You are on page 1of 18

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222633577

Efficient approaches to finite element analysis in earthquake


engineering

Article  in  Computers & Structures · May 2003


DOI: 10.1016/S0045-7949(03)00038-5

CITATIONS READS
43 757

4 authors:

Luc Davenne F. Ragueneau


Université Paris Nanterre Ecole normale supérieure Paris-Saclay
104 PUBLICATIONS   1,442 CITATIONS    123 PUBLICATIONS   1,495 CITATIONS   

SEE PROFILE SEE PROFILE

Jacky Mazars Adnan Ibrahimbegovic


French National Centre for Scientific Research Université de Technologie de Compiègne
203 PUBLICATIONS   4,853 CITATIONS    380 PUBLICATIONS   7,363 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PARAMetric UNCertainty Summer School and Workshop - Budapest, 3-7 July, 2017 View project

Concrete structures under impact and severe loading View project

All content following this page was uploaded by Luc Davenne on 08 April 2019.

The user has requested enhancement of the downloaded file.


Computers and Structures 81 (2003) 1223–1239
www.elsevier.com/locate/compstruc

Efficient approaches to finite element analysis


in earthquake engineering
a,*
L. Davenne , F. Ragueneau a, J. Mazars b, A. Ibrahimbegovic a

a
Laboratoire de M
ecanique et Technologie, Ecole Normale Superieure de Cachan, 61, Avenue du Pr
esident Wilson,
94235 Cachan Cedex, France
b
Laboratoire Sols, Solides, Structures, INPG––BP 53, 38041 Grenoble Cedex 9, France

Abstract
This paper deals with the modeling of reinforced concrete structures subjected to earthquake ground motion. Due to
the complex behavior of both materials and structures, efficient numerical tools are developed herein in order to keep
accuracy and robustness for large scale computations. We focus our attention on the use of simplified multifiber beam
element describing the response of structural components and on macro-element accounting for soil–structure inter-
action.
 2003 Elsevier Science Ltd. All rights reserved.

1. Introduction of linear elastic static analysis. The nonlinear behavior


and energy dissipation, can be accounted for in a trivial
Earthquake engineering design of a civil engineering manner by a force-based approach, where the level of
structure concerns in general several levels of strong seismic loading is computed by analyzing the elastic
motion earthquakes, each one with its specific design response spectra on a single degree of freedom system
requirements. In particular, one should distinguish be- and the so-called R reduction factor method is then used
tween: to introduce the ductility of materials (steel and con-
crete). A more rational concept developed during this
– A damage limitation criterion which allows to keep last decade, named displacement-based procedure [19],
the operational features of the building intact for a turns towards the design based on the critical limit states
seismic loading which has a large occurrence probabil- of the structural elements. In other words, defining the
ity. ultimate behavior of materials, one has to deal with
– A collapse criterion which insures that for the nominal maximum strains or curvatures, thus taking into ac-
design seismic action the structure remains stable count the structure ductility demand in a more physical
without significant local collapse. way. Whereas the force-based design uses the elastic
– A life safe objective requiring that for the strongest properties of the structure, the displacement-based de-
possible earthquake feasible for the site, the global sign characterizes it through the secant stiffness at
collapse is prevented and the life is preserved. maximum displacement. The evaluation of vibration
frequencies during the loading is more appropriate than
In the traditional approach to earthquake engineer- for the purely elastic analysis.
ing design, the computations are carried out on the basis These classical procedures for design (force or dis-
placement formulation) are based on the study of an
equivalent single degree of freedom system or in fact a
*
Corresponding author. Tel.: +33-1-4740-2243; fax: +33-1- simplified substitute structure which is not capable to
4740-2240. account for the load redistribution inside the structures
E-mail address: luc.davenne@lmt.ens-cachan.fr (L. Dav- due to local nonlinearities. This is one of the major
enne). drawbacks preventing a realistic description of the

0045-7949/03/$ - see front matter  2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0045-7949(03)00038-5
1224 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239

global and local behavior of a structure up to failure. Section 4 we present several numerical simulations
One remedy consists in using nonlinear finite element which illustrate the performance of the proposed models
method to perform push-over analysis, i.e. a nonlinear with respect to the experimental results. Concluding
static analysis under monotonic increasing lateral load. remarks are given in Section 5.
It allows to determine the maximum carrying capacity of
the structure in terms of forces, displacements, ductility
(deformation), crack pattern and failure mode. The 2. Multifibers beam elements for nonlinear transient
main assumption of such a computation is that the re- analysis of reinforced concrete structures
sponse is related to an equivalent single degree of free-
dom and thus, controlled by a single mode which In this work we seek to develop the models capable of
justifies that the position and amplitude of horizontal representing in a reliable manner nonlinear behavior of a
lateral loads are typically defined by an elastic analysis. structure damaged by a strong earthquake, with the
An alternative choice to perform the earthquake re- main goal of providing a physically based description of
sistant design is by making use of the nonlinear time- damping (as opposed to a more convenient, but very
history analysis, assuming a physical description of adhoc Rayleigh damping model). This kind of approach
materials and applying transient loadings on the struc- is especially suitable for a novel concept in earthquake
ture in terms of natural or simulated ground motion. resistant design of shear-wall type reinforced concrete
The evolution of eigenmodes concomitant to the stiffness structures; referred nowadays as ÔFrench wallsÕ, where
degradation governed by local yield criterion provides the choice of reinforcement is made so that the damaged
currently the most refined method of analysis for ulti- and cracking spreads as much as possible through the
mate behavior of concrete structure [16,23]. For a simple structure which allows that the main part of the energy
reason of excessive computational costs, such ap- generated by an earthquake be absorbed through in-
proaches to structural dynamics in civil engineering are elastic material behavior. In short we seek to develop the
not commonly used. Several strategies have been de- structural models capable of integrating the advanced
veloped in order to lower the CPU time in transient damage models for concrete (e.g. [17]), in order to de-
analysis by solving the equations of motion by a mode scribe the nonlinear behavior of a reinforced concrete
superposition techniques even in the nonlinear range, structures. With respect to the large spreading of the
either by dynamic substructuring for limited spread of zone with nonlinear behavior we further seek to limit the
nonlinear zones (e.g. see [10,11]) or using the tangent model complexity (and resulting computational costs)
stiffness to modify the basis [14]. by limiting the diversity of possible deformation global
Nonlinear dynamic analysis of complex civil engi- patterns which is achieved by choosing a multifiber
neering structures based on a detailed finite element beam model, with all the fibers restricted to beam ki-
model requires large scale computations and handles nematics and with each fiber employing its own con-
delicate solution techniques. The necessity to perform stitutive model (see Fig. 1). The main advantage of using
parametric studies due to the stochastic characteristic of a multifiber type finite element concerns the simple
the input accelerations imposes simplified numerical uniaxial behavior which allows a very efficient imple-
modeling which will reduce the computation cost. In this mentation of inelastic constitutive. This is no longer
work the latter is achieved by selecting the classical possible for thick beams where shear strains play a
Euler–Bernoulli beam model for representing the global major role [7].
behavior of the structural components of a complex civil
engineering structure. We note that no simplification is 2.1. Cross-section behavior
made other than this choice of the beam model. In
particular, the constitutive bahavior of the chosen model The multifiber beam element developed herein em-
remains sufficiently general to take into account all the ploys the standard Hermite polynomial shape functions
different inelastic phenomena (cracking by damage, ir- to describe the variation of the displacement field along
reversible deformation by plasticity and crack-closing by the beam. The difference with ‘‘classical’’ beam elements
unilateral frictional contact condition). Another effect [21] concerns the cross-section behavior, that is the re-
which is also taken into account concerns the interaction lation between the generalized strains e and the gener-
of structure with the foundation, which provides both a alized stresses s. In the general 3D case the latter
more realistic boundary condition as well as the dam- includes:
ping model due to radiation effect.
The outline of the paper is as follows. In Section 2 we s ¼ ðN Mx My Mz ÞT and e ¼ ð e hx vy vz ÞT
briefly describe the chosen model of multilayer beam ð1Þ
used for representing the structural components. The
development pertaining to the modeling of structure– where N is the normal force, Mx the torque, My and Mz
foundation interaction are described in Section 3. In are the bending moments, e the axial strain, hx the twist,
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1225

Fig. 1. Multifibers beam formulation for reinforced concrete structures.

vy and vz the curvatures. The cross-section behavior is difficulty of developing the finite element implementa-
expressed with the constitutive matrix: tion of the Timoshenko beam model concerns the so-
2 3 called shear locking phenomena (e.g. [26]), or inability of
C11 0 C13 C14
6 the standard finite element approximations to represent
C22 0 0 7
C¼6 4
7 ð2Þ pure bending vanishing shear modes. A number of dif-
C33 C34 5
ferent remedies to shear locking problem has been pro-
sym C44
posed (e.g. see [9]), ranging from selective or reduced
where the coefficients are obtained by integrating over integration, assumed shear strain, enhanced shear strain
the cross-section (y and z axes): or hierarchical displacement interpolations. A very re-
Z Z cent work of Kotronis [12] extends these ideas in order
C11 ¼ E dS; C13 ¼ Ez dS; to construct shear locking remedies for a multifiber
S
Z S
Z Timoshenko beam.
Remark: In order to retain the simplicity of the
C14 ¼  Ey dS; C22 ¼ Gðy 2 þ z2 Þ dS
S S constitutive model for the Timoshenko formulation, the
Z Z
normal and shear stress resultants are uncoupled and
C33 ¼ Ez2 dS C34 ¼  Eyz dS nonlinearities are only introduced in the stress–strain
ZS S
relationship in the direction of the normal to the cross-
2
C44 ¼ Ey dS ð3Þ section, whereas the tangential behavior is assumed to
S
remain elastic. A more complex constitutive relation-
where E and G are YoungÕs and shear moduli which vary ships can be used for each fiber, expressed within the
in y and z. The chosen moduli can be either initial, secant framework of irreversible processes thermodynamics
or tangent, depending upon the iterative algorithm used which allows physical coupling of all stress components.
to solve the global equilibrium equations. The compo- Such a development for 2D Timoshenko multilayer
nent of the constitutive matrix are computed by nu- beam was carried out by Dube [7], where the shear stress
merical integrations in (3), with one Gauss point per as well as the normal stress in each layer was used to
fiber. For the Euler–Bernoulli element, the shear forces compute the damage of the layer with the 2D version of
are computed at the element level through the equilib- the damage model presented hereafter.
rium equations (included in the Hermite polynomial
shape functions). Reinforcement bars are introduced 2.2. Enhanced strain formulation
within special fibers, whose behavior is obtained as a
combination of those of concrete and steel according to: For a multifiber beam model, as indicated in (2), the
rlayer ¼ ð1  aÞrconcrete þ arsteel ð4Þ nonlinear behavior implies coupling between axial force
N and bending moment M. It is thus necessary that the
where a is the relative area of the reinforcement in the axial force and bending moment have the same variation
layer. along the element. The latter is not satisfied for an
When dealing with structures which slenderness ratio Euler–Bernoulli multifiber beam element, where axial
is far from the classical beam such as shear wall, a more strain remains constant along the element whereas the
reliable representation of shear deformations and shear bending strain is linear. This problem may become im-
stresses has to be provided. One possibility in that re- portant in the nonlinear range where a linear curvature
spect is to use the classical Timoshenko beam model, along the beam implies a linear variation of the axial
which can describe the constant shear strain. The main strain due to the shift of the neutral axis.
1226 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239
 x x u2  u1

uðxÞ ¼ 1  u1 þ u2 ) eðxÞ ¼ ¼ cst c c 1  2b
L L L u_ tþDt ¼ Du þ 1  u_ ðtÞ  u_ ðtÞ


b Dt b 2c
x2 x3 x3 x2

vðxÞ ¼ 1  3 2 þ 2 3 v1 þ  2 þ x h1 c
L L L2 L þ Dt 1  uðtÞ
€ ð10Þ


3 2b
3 2
x x x x2
þ  2 3 þ 3 2 v2 þ  h2 ) jðxÞ
L L L2 L where c ¼ 1=2 and b ¼ 1=4 are typically chosen for


optimal result accuracy. The discrete set of equations,
6 12x 6x 4
¼  2 þ 3 v1 þ  h1 obtained by introducing (9) and (10) into (8), is further
L L L2 L


solved by an iterative solution procedure, where at each
12x 6 6x 4
þ  3 þ 2 v2 þ  h2 iteration the problem reduces to
L L L2 L
ð5Þ 1 c
M þ C þ K Dui ¼ Mr€ ug ðt þ tÞ
b Dt2 b Dt
One way to remove this incompatibility of variations for ui ðt þ DtÞ  C u_ i ðt þ DtÞ  f int ðui ðt þ DtÞÞ
 M€
axial strain and curvature is by using enriched descrip-
tion of the axial strain field with:
uiþ1 ðt þ DtÞ ¼ ui ðt þ DtÞ þ Dui ð11Þ
4 8x
eðxÞ ) eðxÞ þ ~eðxÞ; ~eðxÞ ¼ GðxÞa; GðxÞ ¼  2
L L Different iterative strategies can be used to compute
ð6Þ the tangent stiffness matrix K in (11) above; if we seek to
reduce the computational cost one would not update K
where GðxÞ is an enhanced axial strain which can be at each iteration as for NewtonÕs method, but either
derived from the displacement bubble function. The keep it constant as for modified NewtonÕs method or use
variational basis of such an enhancement is provided by a secant stiffness for quasi-Newton method. The details
Hu–Washizu principle (e.g. [26]), which can be presented of computing K and f int pertain to the constitutive
in the same manner as the modified method of incom- model which is specified for each component as ex-
patible modes of Ibrahimbegovic and Wilson [10,11] plained subsequently.
Z Z
0 ¼ ðe ðxÞN ðxÞ þ j ðxÞMðxÞÞ dx ¼ u ðxÞf ðxÞ dx 2.4. Materials constitutive relations for dynamics
L L
Z
0¼ ~e ðxÞN ðxÞ dx ð7Þ Both steel and concrete materials are described
L
within the thermodynamic framework for irreversible
processes (e.g. see [15]).
2.3. Global solution procedure In describing the nonlinear behavior of reinforcement
bars, we choose the classical plasticity model and take
Dynamic analysis for earthquake ground motion for into account the nonlinear kinematic hardening of
this kind of structural model reduces to solving the set of Armstrong and Frederick [2] in order to be able to better
nonlinear equations which can be written as: describe the observed hysteresis loops. The free energy
for this model can be written as:
uðtÞ þ C u_ ðtÞ þ f int ðu; tÞ ¼ Mr€ug ðtÞ
M€ ð8Þ
qw ¼ 12ðe  ep Þ : H : ðe  ep Þ þ 12ba : a ð12Þ
where M and C are, respectively, mass and damping
in which H is the HookeÕs elasticity tensor, ep is the
matrix, €uðtÞ and u_ ðtÞ are nodal accelerations and veloc-
plastic strain and a the hardening internal variable
ities, f int ðu; tÞ is the internal force vector obtained by
which describes hardening. The constitutive equations
discretizing (6) and r€ug ðtÞ is the ground acceleration ef-
for this kind of model can be written as:
fect applied on the structure. By using a time-integration
scheme, the differential equation of motion in (8) is re- oðqwÞ oðqwÞ
r¼ ¼ C : ðe  ep Þ; X ¼ ¼ ba ð13Þ
duced to an algebraic equation. In particular for low oe oa
frequency response in earthquake engineering analysis
where X is the stress-like hardening variable. The latter
one uses an implicit scheme such as Newmark one-step
is used to describe a modified form of the plasticity
scheme (see [18]) which allows to express the velocity
criterion:
and acceleration vectors at time t þ Dt as the functions
of their corresponding values at time t and incremental f ¼ J2 ðr  X Þ þ 34aX : X  ry 6 0 ð14Þ
displacement vector:
where a, b and ry are material parameters. Due to the
1 1 1  2b particular geometric characteristics of steel bars, only a
utþDt ¼
€ Du  u_ ðtÞ  €uðtÞ ð9Þ
b Dt2 b Dt 2c 1D implementation of the model is carried out. A typical
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1227

Fig. 2. Hysteresis loop for the steel modeling.

stress–strain hysteresis loops predicted by this model is


given in Fig. 2. Fig. 3. Unidimensional constitutive relations for concrete.
The constitutive model for concrete for earthquake
engineering ought to take into account some observed 8
phenomena, such as decrease in material stiffness due to < r > 0 ! f ðrÞ ¼ 1
rf < r < 0 ! f ðrÞ ¼ 1  rrf ð18Þ
cracking, stiffness recovery which occurs at crack closure :
and inelastic strains concomitant to damage. r < rf ! f ðrÞ ¼ 0
To simulate this behavior we employ the damage
where rf is the stress where the cracks are completely
model of La Borderie [13], which uses two scalar damage
reclosed. The role of the crack closure function is:
variables, D1 for damage in tension and D2 for damage in
compression. Unilateral effect and stiffness recovery
– To restore the material stiffness, passing from tensile
(damage deactivation) are also included. Inelastic strains
to compressive behavior.
are taken into account thanks to an isotropic tensor. The
– To null the tensile inelastic strains creating after ten-
Gibbs free energy of this model can be expressed as:
sion when excessive (over rf ) compression is reached.
hriþ : hriþ hri : hri t
v¼ þ þ ðr : r  Trðr2 ÞÞ Damage criteria are expressed as fi ¼ Yi  Zi (i ¼ 1
2Eð1  D1 Þ 2Eð1  D2 Þ E for tension or 2 for compression, Yi is the associated
b 1 D1 b2 D2 force to the damage variable Di and Zi a threshold de-
þ f ðrÞ þ TrðrÞ ð15Þ
Eð1  D1 Þ Eð1  D2 Þ pendent on the hardening variables). The evolution laws
for the damage variables Di are written as
where b1 and b2 are material coefficients, whereas h
iþ 1
and h
i denotes the positive or negative values of the Di ¼ 1  ð19Þ
given variable. f ðrÞ and rf are the crack closure func- 1 þ ½Ai ðYi  Y0i Þ Bi
tion and the crack closure stress respectively. E is the where Y0i is the initial elastic threshold and Ai , Bi are
initial YoungÕs modulus and m the Poisson ratio. material constants.
For the 1D expression, the total strain is: e ¼ ee þ ein A typical stress–strain response produced by this
rþ r model for a uniaxial cyclic loading which passes through
ee ¼ þ traction, compression and again traction is given in
Eð1  D1 Þ Eð1  D2 Þ
ð16Þ Fig. 3.
b 1 D1 b2 D2
ein ¼ f ðrÞ þ The material coefficients Y0i , Ai ; Bi , bi (i ¼ 1, 2) and rf
Eð1  D1 Þ Eð1  D2 Þ can be identified from uniaxial tests. They are only
function of intrinsic material characteristics (aggregate
where ee the elastic strain and ein the inelastic strain. The size, concrete mix, . . .).
partition of the stress is obtained as:

r > 0 ! rþ ¼ r; r ¼ 0 3. A soil–structure interaction macro-element
ð17Þ
r < 0 ! rþ ¼ 0; r ¼ r
It has already been noted that, after severe earth-
and for the crack-closure function: quakes, some structures with different properties from
1228 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239
0 1 0 1
surrounding ground have been preserved from any V0 V
damage. The interaction of the structure with the soil B C 1 B C
F  @ H0 A  @ H A and
can lead to a less severe response of the structure 0
Bqmax
M M=B
either by modifying the support condition with respect 0 01 0 1
to a clamped boundary or accounting for modification z z
B C 1B C
of the properties of the interacting system. In particu- u  @ x0 A  @ x A ð20Þ
B
lar the low stiffness of the soil reduces the system fre- h0 Bh
quency, and may, depending on the frequency content
of the input motion, drive outside of the resonance where B is the width and Vmax ¼ Bqmax the bearing ca-
zone of the input spectrum. Moreover, the nonlineari- pacity of the foundation. Since it is a strip foundation,
ties in the surrounding soil and interface (yielding of all terms are expressed for a unit transverse length (unit
the soil and uplift of the structure) lead to a strong depth of the footing).
energy dissipation and further reduce the forces in the Three different phenomena are represented by the
structure. forces acting on the foundation: the plasticity of the soil
To model these phenomena we developed a simplified surrounding the foundation or near field, the uplift of
method for soil–structure interaction. The overall linear the foundation and the elasticity of the soil in far field.
and nonlinear behavior in the soil and at the interface is They are modeled through three interacting models (Fig.
reduced through a macro-element expressed in three 4). Plasticity of the soil and uplift are strongly nonlinear
degrees of freedom on the strip foundation. The con- and coupled since they both concern the near field while
stitutive laws governing the behavior of this macro- elasticity concerns the far field.
element are developed within the framework of the The total displacement increment is partitioned into
plasticity theory. The quasi-static cyclic loading model is elastic, plastic and uplift components, which can be
developed first (see [4,5] followed by an extension to written choosing the corresponding rate decomposition:
dynamic loading case [6].
u_ tot ¼ u_ el þ u_ pl þ u_ up ð21Þ
The only variables of the macro-element are the
forces at the base of the foundation (vertical force V , The global model of constitutive behavior can be written:
horizontal force H and moment M) and the corre-
sponding displacements (vertical displacement z, hori- F_ ¼ ð½Kelpl 1 þ ½Kup 1 Þ1 : u_ tot ð22Þ
zontal displacement x and rotation h) measured at the
centre of the foundation (assumed as a rigid body). where elastoplastic tangent stiffness matrix Kelpl is ob-
These variables normalized by characteristic values to tained from the plasticity model and the tangent stiffness
become dimensionless. matrix Kup from the chosen ‘‘uplift model’’.

Fig. 4. Idealization of the complete model.


L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1229

3.1. Elasticity and damping 3.2. Plasticity model

The energy dissipation mechanism in the near field Finite element analysis with an efficient model for the
arises exclusively from the nonlinearities included in the soil [4] or analytical results, when available, guided us in
plastic and uplift cyclic models. The energy dissipation constructing the macro-model presented in the follow-
in the far field arises from the radiation boundary con- ing.
ditions in semi-infinite medium. The latter is modeled The elastoplastic model provide the elastic and
through the imaginary part of the elastic impedances [8], plastic parts of the displacement increments:
represented by a dashpot, and the real part is the elastic
stiffness matrix. F_ ¼ Kelpl : ðu_ el þ u_ pl Þ ð26Þ
0 0 1
Kzz 0 0 with
Kel ¼ @ 0 Kxx 0
0 A
0 1
0 0 Khh Kelpl ¼ Kel  ðKel : PÞ  ðQ : Kel Þ;
0 K 1 h0 þ h0
zz
0 0 ofi og
B qmax C Q¼ ; P¼ ð27Þ
B Kxx C oF oF
B 0 C
¼B 0 C ð23Þ
B qmax C where fi is the yield surface and g the plastic potential
@ Khh A
0 0 for nonassociated flow rule employed herein.
B2 qmax
One also defines a failure surface from analytical [20]
and numerical results [25], based on limit analysis for


0:73 2 2 structure overturning with uplift mechanisms (the uplift
Kzz ¼ G0 ð1 þ 2aÞ; Kxx ¼ G0 1 þ a ; influence is also included in the plasticity model).
1m 2m 3

2
!2 !2
p B 1
Khh ¼ G0 1þ a ð24Þ H0 M0
2ð1  mÞ 2 3 fi ¼ þ 1¼0
aV 0c ð1  V 0 Þd bV 0e ð1  V 0 Þf
where a is the gradient of the variation of the shear ð28Þ
modulus G ¼ G0 ð1 þ afÞ (G0 the shear modulus at z ¼ 0)
with respect to the depth f ¼ 2z=B and m the PoissonÕs For a soil with constant gradient of cohesion with depth,
ratio. the coefficients in the last expression are given as:
In the same way, the coefficient of the damping ma-
a ¼ 0:32=x; b ¼ 0:37=x0:2 ;
c ¼ 0:25; d ¼ 0:55;
trix are [8]:
qmax
e ¼ 0:8; f ¼ 0:8 with x ¼
3:4 q0 max
Czz ¼ qVs0 Ab ; Cxx ¼ qVs0 Ab ;
pð1  mÞ
where qmax is the actual ultimate bearing capacity of the
3:4
Chh ¼ qVs0 Ibh ð25Þ foundation (soil with cohesion gradient c ¼ c0 þ gz) and
pð1  mÞ q0 max the ultimate bearing capacity for an homogeneous
soil with a constant cohesion c0 .
where q is the mass density of the soil, Ab the areapof theffi
ffiffiffiffiffiffiffiffiffiffi The model is particularly well suited for the evolution
foundation, Ibh its quadratic moment and Vs0 ¼ G0 =q
of the yield surface towards the failure surface with a
the shear velocity at z ¼ 0.
vertical force being almost constant, for reproducing the
Remark: Step by step analysis prevents from includ-
behavior of the soil and the foundation, initially sub-
ing the frequency variation of the dynamic impedances.
mitted to the weight of the structure, and then loaded
Constant values are thus chosen (static value for stiff-
mainly in the (H ; M) plane (seismic loading).
ness, which are already given for the model, and the
value near resonance of the soil structure system for
2
2
H 0  aCcH M 0  bCcM
damping, which gives the better average damping). This fi ¼ þ 1¼0 ð29Þ
qCcH qCcM
choice is acceptable for a semi-infinite soil medium with
a slow and continuous properties variation (nearly ho- with CcH ¼ aV 0c ðc  V 0 Þd ; CcM ¼ aV 0e ðc  V 0 Þf ; c ¼ v þ
mogeneous soil). In this case the variation of the coef- ð1  vÞðq þ sÞ; v ¼ N =Vmax and N is the weight of the
ficients is low over the frequency range of interest (0–10 structure (initialisation under dead load). In (29) above
Hz). For a foundation lying on a thin soil layer, this s  ð 0 a b ÞTqisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the kinematic hardening vector with
approximation is no longer valid. In this case one could
take the values corresponding to the fundamental fre- s ¼ ðs : sÞ1=2 ¼ a2 þ b2 and is the isotropic hardening
quency of the soil building system. parameter.
1230 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239

The consistency rule, the criterion of noninterpene- 0 1


oz0up oz0up
tration of the failure surface with the loading surface 0
B oV 0 oM 0 C
B C
and a relationship obtained from the observation of the F_ ¼ Kup : u_ up ¼ ðCup Þ1 : u_ up ; Cup ¼ B 0 0 0 C
foundation behavior during uplift allow us to define the @ 0up 0up A
oh oh
evolution of the isotropic and kinematic hardening pa- 0
oV 0 oM 0
rameters. The kinematic hardening increment is defined ð34Þ
by its amplitude s_ and its direction l:
The displacements induced by uplift are given as:
l h0 1
s_ ¼ s_ with s_ ¼ hk_i ofi and k_ ¼ 0 Q : F_
jlj
l
: jlj þ ofi h ðd=dmax Þ2
os oq h0up ¼ ð1  V 0 Þh0 and
1  d=dmax
ð30Þ
jh0 j d=dmax
z0up ¼ ð1  V 0 Þ þ lnð1  d=dmax Þ
where 2 1  d=dmax
0 1 ð35Þ
0
B H
C1
cos w  H 0
C ! 0ð0Þ
B c C C1H M 0  bC1M where dmax ¼ ð4=V 0 ÞjMmax
0
 M0 jis the uplift reached
lB
B 1 C H
C and w ¼ a tan
C
for the maximum moment at failure for the given V 0 and
@ CM sin w  M A
0 C1M H 0  aC1H
for H 0 ¼ 0, that is to say for Mmax
0
¼ bV 0e ð1  V 0 Þf .
CcM 0
For a given V , the uplift ratio d is obtained as fol-
ð31Þ lows:
0ðiÞ
The isotropic hardening increment is set equal to the • if jM 0 j 6 jM0 j, d ¼ 0
0ð0Þ 0ð0Þ
kinematic one: q_ ¼ s_ . The plastic modulus h0 is fitted on • if jM 0 j P jMp0ðiÞ j, d ¼ ð4=V 0 ÞjM 0  M0 j with M0 ¼
the M 0 –h0 curves since it is the most important behavior ðV 0 =4Þ expAV
0

0ðiÞ 0ðiÞ
of the foundation • if jM0 j < jM 0 j < jMp0ðiÞ j, d ¼ ð4=gV 0 ÞjM 0  M0 j

K el ðdM 0 =dh0 Þ2M 0 of dV 0 of dH 0 of 0ðiÞ
where M0 is such that dðiÞ 0 0ðiÞ 0ðiÞ
h0 ¼ elhh þ þ p ¼ ð4=gV ÞjMp  M0 j ¼
Khh  ðdM 0 =dh0 Þn2 oV 0 dM 0 oH 0 dM 0 oM 0 0ð0Þ
ð4=V 0 ÞjMp0ðiÞ  M0 j with Mp0ðiÞ the maximum moment
ð32Þ reached during the loading history and dðiÞp is the corre-
sponding uplift.
! The right and left uplift are treated independently
dM 0 0 K 0 h0 supposing they remain uncoupled.
¼ Khh exp  hh0
dh0 M1 Note: The macro-element treats uplift in a continuous
global model. No coefficient of restitution to describe the
The external normal to the yield surface is not conve- energy dissipation upon impact while the uplifted foun-
nient for defining the direction of the plastic displace- dation will contact the soil has been introduced.
ments and a nonassociated flow rule is required. The
plastic potential g has been chosen as: 3.4. Use of the model


0 2


0 2
H M This SSI macro-element which was implemented in
g¼ þ þ V 02  1 ¼ 0 ð33Þ
j n computer code FEAP [24] is well suited for the dynamic
simulation of slender structures with potentially large
uplift and moderate yielding in the soil (and no sliding of
3.3. Foundation uplift model the footing on the soil). When using this model it is
recommended to have a safety factor SF ¼ Vmax =V > 3
The foundation uplift behavior on an elasto-plastic (where Vmax is the bearing capacity of the foundation, V
soil is a very complex phenomenon which depends the vertical load on the foundation), and a slenderness
strongly on the level of soil yielding. A cyclic model ratio h=B > 1 (where h is the height of the dynamic
which takes into account the irreversibility of the uplift centre of gravity and B the width of the foundation).
behavior due to plasticity is developed as a direct mod-
ification of the uplift model of Cremer [4] developed for
elastic soils. 4. Numerical simulations and experimental results (CA-
The uplift model operates only on the uplift part of MUS program)
the total displacement introducing coupling between the
vertical displacement zup and the rotation hup of the In this section we briefly present the main results of
foundation: the experimental program carried out within the re-
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1231

search network CAMUS along with the comparative narrower effective band width. The complete experi-
numerical simulations. mental sequence was: Nice 0.25 g, San Francisco 1.13 g,
The main purpose of this experimental program Nice 0.4 g and Nice 0.71 g.
consists in demonstrating the ability of lightly reinforced
concrete bearing walls to sustain very severe earthquake. 4.1. Model calibration (modal analysis and damping
The design is based on a concept favoring damage calibration)
spreading over several storeys of a lightly reinforced
concrete wall. This kind of design leads to a lower per- Eigenfrequencies of the reduced scale numerical
centage of reinforcements with an optimized distribution model anchored to the shaking table show a higher
generating a wide cracks pattern allowing for dissipation global stiffness. Namely the frequency of the first ei-
of great amounts of energy. As a consequence, one ob- genmode measured before the test is 7.3 Hz, as opposed
tains a vertical rising of the masses resulting in energy to 10.3 Hz predicted by the numerical model. This in-
transformation (from kinematic to potential instead of dicates the necessity to take into account the shaking
storing strain energy in the structure) and an increasing table compliance as well as the anchorage system which
ductility thanks to this particular way of dissipating the can never be considered as perfectly clamped. The sec-
earthquake energy. In order to test this concept experi- ond eigenmode corresponds to pumping with an eigen-
mentally, a one-third scaled model (see Fig. 5) has been frequency of 40 Hz as predicted by the finite element
installed on the shaking table of CEA, composed of two model. During the test, the vertical displacement of the
parallel braced walls linked by six square slabs. A mass induced by the openings of cracks excited this
massive reinforced concrete footing allows the anchor- second mode. The resulting vibration measured through
age to the shaking table. the vertical acceleration of the shaking table, with a
The similarity laws to a full-scale structure and the frequency of 20 Hz leads to variation of the vertical
mock-up require additional masses of 6.55 t at each dynamic forces which can be very important for rein-
storey. forced concrete structure failure. This effect must clearly
The problem is assumed to be two-dimensional since be taken into account through the vertical stiffness of
the mock-up is loaded through horizontal acceleration the connecting rods. A simplified model for the base-
parallel to the walls and the presence of steel bracing ment footing is assumed consisting in an horizontal
systems at each storey placed perpendicularly to the beam whose properties are selected in accordance with
loading direction prevent occurrence of any torsional the shaking table stiffness with Kv ¼ 48EI=l3 and Kh ¼
modes.
pffiffiffiThe accelerograms are scaled in time with a ratio 12EI=l. This kind of footing element adds two eigen-
of 1= 3 in order to account for the similarity rules. Two modes, as required for the CAMUS structure, since as
types of accelerograms are applied: a synthetic one already discussed, the second vertical mode plays a very
generated from the french code [1] Nice S1 as the far important role. Finally, YoungÕs modulus of the footing
field type earthquake and the San Francisco earthquake beam was set to 10 000 MPa in order to introduce an-
of 1957 (Fig. 6) as the near field one. The response chorage and contact defects.
spectra show clearly the main difference of these two Table 1 summarizes the correction of fundamental
earthquakes: on one hand Nice has a broad frequency frequencies of the system due to these new boundary
content and on the other hand, San Francisco has a conditions.

Fig. 5. CAMUS structure.


1232 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239

Fig. 6. Nice and San-Francisco accelerograms and response spectrum.

Table 1 4.2. Comparisons between experimental and numerical


Eigenmode adjustment results
Initial Adjusted Measured
modeling modeling frequencies The complete experimental sequence of loading (four
(Hz) (Hz) (Hz) earthquakes) was computed using the proposed numer-
First mode 10.3 7.4 7.3 ical model, with the material parameters following:
Second mode 40.0 19.0 20.0 YoungÕs modulus of concrete Ec ¼ 30 000 MPa for a
maximum compressive strength of fc ¼ 35 MPa and
tensile strength ft ¼ 3 MPa, YoungÕs modulus of steel
Es ¼ 200 000 MPa with yield stress ry ¼ 414 MPa ac-
Damping is another important concern. In particu- counting for hardening with ultimate stress ru ¼ 480
lar, the classical viscous damping is assumed to account MPa.
for various sources of energy dissipation not related to The numerical and experimental results are com-
concrete cracking [22]. Despite the lack of clear physical pared in terms of time-history displacements, ductility
interpretation, the damping is often introduced in the and local damage criterion (Figs. 7–9).
analysis through the Rayleigh damping matrix which is For the lowest level of loading (Nice 0.24 g), when
expressed as a linear combination of the mass matrix the concrete cracks in tension and the reinforcement
and the initial stiffness matrix: bars reach the yield stress, the global behavior of the
C ¼ aM þ bK structure is well represented by the numerical model.
The loss in stiffness during the later stages of loading and
The particular values of damping coefficients are set to the decrease of the fundamental frequency is also pre-
ensure a relative damping value of 1% in the first mode dicted in a fairly acceptable manner, with a maximum
and 2% in the second mode. Great care is taken to keep displacement always slightly underestimated before and
these damping values as stable as possible during all the overestimated after the maximum of the load.
analysis, despite the fact that cracking induces reduction Table 2 summarizes the comparisons of experimental
of the stiffness and shift of the fundamental frequency. and computed results for reaction forces for different
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1233

accelerograms which indicates globally a very good


agreement.
Another feature of material modeling during cyclic
loading, which is barely taken into account, is the cracks
closing there described by the crack closure function of
the constitutive relation. When re-closing the cracks, the
point where the compressive stiffness is totally recovered
depends on the crack closure stress of the model rf (see
Fig. 10). In the CAMUS experiment, the shocks induced
by cracks closure which activates the vertical eigenmode
leads to variation of vertical dynamic forces. Experi-
mentally, the phenomenon can be quantified by the
measurement of an induced vertical acceleration at the
Fig. 7. Nice 0.24 g: experiment/computation comparisons.
shaking table. The discontinuity of the stiffness when rf
is reached reproduces the shock, two computations
performed with different forms of the crack closure
function (Figs. 11 and 12) allow to reproduce this ex-
perimental observation.
Modeling this kind of feature become very important
for reinforced concrete structures where the design takes
into account the coupling between flexural and axial
loading. Through this short analysis we can easily re-
mark that the local behavior of concrete plays a very
important role on the global structural response and
that the material parameters of the crack closure func-
tions can be identified using different levels of loading
(see Table 3).
Due to cracking of the concrete, the global stiffness of
Fig. 8. San-Francisco 1.13 g: experiment/computation com-
the structure decreases. This variation does not remain
parisons.

Fig. 10. Effect of the crack closure stress on the slope of the
Fig. 9. Nice 0.71 g: experiment/computation comparisons. stiffness recovery during cyclic loading.

Table 2
Global response comparisons
Displacement (cm) Shear load (kN) Moment (kN m) Vertical l (kN)
Exp. Comp. Exp. Comp. Exp. Comp. Exp. Comp.
Nice 0.24 g 0.72 0.61 65.9 65 200 200 202 190
SF 1.1 g 1.2 1.1 106 90 280 240 271 270
Nice 0.4 g 1.35 1.1 86.6 75 280 240 217 225
Nice 0.7 g 4.4 3.9 111 120 345 380 312 310
1234 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239

Fig. 11. Initial model: vertical load variation.


Fig. 13. Nice 0.24 g: displacement spectrum.

1.5

Fig. 12. Modified model: vertical load variation.

constant during the loading but depends on the cracks Fig. 14. Nice 0.71 g: displacement spectrum.
opening situation (opened or closed crack). This stiffness
decrease shifts greatly the natural frequencies. The
analysis of the fundamental frequency of the structure the earthquake motion. One way to estimate the domi-
subject to a white noise after each sequence is not sat- nant frequency during the loading is to plot the response
isfactory because it is applied on the cracked structure spectrum of horizontal displacement on the top of the
but for closed cracks (this is due to the low level of the structure. Considering that this is the spectrum of a
white noise which does not allow to open cracks). Such nonlinear response signal, the results should only have a
experimental measures cannot capture this kind (chan- qualitative sense. These are shown on Figs. 13 and 14 for
ges in stiffness reduction during the earthquake) of the Nice 0.24 g and Nice 0.71 g earthquakes.
structural behavior and provide the first eigenfrequency We observe that such a frequency shift, although far
of 6 Hz instead of the 3 Hz., during the strongest part of from being negligible, is realistic. It indicates that the

Table 3
Vertical load: crack closure function identification (rf ) (kN)
3.5 MPa 1.75 MPa 1.3 MPa 1.0 MPa Experiment
Nice 0.25 g 115 119 120 120 138
SF 1.13 g 150 160 200 218 198
Nice 0.40 g 119 132 155 165 146
Nice 0.71 g 140 190 240 265 248
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1235

local degradations is satisfactorily accounted for and


permits to follows the global stiffness reduction of the
structure.
The final crucial point in modeling pertains to com-
puting corresponding local level of degradation for both
concrete and steel. Fig. 15 shows a typical result of this
kind with the level of damage in concrete and the irre-
versible strains in steel at the end of the complete
loading sequence.
One can observe that the global trend observed ex-
perimentally is recovered in the computations at this
local level. The location of the critical region are posi-
tioned on the upper level. This particular behavior is
mainly due to the effect of the second pumping eigen-
mode, whose main effect results in shifting the failure
region. Fig. 15 clearly demonstrates that the computed
strains always underestimate the experimentally ob-
tained values. The lack of information at local scale (for
example real strain in a steel bar at the location of a
crack) is the major drawback preventing the designer
from expressing physical criteria describing rupture. Fig. 16. Camus IV mock-up.
Due to the small reinforcement ratio, failure phenome-
non can only happen by rupture of the steel bars under
container dimensions limited by the shacking table ca-
tension, thus post-peak behavior cannot be represented.
pacity where chosen 4 m  4 m, but computations have
shown a low influence of the lateral boundaries. As it
4.3. Simulation of Camus IV test was impossible to have a realistic depth of sand capable
of reproducing the real failure mechanisms of the soil,
We present in this paper a comparison of the nu- the goal was limited to find a more or less realistic
merical results obtained with the SSI model with those overturning stiffness of the foundation. The depth was
measured in the test Camus IV [3]. thus chosen equal to 0.4 m.
The mock-up was submitted to the Nice synthetic
accelerogramm applied horizontally at different levels
4.3.1. Test characteristics
and followed by the natural signal of Melandy Ranch.
The five storey mock-up (with the same dimensions
The natural frequency of the interacting system table-
as one used in Camus I test) is mounted on the shaking
sand-mock-up was 3.5 Hz which is much lower than for
table simply supported on a sand container (Fig. 16).
the clamped mock-up of Camus I with frequency of
The main goal is to study the boundary conditions in-
about 7.0 Hz.
fluence due to the presence of soil and to the possible
uplift and sliding.
The two shallow foundations (Fig. 17) where di- 4.3.2. Choice of the parameters
mensioned with a safety factor equal to 3 to avoid too Although the test conditions are far from the
early failure of the soil when uplift will occur. The sand assumptions used for elaboration of the model, by

Fig. 15. Degradation of the structure at the end of the analysis. Location of the cracks on the instrumented wall at the end of the
loading sequence. Measured and computed strains (maximum values).
1236 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239

Fig. 18. Modeling of Camus IV mock-up.

Fig. 17. Details of the foundations.


Table 4
Chosen parameter values
B ¼ 2:1 m, l ¼ 0:8 m, Vmax ¼ 0:8 MN
adapting the parameters (stiffness, failure criterion and qmax ¼ 0:476 MPa
damping) to the particular conditions due to the low Kzzel ¼ 320 MN/m, Kxx el
¼ 50 MN/m
el
Khh ¼ 150 MN m/rad
depth of soil, it is possible to use the macro-element
Czzel ¼ 2 MN s/m, Cxx
el
¼ 2 MN s/m
since the yielding of the soil was limited during the test el
Chh ¼ 2 MN m s/rd
(SF  4; h=B  1:5). The soil acted as an isolator and a ¼ 0:7, b ¼ 0:57
during the test no damage was observed in the structure. c ¼ 1, d ¼ 1, e ¼ 1, f ¼ 1
The structure is thus modeled with elastic beams (Fig. j ¼ 0:25, n ¼ 0:17
18). Concentrated mass and inertial moments are placed
at each storey. The test is modeled in 2D, which is re-
alistic for the structure (since only one wall is modeled)
but brings the strong assumption on a shallow founda- – The bearing capacity was calculated with a safety fac-
tion. tor equal to 3 for a rough foundation on a sand layer
The following fitting procedure for the parameters with a limited depth. Since the sand container is also
has been used to take into account the particular shape laterally limited, the safety factor is estimated to 4.
of the soil in the test (the chosen values of parameters We can thus estimate the bearing capacity Vmax .
are given in Table 4): – The failure criterion coefficients have been adapted
from the maximum moment measured for high excita-
– The horizontal and rotational stiffness have been ob- tion levels of the test and assuming a friction ratio
tained from the very low level response with quasi-lin- equal to 0.7 at the interface.
ear behavior. – Due to the low depth of soil and to its high density,
– The vertical stiffness has been obtained by calibration the uplift in the test was greater than in a foundation
of the vertical natural frequency of the soil–structure lying on an infinite soil. An amplification coefficient of
system. the uplift displacements equal to 3 has been used.
– The radiation damping coefficients should not exist
here since the soil is not semi-infinite, nevertheless a
small value is kept to take into account the damping 4.3.3. Results
in the whole system (since no damping is put in the In Fig. 19a, the S shape hysteresis loops of the mo-
structure). ment––rotation response due to the strong uplift is
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1237

Fig. 19. Comparison between experients and simulation results for Camus IV test.

clearly observed indicating energy dissipation, despite a The shear force and moment are reduced due to the
low depth of soil. The residual settlement (positive) uplift (see Table 5). We see on the other hand a large
visible in Fig. 19b confirms that the soil yielding is also dynamical variation of the vertical load which can reach
present. as much as half of the static load (Fig. 19c). Namely, a
1238 L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239

Table 5 [4] Cremer C. Modelisation du comportement non lineaire des


Comparison between Camus I et Camus IV [3] fondations superficielles sous seisme, PhD thesis, Civil
Storey Camus I Camus IV Camus IV Camus IV Engng, LMT Cachan, ENS Cachan, France. 2001.
0.71 g 0.52 g 0.90 g 1.11 g [5] Cremer C, Pecker A, Davenne L. Cyclic macro-element of
soil–structure interaction: material and geometrical non-
Fifth floor linearities. Int J Num Anal Meth Geomech 2001;25:1257–
M (kN m) 36.5 14.3 25.2 26.4 84.
T (kN) 40.0 15.9 28 29.4 [6] Cremer C, Pecker A, Davenne L. Modelling of non linear
First floor dynamic behavior of a shallow strip foundation with
M (kN m) 345 136.7 137.2 133 macro-element. Earthquake Eng Struct Dyn, in press.
T (kN) 111 51.4 56.5 58.6 [7] Dube J.F. Modelisation simplifiee et comportement visco-
endommageable des structures en beton. PhD thesis, ENS,
Cachan. 1994.
variation of 80 kN is measured while the static load is [8] Gazetas G. Foundations vibrations. In: Fang H-Y, editor.
Foundation engineering handbook. NY: van Nostrand
200 kN. This is due to the uplift and the impact of the
Reinhold; 1991 [Chapter 15].
foundation on the soil.
[9] Ibrahimbegovic A, Frey F. Finite element analysis of linear
Despite it was not developed for this kind of condi- and non-linear planar deformations of elastic initially
tions, the model gives good results after some adapta- curved beams. Int J Num Meth Engng 1993;36:3239–58.
tions of the parameters. [10] Ibrahimbegovic A, Wilson EL. A methodology for dy-
namic analysis of linear structure–foundation systems with
local nonlinearities. Earthquake Eng Struct Dyn 1991;
5. Conclusions
19:1197–208.
[11] Ibrahimbegovic A, Wilson EL. A modified method of
Efficient approaches to finite element analysis in incompatible modes. Commun Appl Num Meth 1991;
earthquake engineering are presented in this paper. They 7:187–94.
have been developed for the nonlinear transient analysis [12] Kotronis P. Cisaillement Dynamique de Murs en Beton
of reinforced concrete structures under seismic loading Arme. Modeles Simplifies 2D et 3D. PhD thesis, ENS,
with various boundary conditions including soil–struc- Cachan. 2000.
ture interaction. Accurate constitutive relations are used [13] La Borderie Ch. Phenomenes unilateraux dans un
with dedicated simplified structural finite elements materiau endommageable: modelisation et application a
(multifiber beam element or soil–structure interaction lÕanalyse de structures en beton. PhD thesis, University of
Paris VI. 1991.
macro-element). This brings robustness and efficiency
[14] Leger P, Dussault S. Nonlinear seismic response analysis
while keeping the accuracy. Numerical simulations of using vector superposition methods. Earthquake Eng
two tests of the European research network program Struct Dyn 1991;20.
CAMUS are presented. The first one focuses on the re- [15] Lemaitre J, Chaboche JL. Mechanics of solids material.
inforced concrete behavior while the second concerns the Cambridge University Press; 1990.
soil–structure interaction influence. The simulation re- [16] Lowes LN, Moehle JP. Evaluation and retrofit of beam-
sults are in good agreement with the experimental ones. column T-joints in older RC bridge structures. ACI Struct
These simplified methods permit to perform para- J 1999;96(4):519–32.
metric studies since they are not overly computer time [17] Mazars J. A description of micro- and macroscale damage
consuming. We can for instance study the influence of of concrete structures. J Eng Fract Mech 1986;25(5/6):729–
37.
the input motion (type, frequency content, . . .), the in-
[18] Newmark NM. A method of computation for structural
fluence of the type of design concept of the structure dynamics. ASCE J Eng Mech Div 1959;85:67–94.
(shear wall with ‘‘plastic hinge’’ versus ‘‘French shear [19] Paulay T, Priestley MJN. Seismic design of concrete and
wall’’) or the influence of the soil characteristics. masonry buildings. New-York: John Wiley & Sons Inc.;
1992. 744 p.
[20] Pecker A. Analytical formulae for the seismic bearing
References capacity of shallow strip foundations. In: Seco e Pinto PS,
editor. Seismic behavior of ground and geotechnical
[1] AFPS. Guide AFPS 92 Pour la protection des ponts. structures. Rotterdam: Balkema; 1997. p. 261–8.
Presse de lÕENPC. France. 1995. [21] Przemieniecky JS. Theory of matrix structural analysis.
[2] Armstrong PJ, Frederick CO. A mathematical representa- New-York: MCGraw-Hill; 1985.
tion of the multiaxial Bauschinger effect. GEGB Report [22] Ragueneau F, La Borderie Ch, Mazars J. Damage model
RD/B/N 731. 1966. for concrete like materials coupling cracking and friction,
[3] Combescure D, Chaudat T. Icons European program contribution towards structural damping: first uniaxial
seismic tests on R/C walls with uplift; Camus IV specimen. application. Mech Cohes Frict Mater 2000;5:607–25.
ICONS project, CEA/DRN/DMT report, SEMT/EMSI/ [23] Rashid JYR, Dameron RA, Dunham RS. Finite element
RT/00-27/4. 2000. analysis of reinforced concrete in bridge seismic design
L. Davenne et al. / Computers and Structures 81 (2003) 1223–1239 1239

practice ASCE Commitee Report In: Shing PB, Tanabe T, [25] Ukritchon B, Whittle AJ, Sloan SW. Undrained limit
editors. Modeling of inelastic behavior of RC structures analysis for combined loading of strip footings on clay. J
under seismic loads. Structural Engineering Institute; 2001. Geotech Geoenv Engng March 1998:265–76.
p. 217–33. ISBN 0-7844-0553-0. [26] Zienkiewicz OC, Talyor RL. 5th ed. The finite element
[24] Taylor RL. FEAP: A finite element analysis program, method, Vols. I, II and III. Oxford: Butterworth Heine-
version 5.01 manual. University of California, Berkeley. 1996. mann; 2000.

View publication stats

You might also like