You are on page 1of 26

INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS

Int. J. Numer. Anal. Meth. Geomech. (2011)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/nag.1096

Review and enhancement of 3D concrete models for large-scale


numerical simulations of concrete structures

B. Valentini, G. Hofstetter*,†
University of Innsbruck, Institute of Basic Sciences in Civil Engineering, Unit for Strength of Materials and Structural
Analysis, Technikerstr. 13, A-6020 Innsbruck, Austria

ABSTRACT
The present paper focuses on selected plasticity and damage-plasticity models for describing the 3D
material behavior of concrete. In particular, a plasticity model and a damage-plasticity model are reviewed
and evaluated. Based on the results of the evaluation, enhancements are proposed, aiming at improving the
correspondence between predicted and observed material behavior and aiming at implementing a robust and
efficient stress update algorithm in a finite element program for performing large-scale 3D numerical
simulations of concrete structures. The capabilities of the concrete models are demonstrated by 3D numerical
simulations of benchmark tests with combined bending and torsional loading and combined compression and
shear loading and by a large-scale 3D finite element analysis of a model test of a concrete arch dam. Copyright
© 2011 John Wiley & Sons, Ltd.

Received 4 May 2011; Revised 22 July 2011; Accepted 22 July 2011

KEY WORDS: concrete model; constitutive model; plasticity theory; damage mechanics; damage-plasticity
model; concrete structure; arch dam; finite element method; numerical simulation

1. INTRODUCTION

In the past twenty years many 3D constitutive models for concrete were proposed. They describe
the mechanical behavior of concrete for a wide range of 3D stress paths, taking into account
hardening and/or softening material behavior, cracking due to tensile loading and crushing
due to compressive, shear or mixed mode loading. Comprehensive overviews of constitutive
models for concrete were published in [1–3] and a comparison of 3D concrete models can be
found, e.g., in [4].
Most of the concrete models are based on the flow theory of plasticity, the theory of continuum
damage mechanics, a combination of both theories or on micro-plane theory.
One of the first three-dimensional material models for plain concrete based on the flow theory of
plasticity was developed by Willam and Warnke [5]. It is characterized by a single failure surface
for describing failure due to compressive stresses. Based on the strength criterion by Leon [6],
Pramono [7] proposed a plasticity model for describing the triaxial behavior of concrete with
hardening in the pre-peak regime and fracture energy based softening in the post-peak regime.
Based on the constitutive models by Willam [5] and Pramono [7] Etse [8, 9] developed the
Extended Leon model. Similar concrete models characterized by a single failure criterion and
nonlinear hardening/softening laws were presented by Menetrey [10] and Kang [11, 12] and a
single surface plasticity model for soil, rock and concrete was proposed by Lade [13]. Bićanić

*Correspondence to: G. Hofstetter, University of Innsbruck, Institute of Basic Sciences in Civil Engineering, Unit for
Strength of Materials and Structural Analysis, Technikerstr. 13, A-6020 Innsbruck, Austria.

E-mail: Guenter.Hofstetter@uibk.ac.at

Copyright © 2011 John Wiley & Sons, Ltd.


B. VALENTINI AND G. HOFSTETTER

[14] modified the concrete model of Willam [5] to improve its robustness for the stress update.
Haufe [15] enhanced the evolution equations of the yield function and the plastic potential of the
concrete model developed by Kang [11]. Pivonka [16] extended two elastic-plastic models for
describing concrete behavior due to high compressive stresses. Different from most classical
plasticity models, Grassl [17] developed a model in which hardening is controlled only by the
volumetric part of the plastic strains. Recently, a single-parameter plasticity model for concrete
was presented by Papanikolaou [18].
Viscoelastic and/or viscoplastic models were developed for describing time-dependent material
behavior, e.g. aging and creep of concrete (see, e.g., [19, 20]). Viscoplastic models can also serve
for regularizing rate-independent plasticity models in case of softening behavior.
Constitutive models based on the theory of damage mechanics describe the degradation of the initial
elastic stiffness tensor caused by propagating micro-cracks. Basically, isotropic damage models and
anisotropic damage models can be distinguished. Isotropic damage models for concrete can be
found, e.g., in [21, 22] and a comparative study of different definitions of damage variables is
contained in [23]. Huber [24] proposed a three-dimensional gradient-enhanced isotropic damage
model. Anisotropic damage models were proposed, e.g., in [25] and by Papa [26], who used two
independent second-order symmetric damage tensors for describing damage due to tensile and
compressive strains, respectively. Based on the argument that damage is not related to the loading
sign but to the micro-crack pattern Desmorat [27] developed a nonlocal anisotropic damage model
which contains only one second-order damage tensor. A comparison of different damage models
with regard to mesh bias effects can be found in [28].
By combining the theories of plasticity and continuum damage mechanics, both irreversible (plastic)
strains and the degradation of the elastic stiffness tensor can be described. Constitutive models for
concrete based on the combination of the theories of plasticity and isotropic damage mechanics can
be found, e.g. in [29–32]. A damage-plasticity model with a rate-dependent viscoplastic extension
was proposed by Lee [33]. Huber [24] and Schütt [34] enhanced a multi-surface plasticity model by
scalar isotropic damage. Grassl [35] developed a damage-plasticity model in which material
hardening is described by the plastic part of the model whereas material softening is modeled by a
scalar isotropic damage law. In [36] and [37] concrete models were proposed in which the plastic
part of the model is enhanced by a non-local damage model. A constitutive model for concrete based on
plasticity theory and anisotropic damage mechanics theory was recently published in [38]. It is based on
the strain-equivalence hypothesis using two damage variables for describing separately compressive and
tensile damage. Recently, Voyiadjis [39] proposed a concrete model based on the strain-energy-
equivalence hypothesis and Abu Al-Rub [40] presented a nonlocal gradient-enhanced coupled plasticity
anisotropic damage model.
Constitutive models based on the micro-plane theory belong to the category of meso-scale models.
The micro-plane theory allows describing anisotropic material behavior with basic (isotropic)
constitutive models formulated at the so called micro-planes. The stresses computed at the micro-
planes are integrated over all micro-planes, which represent the surface of a unit sphere, to obtain
the stress state at the macroscopic level. First applications of micro-plane models for describing the
behavior of concrete were presented in [41] and [42], and further developments were published, e.g.,
in [43–46,28].
The present paper focuses on selected plasticity and damage-plasticity models for describing the 3D
material behavior of concrete. In particular, (i) the Extended Leon model, developed by Etse [8] and
modified by Pivonka [16], and (ii) the damage-plasticity model proposed by Grassl and Jirásek [35]
are reviewed and evaluated. Based on the results of the evaluation of the concrete models,
enhancements are proposed aiming at improving the correspondence between predicted
and observed material behavior and aiming at implementing a robust and efficient stress
update algorithm in a finite element code for performing large-scale 3D numerical simulations of
concrete structures.
The capabilities of the developed numerical models will be demonstrated by 3D numerical
simulations of lab tests at the integration point level, of small-scale benchmark tests with combined
bending and torsional loading and combined compression and shear loading and by a large-scale 3D
finite element analysis of a model test of a concrete arch dam.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

2. REVIEW OF SELECTED 3D CONSTITUTIVE MODELS FOR CONCRETE

2.1. Extended Leon model


The Extended Leon model is a single-surface plasticity model with nonlinear hardening and softening
for describing the material behavior of concrete. It was developed by Etse [8, 9] and was modified later
by Pivonka [16]. The latter version is described and employed in this paper.
The yield function of the Extended Leon model (ELM)

"   rffiffiffi #2
1  qh ðah Þ m rr ðθÞ 2 3 rrðθÞ
fELM ðs ; r; θ; ah ; as Þ ¼
m
s þ pffiffiffi þ þ
fcu2 6 2 fcu
  (1)
q2h ðah Þ rr ðθÞ
þ ms ðas Þ s þ pffiffiffi  q2h ðah Þqs ðas Þ
m
fcu 6
is formulated in terms of the mean stress sm, the deviatoric radius r, the Lode angle θ, the internal
strain-like hardening variable ah and the internal strain-like softening variable as; fcu denotes the
uniaxial compressive strength of concrete. The deviatoric shape function r(θ) is given as [5]

4ð1  e2 Þcos2 ðθÞ þ ð2e  1Þ2 h pi


r ðθ Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; θ ¼ 0; ; (2)
2ð1  e2 ÞcosðθÞ þ ð2e  1Þ 4ð1  e2 Þcos2 ðθÞ þ 5e2  4e 3

where e = rt/rc denotes the eccentricity parameter with 0.5 ≤ e ≤ 1. It is defined as the ratio of the
deviatoric part of the strength at the tension meridian rt = r(θ = 0) to the deviatoric part of the
strength at the compression meridian rc = r(θ = p/3). For the limiting cases of e = 0.5 and e = 1 the
yield function in deviatoric planes is of triangular and circular shape, respectively. Pivonka [16]
determines the eccentricity parameter by computing rt and rc from the yield function proposed by
Pramono and Willam [7]. In contrast to Pivonka, Etse [8] used the explicit equation
 
0:08 m ftu
e ¼ 0:7  s   0:035ð1  qs Þ (3)
fcu 3

for determining the eccentricity parameter. In (3) ftu denotes the tensile strength and qs is the
decohesion parameter, which will be explained later in the context of the softening behavior.
According to (3) the eccentricity parameter depends on qs and on the mean stress.
The constitutive equations are given as

s ¼ C : ð«  «p Þ (4)

with s, C, « and «p denoting the stress tensor, the elastic stiffness tensor, the total strain tensor and the
plastic strain tensor, respectively. The plastic strain rate is described by the flow rule:

: p ¼ g:
@gELM
« (5)
@s
with the consistency parameter g: and the plastic potential

"  2 rffiffiffi #2
1  q h ð ah Þ rr ð θ Þ 3 rr ð θÞ
gELM ðsm ; r; θ; ah ; as Þ ¼ 2
sm þ pffiffiffi þ þ
fcu 6 2 fcu
  (6)
q2 ðah Þ rrðθÞ
þ h mg ðsm ; ah Þ þ ms ðas Þ pffiffiffi  q2h ðah Þqs ðas Þ:
fcu 6
Hence, it follows from (1) and (6) that the flow rule (5) is associated in deviatoric planes and non-
associated in meridional planes.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

Hardening behavior of the Extended Leon model is described by the normalized strength parameter
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qh0 þ ð1  qh0 Þ ah ð2  ah Þ ifah < 1
qh ðah Þ ¼ ; (7)
1 ifah ⩾ 1
where qh0 ¼ fcy =fcu denotes the initial value of qh, which represents the ratio of the elastic limit stress
under compressive loading, fcy, to the uniaxial compressive strength.
The evolution law of the internal strain-like hardening variable is given as
: :
jj « p jj : h ðsm; r; θ; ah ; as Þ
ah ðsm ; r; θ; ah ; as Þ ¼ ¼ gh (8)
xh ðsm Þ
with the hardening ductility parameter
 2  
sm sm
x h ðs m Þ ¼ A h þ Bh þ Ch : (9)
fcu fcu

Since xh is increasing with increasing hydrostatic pressure, the rate of the internal hardening variable
ah is decreasing for increasing values of the confining pressure. Recommended values for the model
parameters Ah, Bh and Ch are provided for low confinement by Etse [8] and for high confinement by
Pivonka [16].
Softening behavior is controlled by the decohesion parameter

1 ifah < 1
qs ðas Þ ¼ n ; (10)
eðas =au Þ ifah ⩾ 1
which is driven by the strain-like internal softening variable as. In the present work for tensile softening
n = 1 is assumed and au ¼ GIf =ðftu lchar Þ with GIf as the specific mode I fracture energy of concrete, ftu
representing the uniaxial tensile strength and lchar as the characteristic length of the finite element for
regularizing the softening behavior. For softening in the compressive regime n = 2 is assumed.
The evolution law of the internal strain-like softening variable as is defined as

: :
jj < «p > jj :
as ðsm ; r; θ; ah ; as Þ ¼ ¼ ghs ðsm ; r; θ; ah ; as Þ; (11)
xs ðsm Þ
where < • > represents the MACAULEY-brackets and

 4 ðsm Þ þ Bs R
xs ðsm Þ ¼ As R  2 ðsm Þ þ 1 (12)
is the softening ductility parameter with
   
  1 ftu
Rðs Þ ¼ R speak ¼
m m
s m
 (13)
fcu peak 3
depending on the maximum value of the mean stress sm peak , which is constant during softening.
Recommended values for the model parameters As and Bs are provided in [16].
The friction parameter ms in (1) is defined as

m0 ifah < 1
ms ðas Þ ¼ ; (14)
mr  ðmr  m0 Þqs ðas Þ ifah ⩾ 1

where the residual friction parameter


 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð2 þ aTP Þ m0 þ 4  8aTP þ 4a2TP þ m20
mr ¼ (15)
2ð2 þ aTP Þ

is given according to [16] in terms of the initial friction parameter m0 ¼ fcu2  ftu2 =ðfcu ftu Þ, which can
be derived from the strength criterion of Leon [6] and aTP, the latter accounting for the transition from
brittle to ductile failure. In [16] aTP = 8 for fcu = 22N/mm2 and aTP = 6 for fcu = 35N/mm2 is provided.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

It follows from (10) and (14) that during hardening qs = 1 and ms = m0 holds.
For mg(sm ; ah) in (6) only the first derivative with respect to the mean stress is defined as
@mg
¼ Dðah ÞeEðah ÞR ðs Þ þ F
2 m
(16)
@s m
with
 
1 ftu
Rðsm Þ ¼ sm  (17)
2fcu 3
and the model parameters

Dðah Þ ¼ 8:675 þ 5:115e5½1qh ðah Þ ; E ðah Þ ¼ 14:956 þ 6:736e5½1qh ðah Þ ; F ¼ 6:3; (18)
calibrated from measurement data of the dilatancy from three experiments at different levels of
confinement [8].

2.2. Damage-plasticity model by Grassl and Jirásek


The damage-plasticity model by Grassl and Jirásek [35] is a single-surface concrete model with
isotropic hardening, formulated within the framework of plasticity theory, and isotropic softening,
described on the basis of damage theory.
The yield function of the damage-plasticity model (DPM) is given in terms of the effective
 which are related to the nominal stresses s by the scalar damage variable o:
stresses s,

 ¼ ð1  oÞC : ð«  «p Þ
s ¼ ð1  o Þs (19)
and one strain-like internal hardening variable ap as
"   2 rffiffiffi #2
 1  qh ap 
r 3r 
fp;DPM s  ap
; θ;
m ; r ¼ m þ pffiffiffi þ
s þ
2
fcu 2 f
6 cu
  (20)
q2h ap  Þ
 r ðθ
r 
þ m þ pffiffiffi  q2h ap
m0 s
fcu 6
with s  and θ
m, r  denoting the effective mean stress, the effective deviatoric radius and the effective
Lode angle, respectively.
The shape of the yield function in deviatoric planes is controlled by the deviatoric shape function
(2), replacing in the latter equation θ by θ . The eccentricity parameter e is given in terms of the
uniaxial tensile and compressive strengths ftu and fcu and the biaxial compressive strength fbu as [1]
1þe 2
ftu fbu  fcu
2
e¼ with e ¼ : (21)
2e fbu fcu2  ftu2

The friction parameter

2
fcu  ftu2 e
m0 ¼ 3 (22)
fcu ftu e þ 1
depends on the eccentricity parameter. In contrast to the Extended Leon model, the eccentricity
parameter of the damage-plasticity model is independent of the effective mean stress. Nevertheless,
the shape of the yield surface in deviatoric planes depends on the effective mean stress, since only
 , is multiplied by the deviatoric shape function
the linear term of the yield surface (20), containing r
r(θ). Hence, basically, the yield surfaces of the Extended Leon model (1) and of the damage
plasticity model (20) are of similar shape. Figure 1 shows the yield surface (20) in the principal
stress space for five different states of hardening.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

Figure 1. Yield surface of the damage-plasticity model in the principal stress space for different hardening
states.

The plastic strain rate is described by the flow rule


: : @gp;DPM
«p ¼ g (23)

@s
with the plastic potential
"    rffiffiffi #2
 1  qh ap  2
r 3r 
gp;DPM s ; ap
 ;rm
¼ 2
 þ pffiffiffi þ
s m
fcu 6 2 fcu
   ; (24)
2
qh ap 
r
þ m Þ þ m0 pffiffiffi
m g ðs
fcu 6

which is characterized by a circular shape in deviatoric planes. The shape of the plastic potential in the
meridional plane is controlled by
ftu =3Þ=ðBg fcu Þ
m Þ ¼ Ag Bg fcu eðs
m
m g ðs (25)

with the model parameters


 
3ftu m0 1 þ ftu
fcu
1
3
Ag ¼ þ ; Bg ¼     (26)
fcu 2 ln Ag  ln 2Df  1  ln 3 þ m20 þ ln Df þ 1

with Df = 0.85. It follows from (20) and (24) that the flow rule of the damage-plasticity model is non-
associated in both deviatoric and meridional planes.
Hardening is described by the normalized strength parameter
(  
 qh0 þ ð1  qh0 Þap a2p  3ap þ 3 ifap < 1
qh ap ¼ : (27)
1 ifap ⩾ 1

The evolution of the strain-like internal hardening variable is given as

: m ; r; θ; ap ¼ jj«: p jj 4cos2ðθÞ ¼ gh


ap s
: psm; r; θ; ap (28)
xh ðsm Þ

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

with the hardening ductility parameter



Ah  ðAh  Bh ÞeRh ðs Þ=Ch
m
m Þ ⩾ 0
ifRh ðs
 Þ¼
xh ðs m
m Þ=Fh
Rh ðs ; (29)
Eh e þ Dh m Þ < 0
ifRh ðs
where
m 1
s
m Þ ¼ 
R h ðs  (30)
fcu 3
and Ah, Bh, Ch, Dh, Eh and Fh are model parameters. Recommended values are provided in [35].
Softening material behavior of the damage-plasticity model is described by an isotropic damage law.
The damage loading function is formulated in the strain-space as

fd;DPM ð«; «p ; ad Þ ¼ ~eð«; «p Þ  ad ; (31)

where ~e represents the equivalent strain and ad is the strain-like internal softening variable. The rate of
the strain-like internal softening variable is given as
8
ifap < 1
: < 0:p;vol
ad ¼ ifap ⩾ 1; (32)
: e:
xs ðe p;vol Þ

: :
where e p;vol ¼ e pij dij denotes the volumetric plastic strain rate and the softening ductility parameter

8 : :
: p;vol < 1 þ As 
R2s ep;vol
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ifRs e p;vol < 1
xs e ¼ : p;vol : (33)
: 1 þ As 4 Rs e 3 ifRs e p;vol ⩾ 1

controls the evolution of the internal strain-like softening variable. In (33) As = 15 is recommended and

: e: p;vol
Rs e p;vol ¼ :⊝p;vol : (34)
e

:
The negative part of the volumetric plastic strain rate e⊝p;vol is calculated from the principal values of
:
the plastic strain rate e pI , I = 1, 2, 3, as

:
e⊝p;vol ¼
X
3
:
< e pI > : (35)
I¼1

The evolution law of the damage variable is given in [35] as

oðad Þ ¼ 1  ead =ef (36)

with ef controlling the slope of the softening curve. The latter is determined from the specific fracture
energy and the characteristic length of the finite element [35].
Since the evolution law of the damage variable (36) produces sharp bends of compressive stress-
strain curves in the softening regime, (36) is replaced in the present work by [49]
 m  m 1
 m ðad =ef ;c Þ2
o s d 
d ; ad ¼ 1  X s d e
2  1  X s (37)
1 þ eafd;t

with ef, t and ef, c controlling the slope of the softening curve and

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

8
> fcu
>
>0 sm
if d⩽
 m < m 3
d ¼ 3
X s sd fcu (38)
>
> þ1 if  < s md < 0
> fcu
: 3
1 sd ⩾0
if m

determining the weight of the second and third term in (37). Since s m
d is equal to the effective mean
stress at the onset of softening, X is a constant parameter. For tensile loading with s md ⩾0 the shape
of the softening curve is controlled only by the hyperbolic function which results in a steeper initial
descent of the softening envelope. In contrast, for compressive loading with s m
d ⩽  fcu =3 the shape
of the softening curve is controlled only by the quadratic exponential function. Thus, the sharp bend
of stress-strain curves at the transition from hardening to softening, produced by the original damage
law is avoided. For fcu =3 < s m
d < 0 multi-axial combined tension-compression loading is
controlled by a combination of both functions. Figure 2 shows the ultimate stress envelope of
concrete for plane stress conditions together with the parameter X according to (38).

3. STRESS UPDATE ALGORITHM

The incremental-iterative solution procedure of a nonlinear FE-analysis requires the update of the
stresses and internal state variables at each integration point of the discretized continuum. Assuming
the equilibrium state and, hence, the stresses sn, to be known at time instant tn, for plasticity models
the stresses at time instant tn + 1 follow from (4) as


snþ1 ¼ C : «nþ1  «pnþ1 ¼ sn þ C : Δ«enþ1 (39)

with Δ«enþ1 ¼ Δ«nþ1  Δ«pnþ1 denoting the increment of the elastic strains for the current time step.
:
The equations for the plastic strain rate e p and for the rates of the internal state variables, the latter
:
summarized in the vector q, are integrated from tn to tn + 1 by the backward Euler method. The resulting
system of nonlinear equations is solved within the framework of the return mapping algorithms by
means of Newton’s method [47].
In order to increase the robustness of the stress update algorithm for larger strain increments it is
enhanced by a substepping method proposed by Pérez-Foguet [48]. The basic idea of the
substepping scheme is to split the total strain increment Δ«n + 1 into m subincrements Δ«n + k/m, i.e.,

Figure 2. Ultimate stress envelope for plane stress states together with the parameter X for the enhanced
damage evolution law.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

X
m X
m
Δ«nþk=m
Δ«nþ1 ¼ Δ«nþk=m ¼ bnþk=m Δ«nþ1 with bnþk=m ¼ ; (40)
k¼1 k¼1
Δ«nþ1

and performing the stress update algorithm consecutively for all subincrements of the total strain
increment. In (40), 0 < bn + k/m ≤ 1 defines the ratio between the k-th total strain subincrement and
the total strain increment. The size of the subincrements is controlled adaptively to guarantee
convergence of the stress update and to limit the error of the solution to a user-defined threshold
value [49].
The substepping procedure starts from the converged values sn, «pn and qn of the previous increment
n. By analogy to (39), the stresses for the k-th subincrement are given as
 
snþk=m ¼ C : «nþk=m  «pnþk=m ¼ snþðk1Þ=m þ C : Δ«enþk=m (41)
and a trial stress
 
p
strial
nþk=m ¼ C : «nþk=m  «nþðk1Þ=m ¼ snþðk1Þ=m þ C : Δ«nþk=m (42)

is computed by assuming elastic material behavior for the current subincrement. Hence, the plastic
strains and the internal state variables are identical to the respective converged values of the
previous subincrement n + (k  1)/m.
Evaluation of the yield function on the basis of the trial state determines whether the material
behavior is elastic or plastic for the current subincrement. In the latter case, the stresses sn + k/m,
plastic strains «pnþk=m and internal state variables qnþk=m are updated by means of the return mapping
algorithm described in the following subsection. They serve as starting values for the subsequent
subincrement. This scheme is applied consecutively for all subincrements. The solution for the last
subincrement k = m provides the stresses sn + 1, plastic strains «pnþ1 and internal state variables qnþ1
at tn + 1.
Once the increment of the plastic strains Δ«pnþ1 is known, the damage variable on + 1 for the damage
plasticity model can be updated directly from (36) or (37), respectively. Substituting on + 1 into the
constitutive equations for the plastic damage model at tn + 1, which follow from (19), yields the
nominal stresses

 nþ1 ¼ ð1  onþ1 ÞC : «nþ1  «pnþ1 :
snþ1 ¼ ð1  onþ1 Þs (43)

3.1. Return mapping algorithm with adaptive substepping


In case of plastic material behavior the flow rule (5) or (23) and the hardening/softening laws (8) and
(11) or (28) are integrated for the time subincrement Δtn + k/m by the backward Euler method yielding
the plastic strains and the internal hardening variables

:
«pnþk=m ¼ «pnþðk1Þ=m þ «pnþk=m Δtnþðk1Þ=m ¼ «pnþðk1Þ=m þ Δgnþk=m 
@gnþk=m
; (44)
@snþk=m
f f

¼Δepnþk=m

:
qnþk=m ¼ qnþðk1Þ=m þ qnþk=m Δtnþk=m ¼ qnþðk1Þ=m  HΔgnþk=m hnþk=m ; (45)
¼ Δqnþk=m

with H representing the generalized hardening/softening moduli and Δ«pnþk=m, Δqnþk=m and Δgnþk=m ¼
:
gnþk=m Δtnþk=m denoting the subincrements of the plastic strains, the internal state variables and the
consistency parameters, respectively.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

The stresses at time tn + k/m (41) can be rewritten by means of (42) and (44) in terms of the elastic
p
predictor strial
nþk=m and the plastic corrector C : Δ«nþk=m as

p @gnþk=m
snþk=m ¼ strial
nþk=m  C : Δ«nþk=m ¼ snþk=m  C : Δgnþk=m 
trial
: (46)
@snþk=m

Rearranging the terms in (45) and (46) and combining them with the vector f nþk=m , containing the
yield conditions fi, n + k/m = 0 for all active yield functions, results in a system of algebraic equations:

@gnþk=m
snþk=m þ strial
nþk=m  C : Δgnþk=m  ¼ 0;
@snþk=m
(47)
qnþk=m þ qnþðk1Þ=m  HΔgnþk=m hnþk=m ¼ 0;
f nþk=m ¼ 0:

Although both concrete models are characterized by a single yield surface in Eqs. (44) to (47) Δgn + k/m,
gnþk=m and f nþk=m denote vectors of consistency parameters, plastic potentials and yield functions,
respectively. This is due to the fact that for both models the yield function is characterized by vertices
at the points of intersection with the hydrostatic axis in both the tensile and the compressive region.
Hence, cut-off functions are introduced for the treatment of the vertices within the framework of multi-
surface plasticity. They are obtained (1) or (20) by setting the deviatoric radius equal to zero [49].
The system of nonlinear equations (47) is solved for sn + k/m, qnþk=m and Δgn + k/m in an iterative
manner by means of Newton’s method, yielding for an abritrary iteration step s
   
ðsÞ ðsÞ ðsÞ
Jnþk=m xnþk=m Δxnþk=m ¼ Rnþk=m xnþk=m ; (48a)

ðkþ1Þ ðsÞ ðsÞ


xnþk=m ¼ xnþk=m þ Δxnþk=m ; (48b)
j kT j   kT
ðsÞ ðsÞ
with xnþk=m ¼ sn þ k=m qn þ k=m Δgn þ k=m and Δxnþk=m ¼ Δsn þ k=m Δqn þ k=m Δ Δgn þ k=m
denoting the vector of unknowns and the update of the vector of unknowns, respectively; Rnþk=m is
the residual vector, which corresponds to the left hand side of the equation system (47) and
2 3
@ 2 gnþk=m @ 2 gnþk=m @gnþk=m
6 I þ C : Δgnþk=m  C : Δgnþk=m  C: 7
6 @s2nþk=m @snþk=m @qnþk=m @snþk=m 7
6 7
6 @hnþk=m @hnþk=m 7
6 7
Jnþk=m ¼ 6 HΔgnþk=m  I þ HΔgnþk=m  Hhnþk=m 7 (49)
6 @snþk=m @qnþk=m 7
6 7
6 @f nþk=m @f nþk=m 7
4 5
0
@snþk=m @qnþk=m

is the Jacobian of the residual vector. The trial state is chosen as starting vector for the iteration. If the
iteration diverges then the size of the actual strain subincrement Δ«n + k/m is reduced and the stress
update procedure is repeated for the current subincrement.

3.2. Damage-elastic-plastic tangent moduli consistent with the adaptive substepping procedure
For achieving a quadratic rate of asymptotic convergence at the structural level the (damage-)elastic-
plastic tangent moduli consistent with the employed stress-update algorithm are required. They are
obtained by exact linearization of the stress-strain relations for the substepping scheme.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

Linearization of the constitutive relations, the flow rule, the hardening/softening law and the
consistency conditions - formulated for subincrement k - leads to the system of equations
0 1 0 1
dsnþk=m dsnþðk1Þ=m þ C : dΔ«nþk=m
B dqnþk=m C
 A ¼ Jnþk=m @ A
1
@  dqnþðk1Þ=m (50)
d Δgnþk=m 0

with the Jacobian Jnþk=m according to (49) computed for the converged values of subincrement k. The
consistent elastic-plastic tangent moduli Cep
nþk=m are derived from (50) as

dsnþk=m
Cep
nþk=m ¼ ¼
d«nþ1

2 0 1 3
6 B C 7
6 s Bdsnþðk1Þ=m C dqnþðk1Þ=m 7
¼ ðPs ÞT J1 6 B þbnþk=m CC þ Pq : 7;
nþk=m 6P :B C 7 (51)
4 @ d«nþ1 A d«nþ1 5
f

f
¼Cep
nþðk1Þ=m
¼Qep
nþðk1Þ=m

h   i
q T 1 s q
Qep ep ep
nþk=m ¼ ðP Þ Jnþk=m P : Cnþðk1Þ=m þ bnþk=m C þ P : Qnþðk1Þ=m (52)

with Ps ¼ ½ I 0 0 T and Pq ¼ ½ 0 I 0 T representing projection matrices according to [50].


For the first subincrement k = 1 the tensors Cep ep ep
nþ0=m and Qnþ0=m are zero. With the aid of Cnþ1=m and
ep
Qnþ1=m , obtained from (51) and (52) for the first subincrement, for the second subincrement k = 2 the
respective tensor Cep
nþ2=m can be derived from (51). Thus, the consistent elastic-plastic tangent moduli
for the total increment, Cep nþ1 , are computed by consecutively applying (51) and (52) for all
subincrements k = 1, . . ., m.
For the damage-plasticity model, the damage-elastic-plastic tangent moduli Cep;d
nþ1, consistent with the
substepping method, are derived from (43) as

dsnþ1  nþ1
ds donþ1
Cep;d
nþ1 ¼ ¼ ð1  onþ1 Þ  nþ1 
s ; (53)
d«nþ1 d«nþ1 d«nþ1
f

Cep
nþ1

where Cepnþ1 is computed according to the update procedure for the plasticity model describe above. The
term don + 1/d«n + 1 in (53) is also updated consecutively at the end of each subincrement k = 1, . . ., m,
similar to the consistent elastic-plastic tangent moduli in (51) [49].

4. VALIDATION EXAMPLES

4.1. Validation by material tests


The Extended Leon model, modified by Pivonka, and the damage-plasticity model by Grassl &
Jirásek as well as the modified version of the latter model were validated by a series of tests
on concrete specimens subjected to different stress paths. Among them were tests conducted by
Hurlbut [51], Gopalaratnam [52], Imran and Pantazopoulou [53], van Mier [54], van Geel [55] and
Burlion [56]. However, for brevity, the validation by material tests is presented here only in an
exemplarily manner by repeated loading tests, documented in [52] and triaxial compression tests,
described in [53].

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

4.1.1. Uniaxial tensile tests with repeated loading. From a series of direct tension tests with
monotonic and repeated loading, conducted by Gopalaratnam and Shah [52], the repeated loading
test CN of the test series C1 is chosen. The dimensions of the notched prismatic specimens are
76  19  305 mm. The material parameters are summarized in Table I. In this test five loading/
unloading cycles between zero stress and the stress envelope curve for monotonic loading were
performed. A single quadrilateral finite element with linear shape functions and reduced numerical
integration was employed for the numerical simulations. Its characteristic length is referred to the
gauge length of 82.6 mm for measuring the axial displacements due to crack opening in the
experiment, i.e., lchar = 82.6 mm. The repeated loading was represented by prescribing varying
vertical displacements at the top edge of the finite element.
Figure 3 shows the experimental and the numerical results, plotting the axial stress s11 in terms of the
axial strain e11. Since the Extended Leon Model (ELM) is unable to describe the reduction of stiffness
degradation in the post-peak region, only monotonic loading is considered for the Extended Leon
model. Cyclic loading is considered only for the damage-plasticity model (DPM) by Grassl & Jirásek.
The uniaxial tensile strength is predicted very well by both models. However, for the given ratio
ftu/fcu the Extended Leon model, modified by Pivonka, erroneously would predict a plateau at peak
stress. This shortcoming is a consequence of the hardening ductility parameter (9), which was
determined from the tensile tests by Hurlbut [51] for a ratio of ftu/fcu = 1/8. This shortcoming was
eliminated by replacing the uniaxial compressive strength by fcu = 8ftu. This modification would result
in erroneous predictions of the compressive behavior, which, however, is not relevant in this test.
The damage-plasticity model (DPM) by Grassl & Jirásek predicts the response of the specimen very well.
For this test the version with the enhanced evolution law for the damage variable (37) (DPM-enh.) gives a
similar approximation of the observed softening envelope.

4.1.2. Triaxial compression tests. From the extensive test series on cylindrical concrete specimens,
documented in [53], triaxial compression tests with different levels of confinement are chosen. The
diameter and height of the specimens are 54 mm and 108 mm, respectively. The material parameters
are summarized in Table II, where the specific mode I fracture energy is estimated according to [57]
from the maximum aggregate size of dmax = 10 mm. A single axisymmetric finite element with linear
shape functions and reduced integration was employed for the numerical simulations. Its
characteristic length is chosen as lchar = 108 mm. The loading was represented by prescribing the
confining stress at the lateral surface, which was kept constant during the test, and by prescribing
increasing vertical displacements at the top surface of the finite element.
Figure 4 shows a comparison of experimental and computed results. The peak stresses at different
levels of confinement are predicted well by both models. However, the Extended Leon model
(ELM) underestimates both, the axial and lateral strain, in particular for higher levels of confinement.
In contrast to the Extended Leon model, the damage-plasticity model (DPM) by Grassl & Jirásek
yields good agreement of measured and predicted axial and lateral strains for different levels of
confinement. A further slight improvement is achieved by the modified version of this model with
the evolution law for the damage variable (37) for the softening response, as the artificial sharp
bends at the transition from hardening to softening are eliminated.
Because of its superior performance, in the subsequent subsections only the damage plasticity model with
the modified evolution law for the damage variable (37) will be validated by structural tests and will finally
be applied for the large-scale numerical simulation of the structural behavior of an arch dam model.

Table I. Material parameters for the repeated uniaxial loading tests [52].
parameter (mean) value

Ec 32012.00 N/mm2
nc 0.18
fcu 43.88 N/mm2
ftu 3.53 N/mm2
GIf 0.0564 Nmm/mm2

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

4.0
experiment
ELM
3.5 DPM
DPM-enh.
3.0

2.5

2.0

1.5

1.0

0.5

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Figure 3. Experimental and numerical results of a uniaxial tension test with repeated loading according to [52].

Table II. Material parameters for the triaxial compression tests [53].
parameter (mean) value
Ec 30000.00 N/mm2
nc 0.15
fcu 47.40 N/mm2
ftu 4.74 N/mm2
GIf 0.0780 Nmm/mm2

4.2. Validation by structural tests


4.2.1. PCT-3D test. In the PCT-3D tests, conducted at the University of Innsbruck, prismatic
concrete specimens with dimensions of 180  180  600 mm were subjected to combined bending
and torsional loading. The dimensions of the specimen together with the boundary conditions and
the loading are shown in Figure 5.
The test layout is described in detail in [58] and [59]. Hence, here only a brief overview of the test
arrangement is provided. Figure 6 shows the FE-mesh of the test specimen together with the support
rollers and the load application roller, the latter located at the bottom face of the specimen. Both, the
concrete specimen and the steel components are discretized by 69372 3D isoparametric 20-node
elements with reduced numerical integration. The vertical displacements of the support rollers are
constrained at the outermost seven nodes located at the roller axes. At the end of one support roller
also constraints in horizontal direction are prescribed to avoid rigid body motions.
A notch of isosceles triangular shape of 60 mm length in both vertical and horizontal direction with a
notch width of 5 mm was provided at midspan of the specimen at the tensile faces for triggering the
location of crack initiation. At an offset of 30 mm from the vertical face of the specimen a vertical point
load was applied to the load application roller, which resulted in combined bending and torsional loading.
The employed material parameters for the concrete specimen and the support rollers and the load
application roller, both made of steel, are summarized in Table III. According to [59], the uniaxial
tensile strength of ftu = 4.83 N/mm2, determined on prismatic samples of the same concrete as the
PCT-3D specimens, overestimates the apparent tensile strength of the latter. As shown in [59] this
deficiency is confirmed by a comparison of the measured uniaxial tensile strength with estimates
computed from the flexural tensile strength, the splitting tensile strength and the compressive
strength. Hence, for the numerical model the uniaxial tensile strength and the specific mode I
fracture energy of concrete are estimated according to [57]. A similar assumption for the tensile
strength was made in the numerical simulation conducted by Gasser [60]. The latter was performed
on the basis of the X-FEM for modeling crack propagation, assuming, however, linear-elastic
material behavior of concrete in compression.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

-25

-50

-75

-100 ELM
DPM
DPM-e nh.
-125

-150

-175

-200

-225
-70 -60 -50 -40 -30 -20 -10 0 10 20

Figure 4. Experimental and numerical results of triaxial compression tests according to [53].

Figure 5. Dimensions, boundary conditions and loading of the test specimen.

In the numerical simulation after application of the dead load the point load is applied by prescribing
a vertical displacement at a single node of the load application roller, representing the point of load
application.
The scatter of the experimental results as well as the numerical results for the PCT-3D test are shown
in Figure 7, plotting the point load P in terms of the crack mouth opening displacement (CMOD). The
latter is measured across the notch at the front edge of the top face of the specimen.
In the numerical simulation the onset of cracking is predicted at the center of the base of the
triangular notch. With increasing vertical displacement of the point of load application, the crack

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

support roller

load application roller


support roller

Figure 6. Finite element mesh of the PCT-3D test.

Table III. Material parameters for the numerical simulation of the PCT-3D test.
concrete steel
param. mean value (standard deviation) param. mean value
rc 2449 () kg/m3 rs 7850 kg/m3
Ec 37292.60 ( 2055.45) N/mm2 Es 210000.00 N/mm2
nc 0.1927 ( 0.014) ns 0.30
fcu 40.12 ( 0.83) N/mm2
ftu 3.05 () N/mm2
GIf 0.0661 () Nmm/mm2

40
experimental scatter
35 PCT-3D-Gasser
PCT-3D-DPM-enh.
30

25

20

15

10

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40

Figure 7. Measured and computed load-crack mouth opening displacement curves for the PCT-3D test.

starts propagating along the base of the triangular notch and subsequently along the tensile face (top
surface) towards the rear side of the finite element model. The ultimate load is reached at a crack
mouth opening displacement of CMOD = 0.022 mm. At peak load the predicted crack extends from
the notch to the rear side of the model. For a detailed comparison of the computed crack faces,
displayed in Figure 8, with the crack faces, determined from the tests on four specimens, it is
referred to [59]. As can be seen in Figure 8, in contrast to the vertical tensile face, at the top face
the predicted crack propagates in one row of elements. However, in the experiments and the

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

Figure 8. Predicted crack pattern for the front face (top) and the rear face (bottom) of the PCT-3D test spec-
imen at failure (SDV9  o, deformations 50-fold).

numerical simulation by Gasser a curved crack was observed at the top face. Hence, the present model
shows a mesh induced bias as a consequence of the employed smeared crack approach. Nevertheless,
the predicted peak load of Pmax = 33 kN, which overestimates the mean value of the experiments by
about 18 % is in good agreement with both, the experimental results and the numerical result by
Gasser, and both the pre-peak and the post-peak part of the load-CMOD curve is predicted very
well by the present model.

4.3. Squat bridge column subjected to combined axial and cyclic shear loading
From the test series of squat bridge columns, subjected to combined axial and cyclic shear loading,
documented in [61] and [62], the test RCT-R1 is simulated numerically. The column was built on a
scale of 1 : 2.5, a vertical section and the cross section of the test specimen are shown in Figure 9. The
longitudinal reinforcement consisted of 22 Grade 40 bars with a diameter of dlong = 19.05 mm,
extending from the foundation through the column into the load application head. The transverse
reinforcement consisted of Grade 50 stirrups with a diameter dhoop = 6.35 mm at a spacing of
shoop = 127 mm. The load application head and the foundation were tightened by additional reinforcement.
Taking advantage of symmetry, only one half of the test specimen is modeled. The column, the
foundation and the head of the column are discretized by 7816 3D isoparametric quadratic finite
elements with reduced integration, the longitudinal and transverse reinforcement bars are modeled

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

by 1336 quadratic beam elements. The latter are placed along the edges of continuum elements
representing the concrete, assuming perfect bond between reinforcement and concrete.
The material parameters of the concrete and the reinforcement are provided in Table IV. Linear-
elastic material behavior is assumed for the concrete of the load application head and the
foundation. The uniaxial tensile strength and the mode I fracture energy are estimated according to
[57]. To this end, the maximum aggregate size is assumed as dmax = 20 mm. Linear elastic - perfectly
plastic material behavior is assumed for the reinforcement bars.
At the bottom of the foundation the displacements are fixed in both vertical and horizontal direction.
According to the test setup, the load application head does not exhibit any rotations. Hence, kinematic
coupling constraints are used to prohibit rigid body rotations of the top face of the load application head.
To this end, at the top surface the translational degrees of freedom in vertical direction are constrained to
the reference point A, which is shown in Figure 9. Hence, the load application head is only allowed to
translate in horizontal and vertical direction parallel to the plane of symmetry.

Figure 9. Test specimen RCT-R1 of a squat bridge column.

Table IV. Material parameters for the test specimen RCT-R1 of a squat bridge column.
concrete steel

param. mean value param. mean value


rc 2400 kg/m3 rs 7850 kg/m3
Ec 24132 N/mm2 Es 199948 N/mm2
nc 0.20 ns 0.30
fcu 37.92 N/mm2 fsy , long 324.05 N/mm2
ftu 2.91 N/mm2 fsy, hoop 358.53 N/mm2
GIf 0.0763 Nmm/mm2

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

After application of the dead load, an axial load PV = 507.10 kN is applied, which is kept constant
during the test. Subsequently, monotonic and reversed cyclic shear loading is simulated by applying
horizontal displacements to the load application head. The maximum/minimum horizontal
displacements of the considered first three reversed load cycles are given as  2.2 mm,  3.3 mm
and  4.4 mm, respectively. After the third load cycle the prescribed horizontal displacements are
increased monotonically up to failure.
Figure 10 shows the measured and computed load-displacement diagrams, plotting the horizonal
force PH in terms of the applied horizontal displacement uH at the load application head. In addition,
numerical results by Kang [63], referred to as "RCT-R1-Kang" in Figure 10, which are based on an
elastic-plastic concrete model, are shown. In the latter numerical model the concrete cover of the
column was neglected and only monotonic loading was considered.
In the numerical simulation conducted by Kang [63] the stiffness is overestimated at the beginning
and the horizontal displacements at peak load are underestimated. At least partially, this may be the
consequence of the assumed value for the specific mode I fracture energy by Kang [63], which is
about 20-times larger than the estimated fracture energy according to [57].
After the first load cycle horizontal cracks are predicted in the column in the vicinity of the load
application head and the foundation, resulting from tensile stresses due to bending of the column.
After the second load cycle a second row of horizontal cracks and the initiation of a third row of
horizontal cracks are predicted at the top and bottom of the column. After the third load cycle the
concrete interfaces between the column and the load application head and between the column and
the foundation are almost completely cracked through the thickness of the column (Figure 11a).
The numerical simulation of three load cycles shows that the numerical response is robust for
reversed loading. However, if a larger number of load cycles is considered in the numerical
simulation, then the predicted stiffness and ultimate load will be underestimated. This is due to the
fact that the employed modified version of the damage-plasticity model by Grassl and Jirásek is not
designed for reversed loading. However, monotonic loading up to failure gives good agreement of
the computed response with the measured envelope of the cyclic load-displacement curves.
With the subsequently applied monotonic loading new horizontal cracks are predicted at the tensile
faces of the column. In the numerical simulation first yielding of the longitudinal reinforcement is
predicted at PH = 338 kN during monotonic loading after the third load cycle. The horizontal force
at first yielding is underestimated by about 16%. At an applied horizontal displacement of
approximately uH = 5 mm the cracks continue to propagate inclined to the initially horizontal
direction. Subsequently, the primarily horizontal cracks merge in a skew crack band. In the
numerical simulation the column fails at an applied horizontal displacement of uH = 31 mm with a
predicted skew crack band, inclined by approximately 25∘ with respect to the column axis
(Figure 11b). According to [62], in the test the major diagonal crack developed at an inclination of
32∘ with respect to the column axis.

600

400

200

-200

-400 experiment
RCT-R1-Kang
RCT-R1-DPM-enh.
-600
-40 -30 -20 -10 0 10 20 30 40

Figure 10. Measured and computed cyclic response of the test specimen RCT-R1 of a squat bridge column.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

5. NUMERICAL SIMULATION OF A MODEL TEST OF A CONCRETE ARCH DAM

In the course of the design and construction of Zillergründl arch dam, which is one of the highest
concrete arch dams in Austria, model tests were carried out by the department for tests and
measurements of VERBUND - Austrian Hydro Power. The reason for choosing the model dam for the
numerical simulation instead of the full-scale concrete arch dam is given by the fact, that only for
the model test test-data up to failure of the dam is available for a comparison with computed results.

5.1. Test setup


The arch dam model was built on a scale of 1:200. It was made of different mixtures of gypsum,
siliceous earth, dinatriumphosphat and water in order to account for the different material properties
of the dam body and the rock foundation (Figure 12).
After pouring of the rock foundation, the model of the arch dam was casted in one part. Thus, in the
model the arch dam is represented by a monolithic part disregarding joints and inspection chambers.
The joint between the base of the arch dam and the rock foundation was grouted with a synthetic resin.
From the scaling factor for the geometry MG = 200 follows that the height of the arch dam of 186 m,
the thickness at the crest of 6.7 m and the maximum thickness at the base of 42 m are reduced in the
model to 930 mm, 34 mm and 210 mm, respectively. In the model the height of the rock foundation
below the center of the arch dam was 700 mm and the rock foundation on each side of the arch dam
model had a maximum height of 1700 mm. The vertical abutments of the rock foundation were
supported by a suspension construction which was made of reinforced concrete.
The scaling factor for the stresses MS was defined as the ratio of the shear strength of the dam
dam ¼ 3:70 N/mm over the shear strength of the gypsum mixture tdam ¼ 0:54 N/mm
2 2
concrete tD M

yielding MS = 6.85, where superscripts D and M are indicating the real dam and the model, respectively.
The model was loaded by the scaled dead load of the arch dam and the scaled hydrostatic water
pressure acting on the upstream face of the arch dam. The scaled specific weight of the dam
dam ¼ gdam MG =MS ¼ 702 kN/m with gdam ¼ 24 kN/m as the specific
3 3
concrete was computed as gM D D

weight of the dam concrete.


The scaled dead load was applied by 42 steel plates, molded in the arch dam at six horizontal levels.
Concentrated forces were applied to those plates by prestressed steel strings with a diameter of about
4 mm, which were anchored at a steel grid below the base of the model. In order to prevent any

(a) after three cycles ( =0 ) (b) at failure ( = 31 )

Figure 11. Predicted crack propagation for the squat bridge column: (a) after three load cycles, (b) at failure
(SDV9  o, deformations 10-fold).

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

interaction of the steel strings with the arch dam during loading, they were surrounded by a synthetic
cladding tube.
The scaled hydrostatic water pressure was applied on the upstream face of the arch dam model by 77
hydraulic jacks and steel plungers (see Figure 13), which were arranged in nine horizontal levels. The
W ¼ gW MG =MS ¼ 286kN/m with gW ¼ 9:81kN/m
3 3
scaled specific weight of water was determined as gM D D

as the specific weight of water. Hence, the scaled hydrostatic design water pressure at the base of the
W;d ¼ gW hdam ¼ 266 kN/m . The magnitude of the scaled
2
arch dam at full reservoir is determined as pM M M

hydrostatic water pressure was increased during the ultimate load test until the model failed. During the
test, the radial displacements in horizontal planes were measured by displacement transducers, located
at the central cross section at the downstream face of the arch dam 750 mm and 50 mm below the crest,
respectively (Figure 13). The two measurement points are denoted as MP 1700 and MP 1840, the
numbers indicating altitudes above sea level of 1700 m and 1840 m, respectively, for the real dam.
ma

orographical orographical
in
fau

right side rock foundation 1 left side


lt

1 = 1.36

ult
ary fa
upstream side second

m 2 = 1.64
da
arch

rock foundation 2
2 = 1.64
downstream side
m
ain
fa
ul
t

Figure 12. Plan view of the model test layout (taken from [64]).

MP 1840

MP 1700

Figure 13. Test setup of the arch dam model with displacement transducers at MP 1700 and MP 1840 at the
downstream face and hydraulic jacks and plungers at the upstream face (taken from [64]).

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

5.2. Numerical model of the arch dam model test


The finite element mesh comprises the whole test model, consisting of the arch dam and the rock
foundation, however, disregarding the faults shown in Figure 12. The model of the arch dam is
discretized by 267000 3D linear finite elements, increasing the number of elements across the dam
thickness from 6 at the crest to 14 at the center of the base.
The material parameters are taken from the test report [64]. They are adjusted and supplemented
according to the requirements of the employed modified damage-plasticity model. In particular, the
compressive strength, determined on cubical samples is converted to the cylindrical strength and the
measured flexural tensile strength is converted to the uniaxial tensile strength. Since the specific mode I
fracture energy of the gypsum mixture of the arch dam was not determined within the framework of the
model test, it is assumed as GIf ¼ 0:025 N/mm. The material parameters, employed for the numerical
simulation, are summarized in Table V. Since for the rock foundation only the Young’s modulus and the
Poisson’s ratio are available, linear-elastic material behavior is assumed. For the steel strings and the
plungers also linear-elastic material behavior is assumed, because the yield strength is not attained in any
load step.
In the numerical simulation the scaled dead load and the scaled hydrostatic water pressure are applied in
a discrete manner according to the model test by prestressing the steel strings and by applying loads,
equivalent to the scaled water pressure, to the steel plungers. The 42 steel strings are represented by
truss elements. The steel plates are taken into account by coupling the upper node of each truss element
to the nodes of the horizontal faces of the adjacent finite elements, covering the area of the respective
steel plate. The lower node of each truss element is prestressed by a concentrated force, representing
the respective fraction of the scaled dead load. The 77 plungers, located at the upstream face of the arch
dam, are discretized by linear shell elements, sharing the same nodes with the continuum elements,
representing the arch dam. The plungers are loaded by forces, which are equivalent to the respective
fraction of the scaled hydrostatic water pressure.
Displacement boundary conditions are applied to the nodes of both the bottom face and the vertical
faces of the rock foundation.
In the first load step the dead load of the gypsum mixture of the arch dam is applied by distributed
body forces, whereas the dead load of the rock foundation is not considered in the analysis. In the
second load step the scaled dead load of the arch dam is considered by applying prestressing forces
to the vertical steel strings. In the third load step the scaled hydrostatic water pressure is applied to
the plungers on the upstream face of the arch dam model. Subsequently, the magnitude of the scaled
water pressure is increased up to failure.
Figure 14 shows a comparison of the computed load-displacement curves for the two measurement
points MP 1700 (solid curves) and MP 1840 (dashed curves) with the results of the model test, plotting the

Table V. Material parameters for the arch dam model.


gypsum (rock foundation 1) gypsum (rock foundation 2)

param. mean value param. mean value


rrock, 1 1529 kg/m3 rrock, 2 1529 kg/m3
Erock, 1 3658.4 N/mm2 Erock, 2 4411.6 N/mm2
nrock, 1 0.18 nrock, 2 0.18
gypsum (arch dam) steel (strings and plungers)

param. mean value param. mean value


3
rdam 1529 kg/m rs 7850 kg/m3
Edam 2690 N/mm2 Es 210000 N/mm2
ndam 0.18 ns 0.30
fcu 2.77 N/mm2
ftu 0.94 N/mm2
GIf 0.025 N/mm

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

amplification factor lM M
W for the scaled design water pressure in terms of the radial displacements ur in those
horizontal planes, which are determined by the two measurement points (Figure 13).
During application of the scaled dead load of the arch dam first cracks are predicted in the vicinity of
most of the 42 steel plates molded in the arch dam due to the high forces in the steel strings which are
transferred to the arch dam. However, these cracks do not decrease the stiffness of the model
significantly because they are restricted to small regions. At a scaled water pressure of 2:50pM W;d a
horizontal tensile crack is predicted at the base of the upstream face of the arch dam model. At a
scaled water pressure of about 3:00pM W;d damage, caused by combined shear stresses and
compressive stresses, is predicted in the center region of the upper half at the downstream face of
the arch dam. Subsequently, the region of damage is increasing upon further increase of the scaled
water pressure. First damage of the abutments at both sides of the arch dam is predicted at 3:75pM W;d ,
which is caused by compressive stresses. The damaged region is more pronounced at the
orographical left side because of the slightly asymmetric shape of the arch dam with respect to its
central cross section (c.f. Figure 12).
The ultimate load is predicted at 4:52pMW;d, exceeding the measured ultimate load in the model test of
M
4:40pW;d by only 3%. The displacements, computed for MP 1700 and MP 1800, agree well with the

4
0
170
MP

3
0

2
184
MP

experiment (MP 1700)


1 experiment (MP 1840)
DPM-enh. (MP 1700)
DPM-enh. (MP 1840)
0
0 1 2 3 4 5 6 7 8

Figure 14. Comparison of computed and measured load-displacement curves for MP 1700 and MP 1840 of
the arch dam model.

Figure 15. Failure of the model at the ultimate load of 4.40 PM


W,d (taken from [64]).

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

displacements of the model test up to 4:10pM M


W;d and 2:50pW;d , respectively. Upon a further increase of
the scaled hydrostatic water pressure for both measurement points the predicted displacements are
larger than the measured ones. Probably, this is the consequence of missing data regarding the
softening behavior of the gypsum mixture. Since only standard material parameters for the latter
were available, but no load-displacement curves from the tests, similar softening behavior was
assumed as for concrete, although a more brittle material behavior for the gypsum mixture could be
expected.
In the model test the arch dam failed due to rupture of the arch dam body in the vicinity of the crest,
approximately 300 mm from the center of the arch dam at the orographical left side (see Figure 15).
About 150 mm below the crest a crack was running nearly horizontally along the whole arch dam.
In addition, large cracks can be seen in Figure 15 at the orographical left side of the arch dam in
horizontal and vertical direction.
Figure 16 shows the damaged regions at both the upstream and the downstream face of the arch dam
model at failure. Comparing Figure 15 and Figure 16 (bottom) shows that the regions of damage are
predicted in a satisfactory manner. Figure 16 confirms that the discrete application of the scaled
dead load and the scaled water pressure effects the stress state and the distribution of damage within
the arch dam body. In particular, the location of the uppermost row of steel plates can be inferred
from the regions of increased damage of the arch dam, whereas the steel plates, located at lower
horizontal levels have a smaller influence on the crack pattern at the upstream and downstream face
of the arch dam because of the increasing thickness of the arch dam from the crest to the base.

Figure 16. Predicted damage at failure for the upstream face (top) and the downstream face (bottom) (state
variable SDV9  o).

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

6. CONCLUSIONS

The evaluation of selected 3D plasticity and damage-plasticity models for describing the 3D material
behavior of concrete, including the Extended Leon model, developed by Etse and modified by
Pivonka [16] and the damage-plasticity model proposed by Grassl and Jirásek [35], which was
modified regarding the evolution law for the damage variable, clearly revealed the superior
performance of the damage-plasticity model.
Hence, the latter model was implemented in a finite element program placing special emphasis on
the robustness and efficiency of the employed stress update algorithm. To this end, within the
framework of the return mapping algorithms an adaptive substepping procedure with error control
was used.
The capabilities of the developed numerical model were demonstrated by 3D numerical simulations
of lab tests at the integration point level, by the numerical simulation of benchmark tests with
combined bending and torsional loading and combined compression and shear loading and by a
large-scale 3D finite element analysis of a model test of a concrete arch dam. Apart from some mesh
bias effects for the predicted crack paths, which are a drawback of the employed smeared crack
concept, the observed material behavior is represented in a very satisfactory manner by the employed
damage-plasticity model and the stress update has proven as fully robust. The latter feature is
indispensable especially for large-scale nonlinear FE-analyses as demonstrated in the present paper by
the ultimate load analysis of a concrete arch dam model with a FE-model consisting of 267000
elements with altogether about 914000 degrees of freedom.

ACKNOWLEDGEMENTS
This work was supported by the Austrian Ministry of Science BMWF as part of the university infrastructure
program of the Research Plattform Scientific Computing at the University of Innsbruck.

REFERENCES
1. Jirásek M, Bažant ZP. Inelastic Analysis of Structures. Wiley, 2002.
2. Jirásek M. Modeling of Localized Inelastic Deformation. Lecture Notes, September 2007.
3. Babu RR, Benipal GS, Singh AK. Constitutive modelling of concrete: an overview. Asian Journal of Civil
Engineering (Building and Housing) 2005; 6:211–246.
4. Pivonka P, Ozbolt J, Lackner R, Mang HA. Comparative studies of 3D-constitutive models for concrete: application
to mixed-mode fracture. International Journal of Numerical Methods in Engineering 2004; 60:549–570.
5. Willam K, Warnke EP. Constitutive model for the triaxial behaviour of concrete. Seminar on Concrete Structures
Subjected to Triaxial Stresses. Volume 19. Bergamo, Italy, 1975, 1–30.
6. Leon A. Über die Scherfestigkeit des Betons. Beton und Eisen 1935; 34:130–136.
7. Pramono E, Willam K. Implicit integration of composite yield surfaces with corners. Engineering Computations
1989; 6:186–197.
8. Etse G. Theoretische und numerische Untersuchung zum diffusen und lokalisierten Versagen in Beton. Ph.D. thesis,
Universität Fridericiana zu Karlsruhe (TH), Germany, 1992.
9. Etse G, Willam K. Fracture energy formulation for inelastic behavior of plain concrete. Journal of Engineering
Mechanics 1994; 120:1983–2011.
10. Menétrey P. Numerical analysis of punching failure in reinforced concrete structures. Ph.D. thesis. Ecole Polytechnique
Federale De Lausanne, Switzerland, 1994.
11. Kang H. Triaxial constitutive model for plain and reinforced concrete behavior. Ph.D. thesis. CEAE Dept., University of
Colorado, Boulder, Colorado, USA, 1997.
12. Kang HD, Willam K. Localization characteristics of triaxial concrete model. Journal of Engineering Mechanics
1999; 125:941–950.
13. Lade PV. Single hardening constitutive model for soil, rock and concrete. International Journal of Solids and
Structures 1995; 32:1963–1978.
14. Bićanić N, Pearce CJ. Computational aspects of a softening plasticity model for plain concrete. Mechanics of cohesive-
frictional materials 1996; 1:75–94.
15. Haufe A. Dreidimensionale Simulation bewehrter Flächentragwerke aus Beton mit der Plastizitätstheorie. Ph.D.
thesis. Universität Stuttgart, Germany, 2001.
16. Pivonka P. Constitutive modeling of triaxially loaded concrete considering large compressive stresses: application to
pull-out tests of anchor bolts. Ph.D. thesis. Technische Universität Wien, Austria, 2001.
17. Grassl P, Lundgren K, Gylltoft K. Concrete in compression: a plasticity theory with a novel hardening law.
International Journal of Solids and Structures 2002; 39:5205–5223.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
REVIEW AND ENHANCEMENT OF 3D CONCRETE MODELS

18. Papanikolaou VK, Kappos AJ. Confinement sensitive plasticity constitutive model for concrete in triaxial
compression. International Journal of Solids and Structures 2007; 44:7021–7048.
19. Meschke G. Consideration of aging of shotcrete in the context of a 3-D viscoplastic material model. International
Journal for Numerical Methods in Engineering 1996; 39:3123–3143.
20. Bockhold J. 3D material model for nonlinear basic creep of concrete. Computers and Concrete 2007; 4:101–117.
21. Mazars J, Pijaudier-Cabot G. Continuum Damage Theory - Application to Concrete. Journal of Engineering
Mechanics 1989; 115:345–365.
22. Tao X, Phillips DV. A simplified isotropic damage model for concrete under bi-axial stress states. Cement and
Concrete Composites 2005; 27:716–726.
23. Voyiadjis ZG, Kattan PI. A Comparative Study of Damage Variables in Continuum Damage Mechanics.
International Journal of Damage Mechanics 2009; 18:315–340.
24. Huber F. Nichtlineare dreidimensionale Modellierung von Beton- und Stahlbetontragwerken. Ph.D. thesis.
Universität Stuttgart, Germany, 2006.
25. Lemaitre J, Chaboche JL. Mechanics of Solid Materials. Cambridge University Press, 1990.
26. Papa E, Taliercio A. Anisotropic damage model for the multiaxial static and fatigue behaviour of plain concrete.
Engineering Fracture Mechanics 1996; 55:163–179.
27. Desmorat R, Gatuingt F, Ragueneau F. Nonlocal anisotropic damage model and related computational aspects for
quasi-brittle materials. Engineering Fracture Mechanics 2007; 74:1539–1560.
28. Jirásek M, Grassl P. Evaluation of directional mesh bias in concrete fracture simulations using continuum damage
models. Engineering Fracture Mechanics 2008; 75:1921–1943.
29. Lee J, Fenves GL. Plastic-Damage Model for Cyclic Loading of Concrete Structures. Journal of Engineering
Mechanics 1998; 124:892–900.
30. Jason L, Huerta A, Pijaudier-Cabot G, Ghavamian S. An elastic plastic damage formulation for concrete: Application
to elementary tests and comparison with an isotropic damage model. Computer Methods in Applied Mechanics and
Engineering 2006; 195:7077–7092.
31. Nguyen GD, Houlsby GT. A coupled damage-plasticity model for concrete based on thermodynamic principles: Part
I: model formulation and parameter identification. International Journal of Numerical and Analytical Methods in
Geomechanics 2008; 32:353–389.
32. Saritas A, Filippou FC. Numerical integration of a class of 3 d plastic-damage concrete models and condensation of
3 d stress-strain relations for use in beam finite elements. Engineering Structures 2009; 31:2327–2336.
33. Lee J, Fenves GL. A plastic-damage concrete model for earthquake analysis of dams. Earthquake Engineering and
Structural Dynamics 1998; 27:937–956.
34. Schütt J. Ein inelastisches 3D-Versagensmodell für Beton und seine Finite-Element-Implementierung. Ph.D. thesis.
Universität Karlsruhe (TH), Germany, 2005.
35. Grassl P, Jirásek M. Damage-plastic model for concrete failure. International Journal of Solids and Structures 2006;
43:7166–7196.
36. Grassl P, Jirásek M. Plastic model with non-local damage applied to concrete. International Journal for Numerical
and Analytical Methods in Geomechanics 2006; 30:71–90.
37. Mohamad-Hussein A, Shao JF. Modelling of elastoplastic behaviour with non-local damage in concrete under
compression. Computers and Structures 2007; 85:1757–1768.
38. Cicekli U, Voyiadjis GZ, Abu Al-Rub RK. A plasticity and anisotropic damage model for plain concrete.
International Journal of Plasticity 2007; 23:1874–1900.
39. Voyiadjis ZG, Taqieddin ZN, Kattan PI. Theoretical Formulation of a Coupled Elastic Plastic Anisotropic Damage
Model for Concrete using the Strain Energy Equivalence Concept. International Journal of Damage Mechanics
2009; 18:603–638.
40. Abu Al-Rub RK, Voyiadjis GZ. Gradient enhanced Coupled Plasticity-anisotropic Damage Model for
Concrete Fracture: Computational Aspects and Applications. International Journal of Damage Mechanics 2009;
18:115–154.
41. Bažant ZP, Oh BH. Microplane Model for Progressive Fracture of Concrete and Rock. Journal of Engineering
Mechanics 1985; 111:559–582.
42. Bažant ZP, Prat PC. Microplane Model for Brittle-Plastic Material: I. Theory. Journal of Engineering Mechanics
1988; 114:1672–1688.
43. Ozbolt J, Bažant ZP. Microplane Model for Cyclic Triaxial Behavior of Concrete. Journal of Engineering Mechanics
1992; 118:1365–1386.
44. Bažant ZP, Xiang Y, Adley MD, Prat PC, Akers SA. Microplane Model for Concrete: II: Data Delocalization and
Verification. Journal of Engineering Mechanics 1996; 122:255–262.
45. Bažant ZP, Xiang Y, Prat PC. Microplane Model for Concrete. I: Stress-Strain Boundaries and Finite Strain. Journal
of Engineering Mechanics 1996; 122:245–254.
46. Leukart M. Kombinierte anisotrope Schädigung und Plastizität bei kohäsiven Reibungsmaterialien. Ph.D. thesis.
Universität Stuttgart, Germany, 2005.
47. Simo JC, Hughes HJR. Computational Plasticity. Springer, 1998.
48. Pérez-Foguet A, Rodriguez-Ferran A, Huerta A. Consistent tangent matrices for substepping schemes. Computer
Methods in Applied Mechanics and Engineering 2001; 190:4627–4647.
49. Valentini B. A three-dimensional constitutive model for concrete and its application to large-scale finite element
analyses. Ph.D. thesis. Universität Innsbruck, Austria, 2011.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag
B. VALENTINI AND G. HOFSTETTER

50. Ortiz M, Martin JB. Symmetry-preserving return mapping algorithms and incrementally extremal paths: A
unification of concepts. International Journal for Numerical Methods in Engineering 1989; 28:1839–1853.
51. Hurlbut B. Experimental and computational investigation of strain softening in concrete. Master thesis. University of
Colorado, Boulder, USA, 1985.
52. Gopalaratnam VS, Shah SP. Softening response of plain concrete in direct tension. Journal of the American Concrete
Institute 1985; 82:310–323.
53. Imran I, Pantazopoulou SJ. Experimental Study of Plain Concrete under Triaxial Stress. ACI Material Journal 1996;
93:589–601.
54. van Mier JGM. Strain Softening of Concrete under Multiaxial Loading Conditions. Ph.D. thesis. TH Eindhoven, The
Netherlands, 1984.
55. van Geel H. Behavior of concrete in plane strain compression. Research Report TH Eindhoven, Eindhoven, The
Netherlands, 1985.
56. Burlion N. Compaction des Betons: Elements de Modelisation et Caracterisation Experimentale. Ph.D. thesis. Laboratoire
de Mechanique et Technologie ENS Cachan / CNRS / Universite Pierre et Marie Curie, Cachan, France, 1997.
57. CEB-FIP model code 1990. Final draft. Bulletin information 203 du Committee Euro-International du Beton, 1991.
58. Feist C. A Numerical Model for Cracking of Plain Concrete Based on the Strong Discontinuity Approach. Ph.D.
thesis. University of Innsbruck, Austria, 2004.
59. Feist C, Hofstetter G. Validation of 3D crack propagation in plain concrete. Part I: Experimental investigation - the
PCT3D test. Computers and Concrete 2007; 4:49–66.
60. Gasser TC. Validation of 3D crack propagation in plain concrete. Part II: Computational modeling and predictions of
the PCT3D test. Computers and Concrete 2007; 4:67–82.
61. Xiao Y, Priestley M, Seible F. Steel Jacket Retrofit for Enhancing Shear Strength of Rectangular Short Reinforced
Concrete Columns. Structural Systems Research Project. University of California at San Diego, June 1993.
62. Priestley N, Seible F, Xiao Y, Verma R. Steel Jacket Retrofitting of Reinforced Concrete Bridge Columns for
Enhanced Shear Strength - Part 1: Theoretical Considerations and Test Design. ACI Structural Journal 1994;
91:394–405.
63. Kang HD, Willam K, Shing B, Spacone E. Failure analysis of R/C columns using a triaxial concrete model.
Computers and Structures 2000; 77:423–440.
64. Nindl H. Bruchversuch an einem Modell 1:200 der Sperre Zillergründl EZi12 mit Stützkörper im Vergleich mit
Variante EZi11A ohne Stützkörper. Tauernkraftwerke A.G. Prüf- und Messwesen, Research Report, Kaprun,
Austria, 1985.

Copyright © 2011 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. (2011)
DOI: 10.1002/nag

You might also like