm3 , usual in material research, is to
be analysed in this section. Disregarding influence of self-weight
Fig. 3(a) reproduces the prism and their symmetry planes, which
allows just one-fourth of the prism to be reproduced numerically
according to Fig. 3(b) (surfaces labelled as sym regard to
symmetry planes, and the ones marked as bou correspond to
boundary surfaces in contact with the environment). The FE mesh
for the drying shrinkage prism is represented in Fig. 3(c), where
it can be realized that a finer refinement is considered in the
boundaries neighbourhoods, to cope with the higher thermal and
moisture gradients expected at these locations. The symmetry
planes perform as isolated in the normal directions, both for the
thermal and moisture submodels; for the mechanical analyses,
displacements normal to the symmetry planes are prevented. All
calculations (thermal, moisture and mechanical) are conducted
with the same 8-nodded FE brick (with 2 2 2 Gauss points),
although for the thermal and moisture analyses 4-nodded surface
FE (with 2 2 Gauss points) are also considered to account for
the fluxes normal to the boundaries. Time steps are selected in the
sequence that follows: 40 steps of 1 h, 20 steps of 4 h, 20 steps

3446

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

Fig. 3. Prism: (a) symmetry planes; (b) schematic representation; (c) FE mesh.

Fig. 4. Results for a selected group of points on the prism.

Fig. 5. Evolution of H on the prism.

of 16 h, 30 steps of 24 h and 230 steps of 96 h. This results in a

total time of analysis that extends for 968 days (23 240 h), which
was found to be enough for the prisms to attain almost steady state
conditions.
3.2.2. Results of the thermal and moisture submodels
Results pertaining to the temperature evolution during the
cement hydration phase, and for points P1P5 of the prism are
depicted in Fig. 4(a). The highest temperature rise occurs at P1,
the centre of the prism, and has a small value (less than 6 C).
Conversely, the lowest temperature rise occurs at the corner of the
specimen, at P5, amounting to less than 3 C, with the temperature
gradients in the specimen remaining always below 4 C. All

temperature gradients vanish around the age of 3 days, after which

the temperature within the prism becomes constant (T = 20 C).
Evolution of the relative humidity H within the specimen occurs in
a much slower fashion, as plotted in Fig. 4(b) for points P1P5 and
reproduced via the contours of Fig. 5. It is clear that the surface
points P3 and P5 dry much faster than the inner points right after
the exposure at the age of 28 days. P1, the geometrical centre, is
the slowest to converge to the environmental relative humidity
of 50%, reaching H = 51.8% at the end of the analysis. These
gradients of H are responsible for unrestrained potential shrinkage
strains calculated according to Eq. (8), which are bound to induce
self-equilibrated stresses as a consequence of the necessary strain
compatibility between the inner and outer parts of the prism.

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

3447

Fig. 6. Left column: vectors of principal stresses (red colour corresponding to tensile stresses); right column: stresses y (Pa). (For interpretation of the references to colour
in this figure legend, the reader is referred to the web version of this article.)

3.2.3. Results of the mechanical submodel

The computed stress field led to the principal stress vectors
depicted on the left column of Fig. 6 (red colour representing
maximum tensile stresses), and the contours of the normal stresses
y (longitudinal to the specimen) shown on the right. During
the first 28 days only thermally-induced stresses occur, as the
specimen is under moist curing, and so, along this phase the
thermal field evolves as follows: (i) the prism is cast at 20 C; (ii)
then the temperature rises non-uniformly (higher temperatures
in the core and lower in the surface); (iii) after the maximum
temperature is reached the prism starts cooling (faster on the
surface and slower in the core). As a consequence of cooling from
higher temperatures, the core tends to contract more than the

surface of the prism, so the latter becomes compressed and the

interior endures tensile stresses. The stress distribution at the
end of this hydration-induced thermal process can be seen in
Fig. 6 for the age of t = 27.9 days: since the thermal gradients
were less than 4 C, the resulting tensile stresses at the core are
negligible, amounting to 0.05 MPa, which allows to assume that
at the age of 28 days the prism starts to dry from an almost
stress-free state.
The results computed for the prism right after exposure to
drying are shown in Fig. 6 for t = 28.1 days. External surfaces
experience a quite sudden desiccation (in accordance to Fig. 4(b)),
which induces strong potential shrinkage strains at these locations,
whilst the core remains with high relative humidity, thus having
much lower tendency to shrink. As a result of this, the external

3448

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

Fig. 7. Stresses y in points P1P5.

Fig. 9. Computed strains at the centre of the prism vs. predictions using EC2.

Fig. 8. Evolution of cracking strains (vectors normal to the cracks, with length
proportional to the crack opening).

surfaces endure tensile stresses, self-equilibrated by compressive

stresses in the interior of the prism. Given the fact that the
desiccated depth at the beginning of drying is very small, the
relative stiffness of this desiccated skin is much lower than
the internal non-shrinking concrete core. Therefore, high tensile
stresses develop at the skin particularly in the vicinity of the
edges, where drying is stronger , reaching values as high as
1.5 MPa at t = 28.1 days. This tendency is intensified until
the tensile strength of concrete fct = 2.6 MPa is reached, and
cracking starts at the edges of the prism at t = 30 days, as it
can be confirmed in Fig. 7 (where evolutions of the y stresses are
reproduced) and in Fig. 8 (where vectors proportional to the crack
strains are shown). The very low maximum crack strain predicted
at this age, of circa 1 104 , indicates that the cracks are virtually
non-visible, which is consistent with the findings reported in
Refs. [7,11,34,37]. This process keeps intensifying until the age of
approximately 42 days, when crack openings are maximum (Fig. 8)
and tensile stresses are generalized on the skin (Fig. 6).
For ages greater than t = 42 days, and due to the progression
of drying towards the core of the prism, a reversal of the
concrete stress signs occurs, and the maximum y tensile stress
progressively migrate to the interior of the specimen (see Fig. 6
from t = 42 days until t = 964 days). This tendency can be
confirmed also in Fig. 7, where the surface points P2 and P3, having
high tensile stresses at t = 42 days, experience a reduction of those
stresses along time, until they actually reverse at t 400 days,
stabilizing around 0.5 MPa (compression) at t = 964 days. The
opposite occurs in the core points P1 and P4, which reach the end of
the analysis with tensile normal stresses of about 0.5 MPa. Located
in the corner, and thus with a very low restraint to deformation,

point P5 endures almost negligible stresses during the whole

period of analysis. The mentioned process of stress reversal in the
specimen is also visible in the crack openings reproduced in Fig. 8,
which decrease since approximately t = 42 days, and even close
after t = 324 days. This tendency of surface shrinkage cracks
to close along drying had been already documented for similar
conditions by Granger et al. [11].
Feasibility of the results obtained for the analysed prism will
now be checked by comparing the numerically predicted evolution
of the shrinkage strains with the one that would be expectable
upon application of Eurocode 2 (EC2) formulae. For such a goal,
the final shrinkage in the EC2 formulae was adjusted to coincide
with the sh,pot that arises from considering sh, = 500 106
and H = 50% in Eq. (8), that is sh,pot = 459.4 106 . This
comparison does not aim to provide an exact match, since EC2 just
provides, in a simplified way, reference values for the evolution of
strains. The predictable shrinkage strain evolution obtained from
application of EC2, together with the one computed at the core of
the specimen by using the numerical thermohygromechanical
model here proposed, are depicted in Fig. 9. The resemblance of
both evolution curves is notorious.
The analyses in this section allowed checking the feasibility of
the proposed model for calculating shrinkage-induced concrete
strains, in view of existing regulation approaches (EC2).
The EC2 shrinkage model assumes uniform strain distribution
over the cross section. If this assumption was considered, no
stresses would occur in the prism. Therefore, shrinkage prisms
are frequently used without a clear perception of the actual stress
and strain distribution over the specimen. The developed model
allowed us to quantify the stress field in this kind of specimens,
which, as demonstrated, experiences relevant variations along
time that cannot be considered negligible as often assumed. Even
though the shrinkage induced self-equilibrated stresses cannot be
estimated by simple hand calculations (as would be desirable from
the practitioner point of view), the exposed results are expected to
provide a deeper perception of the actual behaviour of shrinkage
prisms.
3.3. RC slab
3.3.1. Overview
The second example to be analysed is a 0.25 m thick RC slab,
with the purpose of checking how the self-induced stresses (due to
heat of hydration and drying shrinkage) will affect its mechanical
response to externally applied axial forces or bending moments.
Particularly, it will be checked how those self-induced stresses
influence the cracking loads and the stiffness of the slab during the
crack propagation phase.
The evolutions of both the temperature and the moisture fields
are presented first, as they are the sources of the self-induced

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

3449

Fig. 10. FE mesh for the mechanical simulation.

stresses. For the thermohygro analyses, interactions with the

environment are supposed to occur only through the bottom and
top surfaces, with 1D heat and moisture fluxes perpendicular to the
slabs middle plane.
Then, taking into account the results of the thermohygro
analyses, the mechanical response of the slab is evaluated and
discussed. Three different situations are considered:
At first the slab is subjected to the self-induced stresses only
(neither external loads nor external restraints exist).
In the second situation the combined effects of self-induced
stresses and axial forces are discussed. The slab is subjected
to an imposed axial deformation, after a time interval during
which it was left to shrink freely.
Finally, the combined effects of self-induced stresses and
bending moments are evaluated by prescribing rotations to the
slab ends, after a time interval during which it was left to shrink
freely.
The cracking loads of the previous situations are compared to
the ones obtained when shrinkage is accounted for by assuming
a uniform distribution along the slab thickness. Relevance of this
comparison lies in the fact that a uniform shrinkage distribution is
traditionally assumed in design, thus the potential inaccuracies of
this simplification are worth being evaluated.
3.3.2. Geometry and numerical model
Fig. 10 reproduces the geometry of the element to be analysed,
with a thickness of 0.25 m, reinforced on the top and bottom faces
with an amount of steel equal to 10.83 cm2 /m (12 mm rebars, with
a 29 mm cover), with Es = 200 GPa and fsy = 500 MPa. A 1 m
wide slab strip is considered, as it is assumed that the imposed
deformations are applied in the longitudinal direction only, i.e. no
restraint is considered in the direction perpendicular to the plane
shown in Fig. 10. The structure under analysis is 1.79 m long, which
corresponds to 10 times the average crack spacing for the stabilized
cracking phase, sr ,m = 0.179 m, quantified according to the MC90
definition sr ,m = 0.67 s /(3.6 s,ef ) (s = 12 mm is the
rebar diameter; s,ef is the reinforcement ratio reported to the
effective area of concrete, that is, the area of concrete effectively
bonded to steel, where tension-stiffening phenomenon develops).
The assumed slab length was chosen just for convenience, as in fact
it plays no role in the problems under study. The leftmost extremity
of the slab is considered fixed as depicted in Fig. 10.
Concrete is discretized by using 4-noded plane stress FE (in
14 layers across the thickness, as reproduced in Fig. 10), whereas
for the reinforcement, 2-noded truss FE are adopted. The effective
area of concrete that surrounds each reinforcement layer is
decomposed into five 1.75 cm tall FE, whereas the remaining core
of the slab (7.5 cm thick) is discretized with 4 FE (see Fig. 10).
The length of the concrete FEs is 17.9 cm, which coincides with
the average crack spacing sr ,m . Furthermore, linear constraints
are imposed to all vertically aligned nodes, so as to assure that
plane sections before deformation remain plane after deformation
(Bernoullis hypothesis). This prevents occurrence of sectional

distortions in the element, as the imposed strains associated to the

temperature and shrinkage fields and the concrete maturity vary
nonlinearly across the thickness of the slab. The adopted FE length
and the imposition of the Bernoullis condition are consistent with
the adopted models for simulation of the tension-stiffening effect,
as the relationships considered for the concrete are assumed
to correspond to an average behaviour between cracks. In this
way the numerical model replicates the intended integral of the
longitudinal concrete strains over the average distance between
successive cracks.
If a constant tensile strength was assumed for the concrete in
all the vertical alignments of FE (see Fig. 10), then cracking would
occur simultaneously over all the slab length when subjected to an
imposed deformation. This result would not be realistic because
experimental results of tensile tests of RC ties with imposed
deformation reveal that cracks do not arise simultaneously and the
last crack occurs for a load approximately 30% higher than that
of the first crack [2,25,38]. Consistency with experimental results
was kept by assuming that fct grows linearly from 2.6 MPa at the
leftmost vertical FE alignment up to 1.3 2.6 = 3.38 MPa at
the rightmost one, but keeping the same value of fct in all the 14
concrete FE of each vertical alignment.
The thermohygro calculations were carried out by using the
same 2D mesh (4-nodded elements) used for the mechanical
analysis, which is enough to simulate the actual heat/moisture
flows which behave as 1D.
All analyses were conducted since the instant of casting until
the age of 48 years.
3.3.3. Results of the thermal and moisture submodels
The small thickness of the slab, as well as its relatively low
cement content, contribute to a small temperature rise in concrete
due to heat of hydration. In fact, upon observation of Fig. 11(a),
where the temperature evolutions at the surface and at the core
are plotted, one can conclude that both the maximum temperature
(32 C) and the maximum temperature gradient (2.3 C) occur
at the same time (10 h). Bearing these results in mind, and
assuming the absence of external restraints to deformation, it is
highly unlikely to expect significant stresses to be associated with
the heat of hydration release.
With the moisture field computed according to Eq. (5) the plots
depicted in Fig. 11(b) were obtained for two points: at the surface
and in the core. After the 28 days period, until which all points
of the slab remained at H = 100%, exposure to an environment
with H = 50% occurred, leading to a sudden drop in the relative
humidity at the horizontal surfaces of the slab. The decrease in H
progresses along time towards the core, causing it to be almost
in equilibrium with the environmental H at the age of 3000 days
(8 years). The gradient of H along the slab thickness increases from
0% up to a maximum of 20% at the age of 88 days, and remains
relatively constant until about the age of 200 days, time after which
it decreases to 1% at 3000 days.
The potential free shrinkage (sh,pot ) evolutions along the slab
thickness that result from the computed H, which will be inputs

3450

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

Fig. 11. Results at the surface and the interior (core): (a) temperatures; (b) relative humidity.

Regarding the relevance of the longitudinal stresses induced

by drying shrinkage, the situation is substantially different, as
observable in Fig. 13(b) and in Fig. 14 between the age of
exposure (28 days) and the time at which H gradients reach
their maximum within the cross-section (88 days): the surface,
which was slightly compressed due to the initial temperature
development, goes into tension due to the rapid increase of the
surface potential shrinkage after exposure, and the core becomes
compressed. At the age of 88 days the surface tensile stresses are
large enough to cause cracking, approaching the tensile strength
of concrete fct = 2.6 MPa (Fig. 14). From then on the relative
humidity gradient within the cross-section decreases (due to the
progressive reduction of H in the interior), and thus the surface
tensile stress alleviates, whilst the core endures a contraction that
is partially restrained by the reinforcement and the outer concrete
layers, which changes its stress state from compression to tension.
At the age of 2000 days (5.5 years) the concrete normal stress
in the surface is almost 0 MPa and the tensile stress in the core
reaches 1 MPa (Fig. 13(b)).
It is instructive to compare the detailed numerical analysis just
described with the one disregarding the effect of hydration heat
release, and adopting the usual design approach of prescribing
shrinkage as a uniform strain field. The prescribed uniform
shrinkage is equal to the average strain obtained with the
differential shrinkage model. The results of the two analyses are
plotted in Fig. 14, which put into evidence the rather different
trends and values of the obtained stresses, particularly evident
at the age of maximum H gradient (88 days), with the standard
procedure being clearly non-conservative as far as the concrete
cracking risk at the surface is concerned. Even for a late age as
48 years noticeable differences are observable between the two
approaches for the self-induced stresses, although in such scenario
the standard procedure stands on the safe side with regards to
surface cracking of concrete.
Self-induced stresses combined with an imposed axial deformation.
The present section analyses the consequences of imposing an
increasing axial deformation m after the slab had been allowed
to develop self-induced stresses, the latter acting alone up to some
extent. Influence of these thermohygro stresses is checked by imposing m (as an imposed deformation at the rightmost extremity of the slab) at three different ages (scenarios): (i) t = 28 days

Fig. 12. Potential shrinkage along the slabs thickness.

to the mechanical model, are shown in Fig. 12: the well-known

tendency of concrete to progressively shrink from the surface to
the core is fully captured by the numerical model.
3.3.4. Results of the mechanical submodel
Self-induced stresses acting alone.
The developed stresses on the slab strip due to the hydrationinduced temperatures and the potential shrinkage are reproduced
in Fig. 13 until the age of 4000 days. As the rising and decreasing of temperature occur mainly within the first 4 days, their
contribution to the stress development may be commented
separately from the one due to the drying shrinkage, which is
only activated at the age of 28 days. As during the initial phase
of cement hydration the temperature in the core is higher than in
the surface, the former tends to expand more and thus it becomes
compressed, whilst the surface endures tensile stresses (Fig. 13(a)).
This tendency occurs until the age of 10 h, but is inverted from
then on since the core will cool more than the surface; this causes
a sign reversal of stresses, with compressive stresses developing
in the surface and tensile stresses in the core. Due to the very
low gradients of temperature within the cross-section, and to the
inexistence of external restraints, the developed thermal stresses
are quite low (less than 0.1 MPa).

Fig. 13. Concrete longitudinal normal stresses: (a) until 4 days; (b) until 4000 days.

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

3451

Fig. 14. Concrete normal stresses, considering both differential (continuous lines)
and uniform (dashed lines) drying shrinkage.
Fig. 16. Nm curves for imposed axial deformations at different ages, considering
both differential (continuous lines) and uniform (dashed lines) drying shrinkage.

Fig. 15. Nm curves for imposed axial deformations at three different ages.

(but just before the beginning of drying), (ii) t = 88 days (instant

of maximum H gradient) and (iii) t = 48 years (infinite time). The
corresponding relationships between m and the axial forces N installed on the slab may be observed in Fig. 15. For guidance dashed
lines are also included in this figure, defining the uncracked stiffness of the slab (state 1), or totally disregarding the contribution
of concrete (state 2). Configurations of the obtained graphics are
coherent with the typical ones reported for experimental works
running under similar conditions [38,39].
Fig. 15 shows that the cracking load Ncr , that is, the axial force at
the beginning of the first descending branch of the Nm diagram,
is markedly influenced by the previously installed self-induced
stresses. For scenario (i) the cracking load computed with the
numerical model is practically coincident with the one obtained
through Ncr = fct Ac (1 + ) = 688 kN, being the modular
ratio and the reinforcement ratio. However, for scenario (ii),
t = 88 days, the effect of self-induced stresses causes Ncr to reduce
to 560 kN, and the first branch of Nm becomes non-linear (see the
enlarged graph in Fig. 16), due to the process of crack formation
that follows the internal normal stress distribution within the
cross-section. Thus, besides causing a decrease of the cracking
load, self-induced stresses lead to a decrease of the stiffness of the
ascending branch of the Nm relationship. If the axial deformation
is prescribed at infinite time (scenario (iii)) Ncr becomes even
lower than when imposed at the age of 88 days. Naturally, the
ultimate tensile strength is the same for the three scenarios, as it
only depends on the yielding strength of reinforcement, and not on
the previous service live history.
Still regarding to Fig. 15, it is worth to remark that in the phase
of stabilized cracking the Nm diagram gradually approaches
the dashed straight line corresponding to state 2 as the axial
deformation m is imposed at later ages, which indicates an

Fig. 17. Mm curves for imposed curvatures at three different ages.

apparent progressive decay of tension-stiffening contribution

of concrete between cracks. This tendency, also reported by
Bischoff [40], is due to the shrinkage that occurs prior to the load
application.
If uniform shrinkage strains are assumed over the cross-section
of the slab, instead of the more realistic differential shrinkage
distribution predicted by the numerical model proposed here, then
higher cracking loads are obtained (see Fig. 16) regardless the age
of loading (for scenario (i) results practically coincide, since drying
shrinkage is not yet active). The lower Ncr loads for the cases where
self-induced differential stresses are considered have to be related
to the fact that cracking occurs sooner at the surface FE, due to the
higher tensile stresses induced by non-uniform drying shrinkage.
Self-induced stresses combined with an imposed curvature.
The present section discusses the effects of superimposing
increasing curvatures m to the self-induced stresses installed
on the slab. Fig. 17 reproduces the evolution of the bending
moment M and the curvature m , the latter being prescribed via
symmetrical rotations imposed to both extremities of the slab
(constant curvature along the span). Similarly to what was done in
the previous section, curvatures will be imposed to the structure
that had been left solely under the effect of self-induced stresses
until three different ages: 28 days (scenario (iv)), 88 days (scenario
(v)) and 48 years (scenario (vi)).
Disregarding influence of shrinkage, it is easy to compute the
elastic cracking moment, which for the present case is equal to
Mcr = 29.5 kN m. Nevertheless, from Fig. 17 it is possible to
notice that when the increasing curvature is imposed before onset

3452

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

along the slabs thickness, instead of the differential ones reproduced by the continuous lines in Fig. 14: except for the case of
loading at the age of 28 days, different Mm diagrams are
obtained during the crack propagation phase, which is relevant for
the service life performance of slabs. For the case of loading applied
at 48 years the apparent cracking bending moment is smaller
when computed with the simplified shrinkage model than with the
differential one, in opposition to what was observed at the age of 88
days. This is due to the fact that in the former case the differential
shrinkage model predicts occurrence of compressive stresses in the
surface areas, which retards crack formation.
4. Conclusions

Fig. 18. Mm curves for imposed curvatures at different ages, considering both
differential (continuous lines) and uniform (dashed lines) drying shrinkage.

of concrete drying, at the age of 28 days (scenario (iv)), the bending

moment that corresponds to the peak of the first ascending branch
of the Mm curve exceeds considerably that value, reaching
43.7 kN m. It should be remarked, however, that at this stage linearelasticity is no longer valid, as evident on the detail of Fig. 18. The
fact that for slender slabs this first peak on the Mm diagram
exceeds the elastic Mcr value is well known in the concrete field.
Guinea [41], who tested concrete beams with different heights,
stored under adequate conditions to prevent drying shrinkage,
showed that the flexural tensile strength is significantly higher
than the axial or splitting tensile strengths. He demonstrated
also that non-linear calculations taking into account the tensionstiffening effect as done in the present applications are able to
explain the differences between the axial and the flexural tensile
strengths.
Imposing the curvatures after onset of drying shrinkage,
markedly different Mm diagrams are observed in Figs. 17 and
18. For instance, for scenario (v) (age of loading equal to 88 days),
the ascending branch of the Mm curve exhibits a progressively
lower stiffness than that corresponding to scenario (iv) (absence
of shrinkage-induced stresses). Moreover, the Mm diagram does
not show descending branches during the crack propagation phase,
which is a consequence of the large self-induced tensile stresses
(and eventually cracking) installed prior to the external loading.
When these stresses are aggravated by the external bending
moments, cracking develops along the entire length of the tensile
surface of the slab (in this case, all the surface FE on the outermost
tensile layer crack when M reaches 8 kN m). As M increases
further cracking propagates inwards along the thickness of the
slab, and the width of the existing cracks progressively increases,
but cracking cannot spread along the span direction, which justifies
the inexistence of descending branches in the Mm diagram.
For the case in which an increasing curvature is imposed at
the age of 48 years (scenario (vi)), Fig. 18 shows that the peak
of the first ascending branch of the Mm diagram occurs for a
lower bending moment than the one for scenario (iv), that is, in the
absence of self-induced stresses. Moreover, the teeth-shaped part
of the Mm diagram is more irregular for scenario (vi), associated
to the successive formation of new cracks, as the loading on the
slab commences with a self-induced stress profile (Fig. 14) with
compression near the surfaces and a moderate tensile stress state
in the core. Therefore, the crack formation along the span and
across the thickness occurs in a more abrupt way than for scenarios
(iv) and (v).
Fig. 18 allows also to analyse the outcomes of the simplified
approach in which uniform shrinkage distributions are assumed

A numerical framework for predicting the service life performance of RC structures was presented, which accounts for the
self-induced effects due to the cement hydration heat release,
as well as to the development of non-uniform drying shrinkage
strains. The field of concrete temperatures induced by cement
hydration is solved with a thermal submodel, and the drying
shrinkage strains are computed on the basis of the relative humidity predicted in the concrete volume by a hygro submodel. Both
submodels are solved as uncoupled, feeding thereafter a mechanical submodel where concrete stresses are computed with due regard to the thermohygro strains, the equilibrium with external
loads, the evolving material properties of concrete (based on the
equivalent age concept) and creep.
Two applications were presented to illustrate the plausibility of
the numerical framework predictions, together with its potential
for application in real structures: an unrestrained drying shrinkage
prism and a RC slab subject to the combined effect of self-induced
stresses and externally applied loads.
Results obtained for the shrinkage prism showed non-uniform
residual stresses associated to heat of hydration and drying
shrinkage, with significant variations in their intensities, including
sign reversals. Evolution of the computed strains revealed quite
coherent with the shape of predictive curves of Eurocode 2, thus
confirming the plausibility of results.
The analysis of the RC slab, with due consideration of softening
and cracking of plain concrete, as well as tension-stiffening of RC,
allowed interesting conclusions to be drawn in regard to the effect
of self-induced stresses on the loaddeformation curves when load
was applied at different ages. It was observed that for loads applied
at instants when important self-induced surface tensile stresses
were already installed in concrete, the structural response of the
slab was significantly different from that corresponding to a zero
stress initial stage, or even from that assuming the simplification
of uniform distribution of shrinkage strains. It was shown that
differential shrinkage stresses lead to a decrease of the calculated
axial cracking loads, regardless the age of loading. Therefore,
the standard procedure suggested by design codes such as the
Eurocode 2 (based on uniform shrinkage strains along the cross
section) underestimates the concrete cracking risk. Codes could be
improved by suggesting reduced cracking loads for design checks
in which lower cracking loads are unfavourable. Moreover, nonuniform shrinkage stresses also cause a decrease of the stiffness
of the ascending branch of the N m relationship, which is more
pronounced for loads applied during the initial drying phase, when
important superficial tensile stresses occur due to differential
shrinkage.
The numerical results also showed that self-induced stresses
are responsible for important stiffness reductions in slabs subjected to increasing bending moments. Furthermore, the flexural tensile strength of slabs is significantly affected by differential
shrinkage stresses. For that reason, the calculation of cracking loads
based on the flexural tensile strength (an hypothesis admitted by

M. Azenha et al. / Engineering Structures 33 (2011) 34423453

design codes such as the Eurocode 2) may frequently lead to overestimations.

Overall plausibility of the obtained results provides good expectancies in regards to the applicability of numerical frameworks
like the one proposed here. Upon further validation, this kind of
numerical tools has a great potential for analysing the service-life
performance of real RC structures. An important field of application concerns to the evaluation of reinforcement needs in order to
control shrinkage-induced cracking.
Acknowledgements
Funding provided by the Portuguese Foundation for Science and
Technology to the Research Units ISISE and LABEST, as well as to the
first and second authors through the Ph.D. grants SFRH/BD/13137/
2003 and SFRH/BD/29125/2006, and to the research projects
PTDC/ECM/68430/2006 and PTDC/ECM/099250/2008, are gratefully acknowledged.
References
[1] CEN. EN 1992-1 European standard Eurocode 2: design of concrete
structurespart 1: general rules and rules for buildings. 2004.
[2] CEB-FIP. Model code 901993.
[3] Kwak H-G, Ha S-J, Kim J-K. Non-structural cracking in RC walls: part I. Finite
element formulation. Cem Concr Res 2006;36:74960.
[4] Kwak H-G, Ha S-J. Non-structural cracking in RC walls: part II. Quantitative
prediction model. Cem Concr Res 2006;36:76175.
[5] Bernard O, Bruhwiler E. The effect of reinforcement in the new layer on hygral
cracking in hybrid structural elements. Mater Struct, Mater Constr 2003;36:
11826.
[6] Martinola G, Sadouki H, Wittman FH. Numerical model for minimizing risk of
damage in repair system. J Mater Civ Eng 2001;13:1219.
[7] Grasley ZC, Lange DA, DAmbrosia MD. Internal relative humidity and drying
stress gradients in concrete. Mater Struct, Mater Constr 2006;39:9019.
[8] Oh BH, Cha SW. Nonlinear analysis of temperature and moisture distributions
in early-age concrete structures based on degree of hydration. ACI Mater J
2003;100:36170.
[9] Oh BH, Choi S. Advanced analysis of stresses for controls of transverse
cracking in early age concrete decks of composite bridges. In: KonstaGdoutos MS, editor. Measuring, monitoring and modeling concrete properties.
Alexandroupolis (Greece): Springer; 2006.
[10] Chen D, Mahadevan S. Cracking analysis of plain concrete under coupled heat
transfer and moisture transport processes. J Struct Eng 2007;133:40010.
[11] Granger L, Torrenti J-M, Acker P. Thoughts about drying shrinkage: experimental results and quantification of structural drying creep. Mater Struct, Mater
Constr 1996;30:58898.
[12] Gawin D, Pesavento F, Schrefler BA. Modelling creep and shrinkage of concrete
by means of effective stresses. Mater Struct, Mater Constr 2007;40:57991.
[13] TNO-DIANA-BV. Diana users manualrelease 9.2. Delft (The Netherlands);
2007.
[14] Reinhardt H, Blaauwendraad J, Jongedijk J. Temperature development in
concrete structures taking account of state dependent properties. In: int. conf.
concrete at early ages. 1982.
[15] Faria R, Azenha M, Figueiras JA. Modelling of concrete at early ages: application
to an externally restrained slab. Cem Concr Compos 2006;28:57285.

3453

[16] Cusson D, Hoogeveen T. Measuring early-age coefficient of thermal expansion

in high-performancee concrete. In: International RILEM conference on volume
changes in hardening concrete: testing and mitigation. 2006.
[17] Viviani M, Glisic B, Smith IFC. Separation of thermal and autogenous
deformation at varying temperatures using optical fiber sensors. Cem Concr
Compos 2007;29:43547.
[18] Azenha M. Numerical simulation of the structural behaviour of concrete since
its early ages. Ph.D. thesis. Faculty of Engineering of the University of Porto.
2009.
[19] Kim JK, Lee CS. Moisture diffusion of concrete considering self-desiccation at
early ages. Cem Concr Res 1999;29:19217.
[20] Bazant ZP. Mathematical modeling of creep and shrinkage of concrete.
Chichester: John Wiley & Sons, Inc.; 1988.
[21] De Borst R. Smeared cracking, plasticity, creep, and thermal loadinga unified
approach. Comput Methods Appl Mech Engrg 1987;62:89110.
[22] Bazant ZP, Oh BH. Crack band theory for fracture of concrete. Mater Struct
1983;16:15577.
[23] Feenstra P, De Borst R. Aspects of robust computational modelling for plain
and reinforced concrete. Heron 1993;38:576.
[24] Winkler B, Hofstetter G, Niederwanger G. Experimental verification of a
constitutive model for concrete cracking. Proc Inst Mech Eng Part L J Mater
Des Appl 2001;215:7586.
[25] Walraven J. Tension stiffening. In: Structural concrete, the textbook on
behaviour, design and performance updated knowledge of the CEB/FIP model
code 1990 volume 1. 2nd ed. fib bulletin, vol. 51. FIB; 2009.
[26] Jonasson J-E. Modelling of temperature, moisture and stresses in young
concrete. Doctoral thesis. Lulea: Lulea University of Technology; 1994.
[27] Azenha M, Faria R, Maekawa K. Influence of moisture losses at early ages
on concrete cover quality: a numerical approach. In: International RILEM
workshop on integral service life modelling of concrete structures. RILEM
Publications; 2007.
[28] Patel R, Killoh D, Parrott L, Gutteridge W. Influence of curing at different
relative humidities upon compound reactions and porosity in Portland cement
paste. Mater Struct 1988;21:1927.
[29] Granger L. Comportement diffr du bton dans les enceintes de centrales
nuclaires. Analyse et modlisation. Ph.D. thesis. Paris: LCPC; 1996.
[30] Mjrnell K. Moisture conditions in high performancee concrete. Gteborg:
Chalmers University of Technology; 1997.
[31] Maekawa K, Chaube R, Kishi T. Modelling of concrete performance. E&FN
SPON; 1999.
[32] Suwito A, Ababneh A, Xi Y, Willam K. The coupling effect of drying shrinkage
and moisture diffusion in concrete. Comput Concr 2006;3:10322.
[33] Sadouki H, Van Mier JGM. Simulation of hygral crack growth in concrete repair
systems. Mater Struct, Mater Constr 1997;30:51826.
[34] Bisschop J, Van Mier JGM. Effect of aggregates on drying shrinkage
microcracking in cement-based composites. Mater Struct, Mater Constr 2002;
35:45361.
[35] Azenha M, Faria R. Temperatures and stresses due to cement hydration on the
R/C foundation of a wind tower-a case study. Eng Struct 2008;30:2392400.
[36] Atrushi D. Tensile and compressive creep of early age concrete: testing
and modelling. Ph.D. thesis. Trondheim (Norway): Norwegian University of
Science and Technology; 2003.
[37] Bazant Z, Raftshol W. Effect of cracking in drying and shrinkage specimens.
Cem Concr Res 1982;12:20926.
[38] Jaccoud J. Armature minimale pour le contrle de la fissuration des structures
en bton. EPFL. 1987.
[39] Farra B. Influence of concrete resistance and its adherence to reinforcement
on cracking. Ph.D. thesis. Switzerland: EPFL; 1995 [in French].
[40] Bischoff PH. Effects of shrinkage on tension stiffening and cracking in
reinforced concrete. Can J Civ Eng 2001;28:36374.
[41] Guinea G. Modelling the fracture of concrete: the cohesive crack. Mater Struct
1995;28:18794.