Engineering Structures 32 (2010) 1964–1975

Contents lists available at ScienceDirect

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

Extended cracked membrane model for the analysis of RC panels

Mário Pimentel a,∗ , Eugen Brüwhiler b , Joaquim Figueiras a
a
Laboratory for the Concrete Technology and Structural Behaviour (LABEST), Faculty of Engineering of the University of Porto, Portugal
b
School of Architecture, Civil and Environmental Engineering, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland

article info abstract

Article history: A new model considering fixed and interlocked cracks for the analysis of reinforced concrete (RC) cracked
Received 7 September 2009 membrane elements is presented. The model can be regarded as an extension of the cracked membrane
Received in revised form model proposed by Kaufmann and Marti. Its value lies in eliminating the need to resort to empirical
23 December 2009
averaged stress–strain relations for the reinforced concrete material, while being sufficiently simple for
Accepted 28 February 2010
Available online 31 March 2010
implementation in a robust finite element formulation. Equilibrium is formulated in terms of stresses at
the cracks, enabling the use of fundamental theoretical models for the individual mechanical phenomena
Keywords:
governing shear behaviour, such as concrete compression softening, aggregate interlock (including crack
Reinforced concrete dilatancy effects) and tensile bridging stresses. Reinforcement stresses at the cracks are obtained from the
Shear average strains assuming a stepped rigid plastic bond shear stress–slip law according to the tension chord
Panels model. This allows for a consistent treatment of the tension stiffening effects and a proper evaluation
Tension stiffening of steel deformation capacity in the post-yielding stage. A rational and quantitative description of the
Tension chord model stress/strain fields both at the cracks and in between the cracks is derived, which allows a deeper
Compression softening understanding of the complex mechanics involved in shear behaviour of cracked membrane elements.
Aggregate interlock A validation campaign was undertaken using a database with the experimental results of 54 RC panels
Shear dilatancy
subjected to in-plane shear and axial stresses. The failure modes, failure loads and stress–strain curves
are, in general, accurately predicted.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction the theoretical models had to be balanced in order to achieve a

Nonlinear finite element analysis (NLFEA) is increasingly rec-
ognized as an effective tool for the accurate assessment of the
load-carrying capacity of concrete structures [1]. By enabling the
consideration of both equilibrium and compatibility conditions,
and if appropriate constitutive laws are adopted, not only ultimate
loads but also realistic force–deformation curves can be calculated.
Within the NLFEA context, many structural concrete elements,
such as bridge I-girders and box-girders, shear walls, transfer
beams, containment structures and offshore oil platforms, to cite a
few, can be idealized as an assembly of membrane elements. There-
fore, an accurate representation of structural concrete behaviour at
the membrane element level is the key for the success of any finite
element model aimed at analysing real scale concrete structures.
In the past 30 years a strong investment has been made on
the research of suitable constitutive laws for RC membranes.
Due to the multiplicity of the involved mechanical effects, it is
easy to understand that the desired accuracy and complexity of
∗Corresponding author. Tel.: +351 225081561; fax: +351 225081854.
E-mail addresses: mjsp@fe.up.pt (M. Pimentel), eugen.bruehwiler@epfl.ch
(E. Brüwhiler), jafig@fe.up.pt (J. Figueiras).
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2010.02.030
satisfactory engineering solution to the problem. Therefore, most
of the existing constitutive models include a series of simplifying
assumptions. One of the most striking aspects of structural
concrete behaviour is that, after cracking, the slip between
concrete and reinforcing bars leads to a highly irregular stress field.
This aspect is commonly dealt with by treating reinforced concrete
as a new material with its own stress–strain characteristics,
which are valid only in spatially averaged terms. These laws
are established based on experimental evidence resulting from
extensive tests on RC panels subjected to in-plane shear and
axial stresses. Although neither the strain nor the stress fields are
uniform throughout each of these panels – local variations occur
between the cracks as will be discussed in this paper – up to
the peak load it is licit to assume an averaged (or homogenised)
state of strain. In fact, once the crack pattern is stabilized one can
assume that these irregularities are repeated and that the stress
and strain fields can be averaged, as indicated in Figs. 4 and 8,
respectively. This is confirmed by the redundant measurements
that are made in properly conducted experiments. The RC panel is
thus considered as the basic unit, or the control volume, over which
the constitutive laws are valid and no spatial discretization is
required for calculating its force–deformation response. Examples
of formulations of this type include the Rotating and Fixed Angle
Softened Truss Models [2,3], and the latest Softened Membrane
 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975 1965

Model [4]. In others, local behaviour at the cracks is acknowledged

to govern the ultimate capacity of RC membranes, and some
kind of equilibrium check in terms of stresses at the cracks is
additionally performed. Examples of these types of models are the
well-known Modified Compression Field Theory [5] and the more
recent Disturbed Stress Field Model [6].
These formulations have greatly contributed to improved
accuracy in the analyses of the shear response of RC elements.
However, they do not give a quantitative description of both local
and average stress/strain fields. Due to this fact it is not possible
to quantify the steel strains in the vicinity of the cracks in the
post-yield regime. Consequently, it is not possible to formulate
a rational criterion for evaluating the deformation capacity of
the reinforcements, which can be crucial for accurate safety
evaluations of structural concrete elements reinforced with less Fig. 1. Cracked membrane element.
ductile steel or corroded reinforcement. It is also commonly
assumed that the direction of the principal average concrete stress
coincides with that of the principal average strain. Although this
assumption greatly simplifies the resulting theories, a lag between
the two directions is observed in the experiments [6]. Moreover,
this assumption is clearly violated in the case of a sliding shear
failure in which part of a specimen is pushed up and away relative
to the other. Additionally, the relationship between the adopted
constitutive relations and the well-established mechanical models
for some specific phenomena taking place at the cracks and which
are known to govern the behaviour of RC elements, such as
the crack shear stress transfer mechanics (including crack shear Fig. 2. Equilibrium expressed in terms of stresses at the crack.
dilatancy effects), is often not transparent.
In the work by Marti and Kaufmann [7] a slightly different ap- the cracks is derived. In order to allow its future implementation
proach was developed. Still relying on a macroscopic approach and in a robust and efficient numerical procedure for finite element
on the consideration of an homogenised strain field, the Cracked analysis, an explicit formulation is proposed, i.e., given a trail
Membrane Model (CMM) combines the basic concepts of the orig- average strain state, the corresponding stress tensor is directly
inal Compression Field Theory [8,9] with the bond stress transfer calculated without having to resort to inner iterative loops in the
mechanics as described by the Tension Chord Model (TCM) [10]. constitutive laws. The issues related with the post-peak response
Equilibrium equations are formulated directly at the cracks, rather and correspondent localization of the failure modes are out of
than using average stresses, and tension stiffening effects result scope of the present work.
from the bond shear stress–slip relationship according to the TCM. The proposed formulation will be hereby designated as F-CMM,
The TCM allows the calculation of the steel stresses at the cracks where the ‘‘F-’’ stands for fixed and interlocked cracks. The CMM
from the average strains, enables a rational description of the will be considered as a particular case, and will be hereby referred
stress/strain field both at the cracks and in spatially averaged as R-CMM, where the ‘‘R-‘‘stands for rotating cracks.
terms, and allows obtaining the crack spacing from first principles.
As a major and significant simplification, stress-free rotating cracks 2. Equilibrium equations
are considered and their direction is perpendicular to the principal
tensile direction of average strains. Consider an orthogonally reinforced RC panel with a set of par-
Other models were proposed later for describing both the aver- allel and uniformly spaced cracks (Fig. 1). The panel is subjected to
age and local stress/strain fields, adopting other formulations for the external in plane stresses σxy = {σx σy τxy }T . For simplicity, the
the bond stress–slip laws and extending the concept to the case global coordinate system xy is aligned with the reinforcement di-
of fixed an interlocked cracks [11–14]. In these models, the crack rections. The sign convention is such that the stresses represented
kinematics and the individual mechanical effects that take place at in Fig. 1 are positive. The local crack coordinates n and t are aligned
the cracks are explicitly considered and allow a deeper understat- with the crack direction. The angle θr defines the direction of the
ing of the complex and interactive mechanics involved in the shear normal to the cracks.
behaviour of RC cracked membranes, at the cost of leading to for- Equilibrium of stresses at the cracks requires (see Fig. 2),
mulations that are more complex than the ones corresponding to σx = σcr ,n cos2 θr + σcr ,t sin2 θr − 2 τcr ,nt sin θr cos θr + ρx σsr ,x
the previously mentioned models.
In this work the CMM is generalized to the case of fixed and σy = σcr ,n sin2 θr + σcr ,t cos2 θr + 2 τcr ,nt sin θr cos θr + ρy σsr ,y
(1)
interlocked cracks. In the context of finite element analysis this τxy = σcr ,n sin θr cos θr − σcr ,t sin θr cos θr
extension is important as, in some cases, rotating crack models
allow an excessive reorientation of the compressive stress field to + τcr ,nt (cos2 θr − sin2 θr ).
the extent where a large portion of the load is taken to the supports The local crack stresses have a clear physical meaning in light of
by direct strut action. In reality this reorientation is delayed, or the concrete mechanics. The concrete stress component parallel to
even prevented, by the slip occurring at the cracks and can only the t coordinate, σcr ,t , is related to the diagonal compressive field.
be evaluated if proper consideration is given to the crack shear The concrete stress component parallel to the n coordinate, σcr ,n ,
stress transfer model. In fact, crack reorientations often occur in is obtained by summing the crack tensile bridging stresses and
panel tests but not so frequently in girder tests. The TCM is adopted the crack dilatancy stresses arising from the crack shear transfer
for calculating the steel stresses at the cracks from the average mechanisms, σcr ,n = σbri + σdil . The crack shear stress, τcr ,nt , is
strains accounting for the tension stiffening effects and a rational given by the sum of the shear stresses arising from the aggregate
description of the stress field both at the cracks and in between interlock and dowel action effects, τcr ,nt = τagg + τdow .
1966 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975
Fig. 3. Mohr circle of stresses expressing the equilibrium in terms of local crack
stresses.
The equilibrium can be visualized in the Mohr circle of stresses
of Fig. 3. In the right circle, the points X = (σx , τxy ) and Y =
(σy , −τxy ) represent the applied stresses in the xy-coordinate
system. The concrete stresses at the cracks are represented by the
left circle, with Xcr = (σcr ,x , τcr ,xy ), Ycr = (σcr ,y , −τcr ,xy ) and
Ncr = (σcr ,n , τcr ,nt ), Tcr = (σcr ,t , −τcr ,nt ) representing the concrete
crack stress tensor in the global and local coordinate systems,
respectively. Due to dilatancy stresses σcr ,n can be negative, as
represented in the figure. The equilibrium between the applied
stresses and the internal stresses is obtained by adding the
reinforcement stress contribution at the cracks, ρx σsr ,x and ρy σsr ,y .
As the reinforcements are aligned with the xy-coordinate system
and as they only carry axial stresses, the equality τxy = τcr ,xy is
valid and the points Xcr /X and Ycr /Y have the same ordinate.
2.1. Relationship between average and local crack stresses
For unbonded reinforcement, steel and concrete stresses are
constant within the membrane element. In the presence of bonded
Fig. 4. Stress field in the cracked membrane element. Notation.
reinforcement, bond stress transfer between reinforcement steel
and concrete is activated when a crack is formed leading to
the existence of an irregular concrete stress field, as depicted
schematically in Fig. 4. The variation of the concrete stress in
the n − t coordinate system can be observed. At the cracks, the
normal concrete stress σc ,n can assume negative values due to
crack dilatancy effects. Equilibrium requires that the compressive
stress parallel to the t-direction σc ,t has its maximum magnitude
at the cracks. The same occurs with the shear stress τc ,nt . In
the reinforcement x- and y-directions the steel stresses σs,x and
σs,y , the concrete stresses σc ,x and σc ,y , and the tensile stresses
transferred from steel to concrete ∆σc ,x and ∆σc ,y can be analysed.
Steel tensile stresses are maximum at the cracks, gradually
decreasing due to bond action. In order to maintain equilibrium,
concrete stresses vary accordingly. In general, the compressive
stresses magnitude is higher at the cracks and lower between
the cracks, the opposite occurring with the tensile stresses. It
must be remarked that the concrete stresses in the reinforcement
directions, σc ,x and σc ,y, are generally compressive (see Fig. 3).
The spatially averaged stress field can be obtained averaging the
concrete and steel stress distributions. The average tensile stresses
transferred from steel to concrete in the reinforcement directions,
i.e, the tension stiffening stresses, are represented by ∆σcm,x and
∆σcm,y . These stresses can be obtained from:
∆σcm,x = ρx σsr ,x − σsm,x


(2)
∆σcm,y = ρy σsr ,y − σsm,y


where σsr ,x/y are the steel stresses at the cracks for the x/y-
reinforcements and σsm,x/y are the corresponding average stresses.
The panel equilibrium can also be expressed in terms of average
stresses, which can still be expressed by the Eq. (1) if the stresses
at the cracks σcr ,n , σcr ,t , τcr ,nt , σsr ,x and σsr ,y are replaced by the
corresponding average stresses σcm,n , σcm,t , τcm,nt , σsm,x and σsm,y .
In the Mohr circle of stresses of Fig. 5, the relationship between
concrete average stresses and concrete stresses at the cracks is
 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975 1967

Fig. 6. Crack kinematics. Notation.

between the cracks can be neglected when compared with

the deformation corresponding to smearing the slip occurring
(c )
Fig. 5. Mohr circle of stresses showing the relationship between concrete average between the crack lips, i.e., γnt = 0.
stresses and concrete stresses at the cracks. The average total engineering strains in the crack local coordi-
nate system can be obtained from the total engineering strains in
shown. The tension stiffening stresses are responsible for the global coordinate system using the strain transformation relation-
deviation of the two circles, which can be expressed analytically ship,
as,
εn c2 s2 εx
 (
sc
( ) )
σcm,n = σcr ,n + ∆σcm,x cos θr + ∆σcm,y sin θr
2 2
εnt = Tε εxy ⇔ εt =  s2 c2 −s c  ε y ,
σcm,t = σcr ,t + ∆σcm,x sin2 θr + ∆σcm,y cos2 θr (3) γnt −2 s c 2sc c 2 − s2 γxy

τcm,nt = τcr ,nt + −∆σcm,x + ∆σcm,y sin θr cos θr . s = sin θr

In general, the principal tensile stress directions of the local and
average concrete stress tensors, θcr and θcm , respectively, do not
coincide and the components of the average concrete stress tensor
have a different meaning from their counterparts at the cracks.
Apart from the crack bridging stresses and whenever there is no
crack shear slip, σcm,n is the projection of the tension stiffening
stresses in the n-direction. However, if shear slip at the cracks
is activated, crack dilatancy stresses introduce an extra stress
component along the n-axis. When the tension stiffening stresses
are different, a shear component is also introduced. Consequently
the crack shear stress τcr ,nt and the average shear stresses τcm,nt are
not equal, which is reflected by the different ordinate of the points
Ncr and Ncm .
3. Compatibility equations
In this work, an approach involving average, or smeared, strains
is adopted. The bond shear stress–slip relationship is formulated in
such a way that the slip between reinforcements and concrete does
not need to be explicitly calculated. This concept agrees with the
goal of formulating a material model suitable for implementation
in a finite element code dedicated to the analysis of real scale
structures.
The total average engineering strains, εnt = {εn εt γnt }T , can be
divided into average strains due to deformation of the concrete
(c )
between the cracks, εnt , and smeared strains due to local crack
(r )
displacements, εnt :
εnt = ε(ntc ) + ε(ntr ) .
In the crack local coordinate system the smeared strain compo-
nents due to crack opening, δr ,n (or simply wr ), and crack slip, δr ,t ,
are given by (Fig. 6):
δ r ,n δ r ,t
εn(r ) = εt(r ) = 0 γnt(r ) =
srm srm
with srm being the average crack spacing.
In order to avoid internal iterative loops at the constitutive level
for performing the tensorial split given by Eq. (4), two simplifying
assumptions are adopted regarding the concrete strains: (1) after
cracking, the strut’s axial contraction due to transversal concrete
tensile stresses is neglected; (2) concrete shear deformation in
(6)
c = cos θr .
4. Constitutive relations
A general model for complete load–deformation analysis of RC
membrane elements subjected to in-plane shear and axial stresses
needs to consider the equilibrium conditions, the compatibility
conditions and suitable constitutive laws for: (1) concrete non-
linear behaviour under compressive stresses; (2) crack bridging
stress; (3) shear stress transfer between rough cracks; (4) bond-
stress transfer between the reinforcing bars and surrounding con-
crete. Prior to cracking linear elastic behaviour is considered.
In the F-CMM equilibrium is expressed in terms of crack
stresses and compatibility is expressed in terms of average strains.
Therefore, the adopted constitutive laws shall relate these two
stress and strain tensors.
4.1. Compressive behaviour
The compressive strength of plain concrete is reduced (or
softened, as usually termed) in the presence of tensile strains in the
orthogonal direction. The softened compressive strength is here
considered as the product of two coefficients, ςf and ςe , by the
cylinder compressive strength fc0 :
fc ,ef = ςe ςf fc0
 31
(7)

1 fc0
ςe = ≤1 ςf = ≤ 1 (fc0 in MPa)
Cm (1.08 + 81εn ) f0
c
(4)
where fc0 = 20 MPa, εn is the average tensile strain in the direction
normal to the cracks, and Cm is a constant.
The coefficient ςf was defined according to Muttoni et al. [15]
and expresses the dependency of the uniaxial compressive
strength of unconstrained concrete panels or wall elements
(5) on fc0 . The expression for the softening coefficient ςe is based
on the proposal of Kaufmann [7,16]. The original expression
was calibrated for the CMM, which is a rotating crack model,
and it is equivalent to considering Cm = 1.0. However, as
discussed by Vecchio [6], rotating crack models require more
compression softening than models which explicitly account
for the crack shear slip kinematics. This is due to the fact
that the principal compressive average strain includes the
1968 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975
a b
Fig. 7. Softened compressive stress–strain relationship for cracked concrete: (a) ratio fc ,ef /fc
curve for varying softening degrees.
contributions of both concrete deformations and smeared crack
shear slip displacements. The parameter Cm allows adjusting the
expression for ςe in order to eliminate the slippage influence
on the compression softening coefficient. As in the F-CMM
both crack shear slip and crack dilatancy effects are taken into
account by independent models, the compression softening
relationship must be calibrated with experimental data obtained
from tension–compression tests on RC panels. These tests allow
an independent calibration of the compressive softening law
without any assumption regarding the tensile behaviour in the
perpendicular direction or regarding the slip between the crack
lips. In Fig. 7(a) it can be seen that with the adoption of Cm = 0.9
an improved fit to the experimental results can be obtained.
As it was shown by Belarbi and Hsu [17], in cases of proportional
loading the strain at peak stress is also reduced. In the present
model, the same softening coefficient used for the compressive
strength was also adopted for the uniaxial peak strain ε00 :
ε0,ef = ςe ςf ε0 .0
(8)
The uniaxial peak strain can be related to the compressive
strength by ε00 = 0.0017 + 0.0010fc0 /70, with fc0 in MPa [18].
The stress–strain curve is defined on the basis of normalized
values of the compressive stress and compressive strain:
σcr ,t εt
S=− E=− . (9)
fc ,ef ε0,ef
The stress–strain relations in the ascending and post-peak
branches are given by:
k E − E2

, E≤1



1 + (k − 2) E
S= 1 (10)

 η
, E > 1.
1 + 2 1 − 2E + E 2

For the ascending branch, E ≤ 1, Sargin’s law was adopted [18],
while the descending branch was defined ensuring that S is
continuously differentiable. The parameter k in the expression for
the ascending branch controls the shape of the curve. With k = 2,
a parabolic relationship is obtained. In general, k = Eci ε0,ef /fc ,ef ,
with Eci being the tangential modulus of elasticity at σc = 0, which
can be estimated according to the fib recommendations [18].
It was shown experimentally [19,20] that even high strength
concrete exhibits ductile behaviour when crushing in the presence
of large lateral strains. In the proposed post-peak relationship this
effect is taken into account by the parameter η, which was made
dependent on the softening coefficients:
π2
4
η= ςf ςe
. (11)
4 tact Density Model [27] is adopted. The shear stress and the ac-
The compressive stress–strain curves for varying softening
coefficient values are depicted in Fig. 7(b).
02/3
versus lateral strains (experimental data from Refs. [20–22]) ; (b) stress–strain
4.2. Tensile crack bridging stresses
Up to tensile strength, concrete is assumed to behave elastically.
When the tensile strength is reached, the fracture process develops
until a macroscopic crack is formed. A lower bound estimate of
the tensile strength, already accounting for the self equilibrating
residual stresses, can be obtained from
fct = 0.33 fc0
p
[MPa]. (12)
Tensile stresses arising from crack bridging effects can still be
transmitted during the fracture process and, although negligible
for well-reinforced structures, can play an important role when
modelling poorly reinforced or plain concrete elements. In the
present model, the expression proposed in Ref. [23] is adopted for
the post-peak branch:
" 3 #
σbri wr wr

wr
−6.93 w
= 1 + 27 e c − 0.027384 ,
fct wc wc
(13)
GF
being wc = 5.14 , wr = hεn .
fct
The fracture energy GF is determined according to the CEB
recommendations [24]. The parameter h is the crack band width,
which in the case of finite element analysis depends on the finite
element size [25]. In the presented calculations, it was assumed to
be equal to an estimate of the average crack spacing.
4.3. Shear transfer through rough cracks
The existing theoretical models for shear transfer through
rough cracks are based on experimental tests in pre-cracked
notched specimens—the push-off tests. In these tests a straight
crack is formed along a predefined notch and the crack rough-
ness is only due to the protruding aggregates. This can be defined
as local crack roughness. However, due to concrete heterogene-
ity, in RC panels tested under in-plane stress conditions, as well
as in real structures, cracks present a crooked shape. This global
roughness is responsible for another interlocking mechanism in
addition to aggregate interlock [3,15]. Additionally, as reported by
Walraven [26], cracks reinforced with deformed bars exhibit an ex-
tra shear stress transfer mechanism, which is mobilized due to the
crack width reduction in the neighbourhood of the reinforcing bars.
In conclusion, for given crack opening and shear displacements,
larger shear and dilatancy stresses are to be expected across struc-
tural concrete cracks when compared to the ones obtained from
push-off tests.
In the present model, a closed form solution based on the Con-
companying normal compressive component (the crack dilatancy
stress) are given by:
 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975
a b
Fig. 8. Tension chord model: (a) Bond shear stress–slip relationship; (b) Stress–strain for the reinforcement bars; (c) Chord element for increasing load levels corresponding
to regimes 1, 2 and 3.
β2
τagg = τLIM
1 + β2
π β
 
σdil = −τLIM − cot−1 β −
2 1 + β2
τLIM = 3.83fc01/3 g (εn , srmθ ) ,
0.5
g (εn , srmθ ) = ≤1
0.31 + 200 srmθ εn
where β = δr ,t /wr . If a regular array of cracks is formed, it is pos-
sible to conclude from (5) that β = γnt /εn . In the previous ex-
pression, τLIM is the maximum attainable shear stress and srmθ is
an estimate of the maximum crack spacing:
cos θr sin θr
  −1
srmθ = + (15)
srmx0 srmy0
with srmx0 and srmy0 being the maximum uniaxial crack spacings in
the reinforcement directions, which must be determined accord-
ing to expression (16).
In the original approximate closed form solution of the Contact
01/3
Density Model a constant value τLIM = 3.83fc was adopted.
However, this value was found to overestimate the crack shear
capacity for large crack openings. In the present formulation, a
function g (εn , srmθ ) is defined which basically reduces the crack
shear capacity when the normal strains, and therefore the crack
openings, become large. The adopted formulation was developed
modifying the expression proposed by Vecchio and Collins [5]
for the crack shear capacity in order to take into account the
crack width reduction in the neighbourhood of the rebars and
the cracks meandering shape. For simplicity, the influence of the
maximum aggregate size was neglected. It must be recalled that
the parameter 200 in the denominator of g (εn , srmθ ) is closely
related to how the uniaxial crack spacings are determined. This
value was found to lead to good results for maximum uniaxial crack
spacings determined according to the TCM (expression (16)) and
calculating fct from expression (12).
The Contact Density Model was originally formulated consider-
ing the crack surface geometry of normal strength concrete. In the
case of high strength concrete (HSC), the crack surface geometry
1969
is smoother. This is attributable to the fact that the bond strength
between the mortar and coarse aggregates is so high that these
tend to break apart, rather than separate from mortar as in nor-
(14) mal strength concrete. Due to this fact, in HSC less shear dilatancy
is observed. In order to take this effect into account in a simpli-
fied manner fc0 is limited to 40 MPa in the σdil expression (14)2 .
However, also the crack shear stress law (14)1 and crack shear ca-
pacity (14)3 must be updated for HSC. Therefore, further research
must be made in order to develop a suitable closed form expres-
sion similar to (14) that can be applicable for HSC, thus improving
the generality of the model.
The small contribution of dowel action is neglected in the
present model.
4.4. Reinforcement steel and tension stiffening
The effect of bond on the behaviour of structural concrete mem-
bers is reflected by the stiffer post-cracking response exhibited by
the tension chord when compared to the obtained with a naked
steel bar of equal resistance. This effect is called tension stiffen-
ing and its inclusion in the structural analysis is essential both in
the pre- and post-yield regimes. In the F-CMM, tension stiffening
is taken into account by calculating the reinforcement stresses at
the cracks from the average strains. This is performed following the
lines of the TCM [10,16,28,29]. Another interesting perspective on
the modelling of the composite behaviour of RC and the interac-
tion between the reinforcement and the surrounding concrete can
be found in [30,31].
4.4.1. Tension Chord Model—one-dimensional stress conditions
In the TCM, a simple stepped rigid-perfectly plastic bond
stress–slip relationship is used, see Fig. 8(a). This relationship has
been extensively validated with experimental results and numer-
ical simulations with more detailed bond stress–slip laws [10,
28,32] and its applicability has been verified both in pre- and
post-yield regimes. In the case of ordinary ribbed rebars, it was
02/3
suggested by Sigrist [29] to assume τb0 = 0.6fc (prior to rein-
forcement yielding) and τb1 = 0.5τb0 (after reinforcement yiel-
ding). In the following, a bilinear stress–strain relationship is
considered for the reinforcement, as depicted in Fig. 8(b).
1970 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975

schematically depicted for the pre-yielding stage. Immediately

Fig. 9. Steel stresses for a monotonically increasing load according to the TCM.
after cracking, the steel stress at the crack jumps from n · f ct (where
n is the modular ratio Es /Ec ) to σsr0 . Points C and E correspond
to the end of the crack formation phase for λ = 1 or λ < 1,
respectively. The implementation of the TCM in the present model
is such that, in the crack formation phase, the dashed line ADC or
ADE is followed, instead of the ABC or ABE trajectories. This is due
to the numerical stability of the algorithm in the two-dimensional
case.
In Fig. 10 the behaviour of a tension chord according to the TCM
is illustrated for four different reinforcement ratios, assuming λ =
1, fsy = 500 MPa, fsu = 625 MPa, Es = 200 GPa, εsu = 0.05, Ø =
02/3 02/3
16 mm, fc0 = 30 MPa, fct = 0.3fc , τb0 = 0.6fc , τb1 = 0.5τb0 .

Consider a tension chord element subjected to constant axial 4.4.2. Tension Chord Model—generalization to two-dimensional stress
tensile forces inducing a symmetric bond stress distribution along conditions
the element. In Fig. 8(c) the stress and strain distributions along According to Fig. 4 the crack spacings in the reinforcement
the chord element are presented for three different working condi- directions are related to each other by srm = srmx cos θr =
tions, corresponding to increasing applied load levels: (1) fully de- srmy sin θr , where srm is the diagonal crack spacing. In Fig. 4
veloped crack pattern with the reinforcement in the elastic range, the stress distribution between the cracks is also illustrated. At
i.e., σsr ≤ fsy ; (2) partial reinforcement yielding along the chord el- the centre between two consecutive cracks the tensile stresses
ement, i.e., σs,min ≤ fsy ≤ σsr ; (3) reinforcement yielding along the transferred to concrete by bond reach their maximum values
entire chord element, i.e., fsy ≤ σs,min . After reinforcement yielding, ∆σc ,x = λx fct and ∆σc ,y = λy fct , where:
the reinforcement strain distribution becomes even more irregular ∆σc ,x srmx srm
exhibiting strong strain localizations in the vicinity of the cracks. λx = = =
This aspect can be successfully reproduced by the TCM and it is an fct srmx0 srmx0 cos θr
(17)
important feature for an accurate calculation of the deformation ∆σc ,y srmy srm
λy = = = .
capacity of RC members. fct srmy0 srmy0 sin θr
With the adopted bond stress–slip relationship, the equilibrium The maximum crack spacings srmx0 and srmy0 for uniaxial tension
along the chord element for the working regime 1 requires that the in the x- and y-directions, respectively, follow from Eq. (16), and
maximum crack spacing is given by: are obtained by proper substitution of φ and ρ by φx and ρx , or φy
fct φ (1 − ρ) and ρy .
srm0 = (16) The parameters λx and λy are no longer limited by the interval
2 τb0 ρ [0.5;1.0] as in the uniaxial tension case. Rather, the average
where ρ = As /Ac is the reinforcement ratio, Ac is the gross cross diagonal crack spacing srm follows from the observation that the
sectional area and As = π φ 2 /4 is the rebar area. In general, in a maximum concrete tensile stress at the centre between the cracks
fully developed crack pattern, the average crack spacing is smaller cannot be greater than λfct (see Fig. 4):
than srm0 , and is given by srm = λsrm0 , with 0.5 ≤ λ ≤ 1. The factor s
σc ,x + σc ,y σc ,x − σc ,y
2
λ takes into account that sections with the concrete stresses equal

to fct can either crack (λ = 0.5) or remain uncracked (λ = 1),
λ fct = + + τxy2 (18)
2 2
giving a lower and an upper limit to the tension stiffening effect,
respectively. which can be worked out to:
s
Using the adopted constitutive laws and the equilibrium
λx + λy λx − λy τxy
 2  2
conditions along the tension chord, the steel stresses at the cracks, λ=a+ + b+ + (19)
as well as their correspondent distribution between the cracks, 2 2 fct
can be obtained from the average strains in a closed form and where the parameters a and b can be expressed in terms of the local
without explicitly calculating the slip between reinforcements and concrete stresses at the cracks as:
concrete. These expressions are given in Box I (see Fig. 8 for
σcr ,n + σcr ,t σcr ,n − σcr ,t cos 2θr − 2τcr ,nt sin 2θr


notation). a= b= .
In Fig. 9 the evolution of the steel stresses at the crack for 2 fct 2 fct
a tension chord under monotonically increasing applied load is (20)

a b c d

Fig. 10. Response of a tension chord according to the TCM: (a) and (b) steel stresses at the cracks and average steel stresses versus average strains; (c) average concrete
stresses versus average strains; (d) strain localization versus steel strain at the cracks.
 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975 1971

τb0 srm τb0 srm

Regime 1 - σsr = Es εsm + σsm = σsr −
φ φ
r 2
τb0 srm
τb1 srm  τbo  
φ
− fsy − Es εsm φ τb1
− Es
Esh
+ Es
Esh
τbo τb1 srm
φ
Regime 2 - σsr = fsy + 2 τbo Es
τb1
− Esh

2
τb0 srm
τb0 σsr − fsy φ τb0
 
σsm = fsy − + σsr − fsy − −1
φ τb1 4 τb1 srm τb1

τb1 srm τb1 srm
Regime 3 - σsr = fsy + Esh εsm − εsy + σsm = σsr −
φ φ
Box I.

Fig. 11. Polar representations of the maximum diagonal crack spacing srm0 .

Eq. (19) can be solved for the average diagonal crack spacing
srm . This equation was derived based on the equilibrium conditions
of the cracked membrane element and on TCM bond stress–slip
relationship, and includes the solution derived by Marti and
Kauffman [7,16] for the CMM as a particular case. If stress-free
rotating cracks perpendicular to the principal tensile direction of
average strains are considered, then σcr ,n = τcr ,nt = 0 and σcr ,t =
−τcr ,nt (tan θr + cot θr ). Substituting the previous relations in (20)
and (19) yields the crack spacing relation proposed for the CMM.
Fig. 11 provides polar representations of the solution, assuming
λ = 1, ρx = 2%, Øx = 16 mm, fct = 2.9 MPa, τb0 = 2fct and
both stress free (τcr ,nt = 0) and interlocked cracks (τcr ,nt 6= 0).
For the cases where interlocked cracks are considered, shear and
normal stresses at the cracks were determined according to the
Contact Density Model. In Fig. 11, the solid line corresponds to the
crack spacing obtained from the expression proposed by Vecchio
and Collins [5].
Once srm is found, srmx and srmy are obtained from Eq. (17). The
steel stresses in the x- and y-reinforcements are calculated from
the equations given in Box I by replacing εsm and srm by εx and srmx ,
or by εy and srmy , respectively. The crack width wr can be calculated
from:
σcm
 
wr = srm εn + ν εt − . (21)
Ec

5. Calculation example

Both the R-CMM and the F-CMM were implemented in a stand-

alone subroutine for the analysis of RC panels under in-plane shear
and axial stresses. Panel responses were calculated with a stress-
driven incremental iterative scheme using the Newton–Raphson Fig. 12. Flowchart of the main steps for the calculation of the stresses given the
method in conjunction with an arc-length procedure [33] to total strain vector.
constrain the norm of the incremental strain. This allows obtaining
the local peaks as well as the softening branch of the response. The given trial total strain vector are given by the flowchart in Fig. 12.
main steps for the calculation of the stresses corresponding to a Prior to cracking linear elastic behaviour is considered using the
1972 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975
Fig. 13. F-CMM: analysis of panel PP1 tested by Marti and Meyboom [34].
common elasticity matrix. The formulation of the tangent stiffness
matrix, of the unloading/reloading behaviour, and of the uncracked
nonlinear behaviour of concrete will be discussed in a posterior
paper dealing with the implementation of the present model in a
finite element program. While in the case of the R-CMM the crack
angle θr is determined from the principal tensile direction of the
current total average strains, in the F-CMM this angle is set in the
step where cracking occurs and is kept in memory. In the R-CMM
both the crack shear stress τcr ,nt and the dilatancy stress σdil are
null. A converged solution is found when the ratio between the
norm of unbalanced stresses at iteration i and the norm of the
calculated stresses at the first iteration is smaller than a predefined
tolerance.
For a clearer explanation of how the F-CMM works, the or-
thotropically reinforced panel PP1 tested by Marti and Mey-
boom [34] is analysed. The panel, with dimensions 1626 × 1626 ×
286 mm, was submitted to pure shear loading σx : σy : τxy = 0 : 0 :
1. In the experiment, a slow sliding failure occurred along a crack
that had formed in the early load stages. A strip was pushed up
and away relative to the rest of the specimen. The y-reinforcement
yielded, whereas the x-direction reinforcement remained elastic.
Some localized cover spalling was observed. The adopted mate-
rial properties are: fc0 = 27 MPa; fsy,x = 479 MPa; fsu,x =
667 MPa; εsu,x = 0.090; Øx = 19.5 mm; ρx = 1.94%; fsy,y =
480 MPa; fsu,y = 640 MPa; εsu,y = 0.091; Øy = 11.3 mm; ρy =
0.647%. The results are presented in Fig. 13 and, where available,
experimental data is presented for comparison.
The solid lines refer to the analysis considering bonded rein-
forcement (λ = 1), while the dashed lines refer to the analysis
considering λ = 0, which is equivalent to disregarding tension
stiffening effects. From the shear stress–strain curves it can be
observed that tension-stiffening results in a markedly stiffer re-
sponse. The ultimate load is also somewhat increased due to the
reduction of εn , as this affects both the amount of compressive soft-
ening and the crack shear capacity. The stress–deformation curves
compare favourably with the experimental results. The good fit ex-
hibited by the τxy –εx and τxy –εy curves seems to confirm the ade-
quacy of the tension stiffening model. In the numerical analysis, the
peak stress was achieved before concrete crushing and after ex-
tensive crack slipping. Therefore, the calculated failure mode can
be deemed a shear sliding failure followed by concrete crushing,
which agrees with the experimental observations.
In the F-CMM, cracks are normal to the concrete principal
tensile stress direction at impending cracking, which is very close
to the maximum principal applied stress direction. Hence, in this
case, cracks form and remain fixed at θr = 45°. The angle θe ,
which defines the direction of the principal tensile average strains
does not remain constant at 45° due to the crack slip. The crack
shear slip increases sharply during the crack formation phase,
after which follows a steadily increasing phase until the weaker
reinforcement yields. At this point θe starts to increase faster. A
good correspondence with the experimental measurements can
also be observed here. The evolution of the angle defining the
principal average concrete tensile stress direction, θcm , is also
 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975 1973

Fig. 14. Comparison of the calculated with the experimental shear strengths of RC panels tested under in plane shear and axial forces. Basic data given in Table 1.

presented in Fig. 13. It can be observed that the model can describe a
the lag between θcm and θe , as reported by Vecchio [6]. Also relevant
is the evolution of the angle θcr , which defines the direction normal
to the principal compressive concrete stresses at the cracks. After
yielding of the weaker reinforcement, θcr and θe have similar
values. However, near failure, a drift between the two angles is
observed due to the increasing crack shear slip rate. This is a clear
indication of a shear sliding failure.
A comparison between the predicted and measured maximum
crack openings is also presented with a good match being obtained.
Near failure, with the increasing shear slip, the accompanying
opening movement due to crack dilatancy leads the crack width
to increase more rapidly.
Regarding the reinforcement, the average steel stresses and
the steel stresses at the cracks can be discerned. Due to crack
kinematics, the stresses in the y-reinforcement kept increasing b c
after the peak load was reached and only started to decrease after
concrete crushing. The analysis of the local steel strains at the
cracks reveals that, for the y-reinforcement, the local strain εsr ,y
is more than three times the corresponding average strain εy .
In the analysis with λ = 1, the average stresses transmitted
by bond action, ∆σcm,x and ∆σcm,y , can be added to the concrete
stresses at the cracks in order to obtain the average concrete stress
field. Observing the stress components in the n-direction, it can
be discerned the compressive stresses arising from crack shear
transfer mechanisms, σcr ,n , the projection along the n-direction of
the tension stiffening stresses, ∆σcm,n , and the resulting ‘‘apparent’’
tension stiffening diagram σcm,n –εn . The onset of y-reinforcement
yielding is reflected by changes in the slope of these diagrams. Also
the concrete shear stress component in the local n–t coordinate Fig. 15. Failure envelope given by the F-CMM compared to experimental evidence:
(a) three-dimensional view; (b) section by ρx = 1.7%; (c) section by ρx = 2.8%.
system is shown. The difference between the shear stress at the
cracks and the average shear stress can be observed. Regarding
the compressive concrete stresses in the t-direction, it can be The coefficient χ = ωy /ωx is an indicator of the reinforcement
confirmed that the compressive stresses at the cracks are higher orthotropy degree. In Fig. 14 a summary of the results of the
than the average ones. validation campaign is presented. The average values of the
From the above description one can confirm that the proposed τxyexp
,u /τxy,u ratio are close to unity, which shows that both models
calc

model allows a rational description of the complex stress field in a can be used to calculate the shear strength of RC panels with a
cracked membrane element. good level of precision. The coefficients of variation are bellow 10%,
which is also a good result having in mind the complexity of the
6. Validation problem and the diversity of the tested panels.
The failure envelope given by the F-CMM is presented in
Panels from 9 different testing campaigns, tested at 3 different Fig. 15(a). The calculations were made considering: fc0 =
testing facilities, and covering a wide range of reinforcement ratios 45 MPa; fsy,x = fsy,y = 450 MPa; fsu,x = fsu,y = 580 MPa; εsu,x =
and concrete strengths were analysed. The basic data for the 54 εsu,y = 50h and Øx = Øy = 16 mm. The corresponding envelope
panels used in the validation procedure is presented in Table 1. for the R-CMM can be found in Ref. [16]. In Fig. 15(a) and (b)
For further details regarding the panels properties refer to the
two sections of the failure surface are presented, corresponding to
references mentioned in the table. The reinforcement content of
two different x-reinforcement ratios. Experimental failure loads of
each panel is given in terms of the mechanical reinforcement
panels with similar properties are also included. The corresponding
ratios, which are here defined as:
shear stress–strain curves are depicted in Fig. 16 for the F-CMM.
ρx fsy,x ρy fsy,y For the panel series with ρx ≈ 1.7%, both failure envelopes are
ωx = 02/3
ωy = 02/3
(22)
fc fc reasonably fitted to the experimental results. The underestimation
1974 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975

Table 1
Summary of validation results.
Panel Ref. fc0 ωx χ τu,exp τxyexp
,u /τxy,u
calc

1/3
[MPa] [MPa ] [–] [MPa] F-CMM λ = 1 F-CMM λ = 0 R-CMM λ = 1 R-CMM λ = 0

A2 41.3 0.44 1.00 5.38 0.89 0.96 0.90 0.97

A3 41.6 0.63 1.00 7.67 0.92 0.97 0.94 0.98
A4 42.5 1.09 1.00 11.33 0.92 0.94 1.01 1.03
B1 45.2 0.41 0.48 3.97 0.94 1.06 0.89 1.00
B2 [22] 44.1 0.61 0.69 6.13 0.88 0.95 0.87 0.94
B3 44.9 0.60 0.33 4.36 0.95 1.03 0.85 0.93
B4 44.8 1.06 0.19 5.07 0.99 1.09 0.83 0.91
B5 42.8 1.09 0.39 7.16 0.90 0.97 0.84 0.87
B6 43.0 1.08 0.57 9.15 0.93 0.97 0.92 0.95
HB3 66.8 0.46 0.34 4.89 0.99 1.09 0.91 1.01
[35]
HB4 62.9 0.84 0.19 5.71 1.01 1.09 0.87 0.94
VA0 98.8 0.12 1.00 3.35 1.00 1.08 0.99 1.08
VA1 95.1 0.24 1.00 6.29 0.96 1.07 0.98 1.09
VA2 98.2 0.44 1.00 9.73 0.91 0.98 0.92 0.99
VA3 94.6 0.75 1.00 15.13 0.92 0.95 0.93 0.96
VA4 [20] 103.1 1.07 1.00 21.42 0.98 0.99 1.07 1.08
VB1 98.2 0.44 0.54 7.50 0.92 1.01 0.92 1.00
VB2 97.6 0.73 0.33 9.14 0.95 1.02 0.92 0.98
VB3 102.3 1.22 0.19 9.71 0.92 0.98 0.85 0.90
VB4 96.9 0.37 0.33 4.86 0.93 1.04 0.85 0.96
PP1 27.0 1.03 0.33 4.95 0.99 1.06 0.88 0.93
PP2b [34] 28.1 0.97 0.35 5.50 0.99 1.05 0.99 1.03
PP3b 27.7 0.92 0.37 5.50 1.02 1.08 1.01 1.06
SE1 42.5 1.18 0.32 6.77 0.90 0.95 0.84 0.87
SE5 [36] 25.9 1.65 1.00 8.09 0.88 0.89 0.96 0.97
SE6 40.0 1.23 0.11 3.75 0.93 0.99 0.82 0.88
M2 48.3 0.57 0.25 4.28 1.05 1.15 0.95 1.05
M3 48.1 0.57 0.11 3.07 1.03 1.16 1.00 1.10
[37]
M4 44.8 0.99 0.13 4.20 1.01 1.10 0.86 0.94
M5 43.4 1.01 0.06 3.15 1.02 1.12 0.99 1.11
PV4 26.6 0.29 1.00 2.89 1.02 1.11 1.02 1.11
PV10 14.5 0.83 0.56 3.97 1.06 1.10 1.04 1.07
PV12 16.0 1.32 0.14 3.13 1.02 1.06 1.07 1.14
PV16 21.7 0.24 1.00 2.14 1.00 1.11 1.00 1.11
PV19 19.0 1.15 0.26 3.96 0.97 1.00 0.96 1.00
PV20 19.6 1.13 0.32 4.26 0.95 0.97 0.92 0.96
[38]
PV21 19.5 1.13 0.48 5.03 0.92 0.95 0.91 0.94
PV22 19.6 1.12 0.78 6.07 0.88 0.90 0.92 0.93
PV23a 20.5 1.70 1.00 8.87 1.10 1.12 1.20 1.22
PV25a 19.2 2.04 1.00 9.12 1.10 1.12 1.20 1.23
PV27 20.5 1.05 1.00 6.35 0.89 0.90 0.96 0.98
PV28a 19.0 0.93 1.00 6.35 0.96 0.97 1.01 1.03
PC1A 27.9 0.90 0.50 5.61 0.97 1.00 0.93 0.96
PC4a [39] 24.9 0.72 0.65 4.84 0.88 0.96 0.87 0.93
PC7a 28.7 0.56 0.39 3.65 1.00 1.07 0.99 1.05
PA2 43.0 0.70 0.50 6.22 1.01 1.06 0.98 1.02
PHS2 66.1 1.20 0.11 6.66 1.12 1.16 1.13 1.21
PHS3 58.4 1.30 0.22 8.19 1.06 1.10 0.98 1.03
PHS4a 68.5 1.07 0.14 6.91 1.04 1.08 1.04 1.10
PHS5a [40] 52.1 1.32 0.05 4.81 1.10 1.12 1.21 1.29
PHS7a 53.6 1.56 0.31 10.26 1.07 1.11 0.98 1.01
PHS8 55.9 1.34 0.33 10.84 1.15 1.18 1.08 1.11
PHS9a 56.0 1.50 0.20 9.37 1.28 1.32 1.13 1.20
PHS10a 51.4 1.26 0.25 8.58 1.12 1.15 1.06 1.09
Average (54 panels) 0.99 1.04 0.97 1.02
Coefficient of variation [%] 8.0 8.3 9.4 9.4
02/3
a
Panels tested in biaxial tension or compression proportional to shear stress. In these cases ωx/y = (ρx/y fsy,x/y − σx/y,u,exp )/fc .
02/3
b
Prestressed panels in the x-direction. In these cases ωx = (ρx fsy,x + ρp,x fpy,x )/fc .

by the F-CMM of the failure load for panels with very low amounts
of y-reinforcement is due to the consideration of 16 mm rebars in
the weaker direction, while the real panels were reinforced with
small rebars. As expected, in these cases where failure occurs by
shear sliding, the ultimate shear strength was found to be highly
dependent on the crack spacing. The shear stress–strain curves of
Fig. 16 were calculated with the exact rebar diameters and better
agreement can be found. For the panel series with ρx ≈ 2.8%,
the shape of the failure envelope delivered by the F-CMM is closer
to the experimental. The concave shape provided by the R-CMM
shows some overestimation for partially over-reinforced panels.
7. Conclusions
A fixed crack model is proposed for the analysis of RC cracked
membrane elements subjected to in-plane shear and axial stresses.
The present model complements the Cracked Membrane Model of
Kaufmann and Marti by extending its concept to the case of fixed
 M. Pimentel et al. / Engineering Structures 32 (2010) 1964–1975
Fig. 16. F-CMM (λ = 1): shear stress–strain curves for the panels referred in Fig. 15 (see Table 1 for references). Experimental data is given by the curves with symbols.
and interlocked cracks capable of transferring shear and normal
stresses. This enabled a more consistent reproduction of shear
sliding failures in orthotropically reinforced panels. Equilibrium
is formulated in terms of stresses at the cracks and compatibility
is formulated in terms of average strains. The consideration of
the local mechanical effects that take place at the cracks, – such
as aggregate interlock, crack bridging and softened compressed
concrete behaviour –, together with the bond stress transfer
mechanics described according to the Tension Chord Model, allows
a rational derivation of both local and average stress/strain fields
and a deeper understanding of the complex mechanics governing
the behaviour of cracked RC membranes. Nevertheless, the model
was kept sufficiently simple for future implementation in a robust
finite element formulation for structural analysis.
A validation campaign was carried out using a database with
the experimental results of 54 RC panels tested under in-plane
shear and axial stress conditions. Good agreement was found
between the predicted and the observed shear strength, failure
modes and deformational behaviour. The average ratio of the
experimental-to-predicted shear strength of the 54 panels is 0.99,
with a coefficient of variation of 8%. The database covered a wide
range of reinforcement ratios and concrete strengths ensuring that
the model is generally applicable.
Acknowledgements
The support by the Portuguese Foundation for Science and
Technology (FCT) through the Ph.D. grant SFRH/BD/24540/2005
attributed to the first author is greatly acknowledged.
References
[1] fib. A practitioner’s guide to computer-based modelling of structural concrete
(bulletin 45). fib, Lausanne; 2008.
[2] Pang X, Hsu TTC. Behaviour of reinforced concrete membrane elements in
shear. ACI Struct J 1995;92(6):665–79.
[3] Pang X, Hsu TTC. Fixed angle softened truss model for reinforced concrete. ACI
Struct J 1996;93(2):197–207.
[4] Hsu TTC, Zhu RRH. Softened membrane model for reinforced concrete
elements in shear. ACI Struct J 2002;99(4):460–9.
[5] Vecchio FJ, Collins MP. The modified compression-field theory for reinforced
concrete elements subjected to shear. ACI J 1986;83(2):219–31.
[6] Vecchio FJ. Disturbed stress field model for reinforced concrete: formulation.
J Struct Eng 2000;126(9):1070–7.
[7] Kaufmann W, Marti P. Structural concrete: cracked membrane model. J Struct
Eng 1998;124(12):1467–75.
[8] Mitchell D, Collins MP. Diagonal compression field theory—a rational model
for structural concrete in pure torsion. ACI J 1974;71(8):396–408.
[9] Collins MP. Towards a rational theory for RC members in shear. J Struct Div
ASCE 1978;104(4):649–66.
[10] Marti P, Alvarez M, Kaufmann W, Sigrist V. Tension chord model for structural
concrete. Struct Eng Int 1998;98(4):287–98.
[11] Belleti B, Cerioni R, Ivori I. Physical approach for reinforced-concrete (PARC)
membrane elements. J Struct Eng 2001;127(12):1412–26.
1975
[12] Sato Y, Fuji S. Local stresses and crack displacements in reinforced concrete
elements. J Struct Eng 2002;128(10):1263–71.
[13] Soltani M, Maekawa K. Computational model for post cracking analysis of RC
membrane elements based on local stress–strain characteristics. Eng Struct
2003;25:993–1007.
[14] Soltani M, An X, Maekawa K. Localized nonlinearity and size-dependent
mechanisms of in-plane RC element in shear. Eng Struct 2005;27:891–908.
[15] Muttoni A, Schwartz J, Thurlimann B. Design of concrete structures with stress
fields. Basel: Birkhäuser Verlag; 1997.
[16] Kaufmann W. Strength and deformations of structural concrete subjected
to in-plane shear and normal forces. Doctoral thesis. Zurich: Swiss Federal
Institute of Technology Zurich; 1998.
[17] Belarbi A, Hsu TTC. Constitutive laws of softened concrete in biaxial
tension–compression. ACI Struct J 1995;92(5):562–73.
[18] fib. Bulletin no 1: structural concrete. text book on behaviour, design and
performance, vol. 1. fib, Lausanne; 1999.
[19] Zhang LX, Hsu TTC. Behaviour and analysis of 100 MPa concrete membrane
elements. J Struct Eng 1998;124(1):24–34.
[20] Zhang LX. Constitutive laws of reinforced membrane elements with high
strength concrete. Doctoral thesis. Houston: University of Houston; 1995.
[21] Belarbi A. Stress-strain relationships of reinforced concrete in biaxial
tension–compression. Doctoral thesis. Houston: University of Houston; 1991.
[22] Pang X. Constitutive laws of reinforced concrete in shear. Doctoral thesis.
Houston: University of Houston; 1991.
[23] Hordijk DA. Tensile and tensile fatigue behaviour of concrete: experiments,
modelling and analyses. Heron 1992;37(1):1–79.
[24] CEB. CEB-FIP model code 1990. London: Thomas Telford; 1993.
[25] Bazant ZP, Oh BH. Crack band theory for fracture of concrete. Mater Stuct 1983;
16(1):155–77.
[26] Walraven JC, Reinhardt HW. Theory and experiments on the mechanical
behaviour of cracks in plain and reinforced concrete subjected to shear
loading. Heron 1981;26(1A).
[27] Li B, Maekawa K, Okamura H. Contact density model for stress transfer across
cracks in concrete. J Fac Eng Tokyo 1989;XL(1):9–52.
[28] Alvarez M. Einfluss des Verdunverhaltens auf das Vermormungsvermögen
von Stahlbeton. Doctoral thesis. Zurich: Swiss Federal Institute of Technology
Zurich; 1998.
[29] Sigrist V. Zum Verformungsvermögen von Stahlbetonträgern. Doctoral thesis.
Zurich: Swiss Federal Institute of Technology Zurich; 1995.
[30] Manzoli OL, Oliver J, Huespe AE, Diaz G. A mixture theory based method for
three-dimensional modeling of reinforced concrete members with embedded
crack finite elements. Comput Concr 2008;5(4):401–16.
[31] Oliver J, Linero DL, Huespe AE, Manzoli OL. Two-dimensional modeling of
material failure in reinforced concrete by means of a continuum strong
discontinuity approach. Comput Methods Appl Mech Eng 2008;197:332–48.
[32] Kenel A, Nellen P, Frank A, Marti P. Reinforcing steel strains measured by bragg
grating sensors. J Mater Civil Eng 2005;17(4):423–31.
[33] Crisfield MA. A fast incremental/iterative solution procedure that handles
snap-through. Comput & Structures 1981;13:55–62.
[34] Marti P, Meyboom J. Response of prestressed concrete elements to in-plane
shear forces. ACI Struct J 1992;89(5):503–14.
[35] Zhang LX. Constitutive laws of reinforced elements with medium-high
strength concrete. Master thesis. Houston: University of Houston; 1992.
[36] Kirschner U. Investigating the behaviour of reinforced concrete shell elements.
Doctoral thesis. Toronto: University of Toronto; 1986.
[37] Chintrakarn R. Minimum shear steel and failure modes diagram of reinforced
concrete membrane elements. Master thesis. Houston: University of Houston;
2001.
[38] Vecchio FJ. The response of concrete to in-plane shear and axial stresses.
Doctoral thesis. Toronto: University of Toronto; 1981.
[39] Vecchio FJ, Chan CCL. Reinforced concrete membrane elements with
perforations. J Struct Eng 1990;116(9):2344–60.
[40] Aspiotis J. Compression softening of high strength reinforced concrete
elements. Master thesis. Toronto: University of Toronto; 1993.