You are on page 1of 21

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

net/publication/273004389

Numerical simulation of RC infill walls under cyclic loading and calibration


with widely used hysteretic models and experiments

Article  in  Bulletin of Earthquake Engineering · February 2015


DOI: 10.1007/s10518-015-9739-9

CITATIONS READS

4 1,925

2 authors, including:

Beyza Taskin
Istanbul Technical University
32 PUBLICATIONS   139 CITATIONS   

SEE PROFILE

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

Verticallt irregular rc buildings View project

A Novel Post-Earthquake Damage Survey Sheet: Part II- Masonry Buildings View project

All content following this page was uploaded by Beyza Taskin on 03 March 2015.

The user has requested enhancement of the downloaded file.


Bull Earthquake Eng
DOI 10.1007/s10518-015-9739-9

ORIGINAL RESEARCH PAPER

Numerical simulation of RC infill walls under cyclic


loading and calibration with widely used hysteretic
models and experiments

Orkun Görgülü • Beyza Taskin

Received: 5 June 2014 / Accepted: 20 February 2015


Ó Springer Science+Business Media Dordrecht 2015

Abstract Addition of reinforced concrete (RC) infill walls into the structural system has
been a commonly preferred strengthening technique within the last decades for seismic
rehabilitation of RC frames. As a consequence, generating a representative numerical
model of an RC infill wall has become an important issue. As the initial step of this study,
measured structural responses of two selected well-known large-scale RC infill wall ex-
periments subjected to displacement controlled cyclic loading are taken into account.
Later, by calibrating the numerical model prepared in Perform-3D computer program
utilizing fiber cross-sections, a practical and a compatible analytical model is obtained and
proposed herein. Structural systems of the experiments are mathematically modeled by
elements consisting of vertical and horizontal fiber layers to represent the bending/axial
behavior and to control the out of plane displacements, respectively. Nonlinear behavior of
the reinforcing steel is represented by a tri-linear backbone curve without strength
degradation, while a multi-linear hysteretic behavior considering the strength loss is uti-
lized for the structural concrete. Furthermore, those recently conducted experiments are
simulated by a couple of widely used hysteretic models for comparative purposes, which
are preferred in most cases by the researchers during the analytical investigation of RC
structures, so that their adequacy for reflecting the nonlinear behavior of infill walls are
also studied. It is shown with comparisons for the experimentally measured and the
analytically derived results that the calibrated mathematical model proposed herein is more
compatible with the measured values than the widely used hysteretic rules for capturing the
behavior of these types of frames retrofitted by RC infill walls under reversed cyclic
loading. Although numerical simulations are carried out for a limited number of tests and it
is assumed that sufficient amount of anchoring dowels is provided at the interface of the
existing frame and the RC infill, the proposed calibrated model conforms to both

O. Görgülü
Graduate School of Science Engineering and Technology, İstanbul Technical University,
Maslak, Istanbul, Turkey

B. Taskin (&)
Department of Civil Engineering, İstanbul Technical University, Maslak, Istanbul, Turkey
e-mail: btaskin@itu.edu.tr

123
Bull Earthquake Eng

experiments’ measured responses by means of seismic behavior for not only the undam-
aged single bay frames converted to the RC infill wall, but also pre-damaged multi-bay
strengthened structures, which include structural deficiencies like low concrete strength,
inadequate stiffness and insufficient confinement.

Keywords Shear wall  Reinforced concrete infill wall  Fiber element based non-linear
modeling  Seismic behavior  Hysteretic constitutive models

1 Introduction

Most of the existing reinforced concrete (RC) structures, especially in Europe, do not
comply with recent earthquake codes, since most of them were designed and built in the
1970s (Thermou et al. 2011). Due to their insufficient strength, ductility and stiffness, these
buildings should be rehabilitated or replaced by means of reducing the seismic risk to
acceptable levels (Varum et al. 2013). Many retrofitting techniques such as steel bracing,
jacketing of columns, addition of RC infill walls and FRP wrapping are applicable for the
seismic rehabilitation of existing RC structures. To decrease the vulnerability of existing
frame systems, introducing the RC infill walls is one of the most preferred retrofitting
techniques especially in Turkey (Canbay et al. 2002). In this technique, some of the
existing RC frames are converted into the shear walls, therefore seismic capacity of the
existing structure is improved significantly (Fardis et al. 2013).
During the recent years, many experimental and analytical researches have been con-
ducted to investigate the seismic behavior of RC infill walls. Higashi et al. (1980) per-
formed several experiments on RC infill wall specimens and verified the test results by
using nonlinear force controlled analysis. (Canbay 2001) and Anil and Altin (2006) in-
vestigated the cyclic behavior of RC infill walls experimentally and carried out static
pushover analyses to validate the accuracy of the test results. Darwish (2006) worked
through the numerical modeling of RC infill walls in terms of strengthening the bridge
frame and conducted parametric analyses under cyclic loading. Sivri (2011) examined the
calibration of the RC infill wall experiments by using the static pushover analyses. Stre-
pelias et al. (2012) tested the frame strengthened by an RC infill wall subjected to dis-
placement controlled cyclic loading and performed nonlinear dynamic simulation to verify
the pseudo-dynamic test results. As a general approach, most of the researchers preferred
pushover analysis method to simulate the nonlinear behavior of RC infill walls. However,
this method is based on static loading and not able to capture the cyclic behavior of RC
infill wall under earthquake loading. On the other hand, a few researchers performed
nonlinear dynamic analyses using finite element model which is neither feasible nor
practical to investigate the seismic behavior of RC infill walls due to its complexity and
excessive computer time requirements during the analysis.
This paper involves in the generation of a representative and a practical numerical
model for RC infill walls by calibrating two different well-known experiments under cyclic
loading. One of the adopted experiments for the analytical calibration was performed by
Strepelias et al. (2014) to investigate RC frames rehabilitated with RC infill walls. During
the research, 3:4 scaled, one-bay, four-story frames infilled with RC walls was subjected to
quasi-static load cycles and then the damage was repaired before the pseudo-dynamic test
had been applied. The other experimental investigation taken for consideration was

123
Bull Earthquake Eng

conducted by Canbay et al. (2003) to observe the behavior of RC infill walls. In this second
research, test specimen was constructed by two-story, three-bay and 1:3 scale frames but
unlike the first experiment, these frames included the common negative features encoun-
tered in many existing RC buildings such as low concrete strength, inadequate stiffness and
insufficient confinement. The cyclic behavior of the RC infill wall in each conducted
experiment is numerically simulated and calibrated by using fiber element model approach
utilizing the nonlinear analysis program Perform-3D. Other than the proposed mathema-
tical model, two different commonly used hysteretic material models which are mostly
preferred and employed by researchers are also considered and computations are repeated.
Although these models are mostly satisfactory for reflecting the response of RC frames and
shear walls under cyclic loading, their compatibility for the nonlinear behavior of RC infill
walls should be investigated. Finally, numerical results obtained by the proposed model,
commonly used analytical models and experimental findings are compared with each other
and the capability of the proposed model to reflect the actual behavior of RC infill walls is
illustrated in the paper.

2 Overview of the common constitutive hysteretic models employed in analytical


investigations

Numerous constitutive hysteretic models for the cyclic behavior of concrete are proposed
by the analysts in recent years. Two of the earliest and widely used hysteretic models are
referred by the researchers’ names: Clough model by Clough and Johnston. (1966) and
Takeda model by Takeda et al. (1970). According to the first, the unloading stiffness is
assumed to be equal to the initial elastic stiffness of concrete section. For the reloading
cases, the concrete hysteretic model was modified by Otani (1981) providing more ac-
curacy in the structural behavior. In this modified model, the initial stiffness (Ky) depends
on the unloading stiffness (Kr) with the parameters of the unloading stiffness degradation
(a) and the displacements (D) as shown in Fig. 1a. The latter, which is also modified by
Otani (1981) and Kabeyasawa et al. (1983), is another preferred model, based on ex-
perimental investigations. This modified version includes the parameters for strain hard-
ening with stiffness changes at flexural cracking and yielding point. The unloading
stiffness (kr) is defined by the force (Q) and the corresponding displacement (d) at cracking

Fig. 1 Concrete constitutive models: Modified Clough Model (a) and Takeda Model (b)

123
Bull Earthquake Eng

(dc; Qc), yielding (dy; Qy) and the previous maximum displacement (du), respectively as
given in Fig. 1b (Esquivel 1992; Lizuka et al. 2011).
Another widely used model for defining the hysteretic behavior of concrete sections is
known as Modified Kent-Park model, which is introduced by Kent and Park (1971) and
improved by Scott et al. (1982). This model is also employed by many researchers as in
Xiaolei et al. (2008), Martinelli and Filippou (2009) and Zhou et al. (2011) for calibration
of the mathematical model of RC structures under cyclic loading. In this model, the stress–
strain (rc - ec) relationship of concrete is defined in three regions depending on the value
of strain at maximum stress (e0) and strain in %20 of the maximum compressive stress
(e20), as shown in Fig. 2. According to the model, the compressive stress (rc) in concrete is
defined by the coefficients K and Z, which represent the strength increase factor due to
confinement and the slope of the strain softening, respectively.
Alternatively, a stress–strain relationship for concrete members subjected to cyclic
loading is proposed by Mander et al. (1988) and used by researchers and engineering
professionals during numerous analyses (Jalali and Dashti 2010, Xuewei et al. 2011, Jiang
and Liu 2011, Yin et al. 2012). In this constitutive model, the parameter of the unloading
curve is defined based on experimental investigations for not only confined but also un-
confined concrete. For the reloading curve, a linear stress–strain relation followed by a
parabolic transition curve between the degraded point (enew, fnew) and the returning point
(ere, fre) is assumed as given in Fig. 3.
Beside the concrete material model diversity, proposed constitutive hysteretic models
for steel are very limited in the literature. The simplest model for simulating the hysteretic
behavior of steel is the bilinear model with strain hardening used by many researchers. The
only constant required to specify the bilinear model is the strain hardening parameter,
which is defined as the ratio of the post-yield and initial elastic stiffnesses. Unlike the
bilinear model, the well-known Menegotto-Pinto hysteretic model (Menegotto and Pinto
1973) with many parameters to represent the behavior of steel more realistic under cyclic
loading is shown in Fig. 4. This model was modified by Filippou et al. (1983) and preferred
by researchers for the analytic investigation of nonlinear behavior of RC structures in most
cases, as in Orakcal et al. (2006), Martinelli and Filippou (2009), Jalali and Dashti (2010),
Spyrakos et al. (2012), Yang et al. (2013). According to this model, the shape of the
hysteretic curve is defined by the curvature parameter (R), which depends on the initial
shape parameter (R0), coefficients (a1 and a2) and absolute strain difference (n). a1 and a2
coefficients dominate the degradation of curvature, while the parameter R0 the shape of the
transition curve between initial and post-yield stiffness to reflect the Baushinger &
Pinching effect. Filippou et al. (1983) also proposed a modification to account for isotropic

Fig. 2 Modified Kent-Park ⎡ ⎛ε ⎞ ⎛ εc ⎞ ⎤


2

constitutive model for concrete Kf c' ⎢2⎜⎜ c ⎟⎟ − ⎜⎜ ⎟⎟ ⎥ ε c ≤ ε 0


ε ⎠ ⎝ε0
⎣⎢ ⎝ 0 ⎠ ⎥⎦
Kf c' [1 − Z (ε c − ε 0 ) ] ε 0 ≤ ε c ≤ ε 20
0.2 Kf c' ε c > ε 20

123
Bull Earthquake Eng

Fig. 3 Mander model for


concrete (Mander et al. 1988)

Fig. 4 a Menegotto-Pinto constitutive model for steel and b its modified version including the degradation
of the cyclic curve

strain hardening effect by introducing a stress shift to the yield asymptote by introducing
the coefficients a3 and a4 (Orakcal et al. 2006).

3 Simulation of RC infill walls by using the fiber element model approach

Several mathematical models used for the modeling of the inelastic response of frames
strengthened with RC infill walls can be found in the literature. Higashi et al. (1980)
idealized the RC infill walls as bracing elements. Canbay (2001) and Anil and Altin (2006)
used equivalent beam model, while Darwish (2006) and Sivri (2011) preferred finite ele-
ment model for numerical simulations of RC infill wall. Generally, these mathematical
models can be classified into two groups: Microscopic and macroscopic approaches. Mi-
croscopic models such as finite element and multi-layer shell element provide detailed
information of the local response but are not feasible in terms of computational efficiency
and reliability (Orakcal et al. 2004). On the contrary, macroscopic models such as
equivalent beam model; equivalent truss model; strut-and-tie model; multiple-vertical-line-
element model; fiber type model are computationally efficient and are considered being

123
Bull Earthquake Eng

simple models compared to the microscopic procedures (Bhaumik and Raychowdhury


2012). Therefore, to obtain a practical and applicable mathematical model of RC infill walls
for an extensive use, a macroscopic model approach is preferred during the numerical
calibration of the selected two experiments in this study. Since one of the most preferred
macroscopic model is the fiber type model (Guedes et al. 1994), the numerical simulation of
the test specimens are also evaluated by introducing fiber elements in the analytical model.
In this model, the discretization of RC elements is performed at the cross-section by di-
viding the section in an adequate number of elements, namely fibers, for not only confined
and unconfined concrete but also for reinforcing steel (Varum et al. 2013). Concordantly,
hysteretic behavior of each material is assigned to these fibers and the section’s force–
deformation relation is integrated from each fiber by considering the material’s stress–strain
response. The internal force distribution within the element (Dx) is represented by Eq. (1),
in the flexibility based fiber element model. In this expression, Q is defined as the vector of
element forces, while b(x) is the force interpolation functions which is determined con-
sidering the longitudinal axis of the element (x) and the entire element length (L), as given in
Eq. (2). Consequently, the sectional forces are calculated from the element force which is
followed by the determination of the corresponding fiber stresses, while the fiber strains and
the section deformations are calculated by using the fiber stress–strain relations and the
virtual force principle, respectively (Taucher et al. 1991).
Dx ¼ bðxÞQ ð1Þ
" #
 x 0  0x 1
bx ¼ ð2Þ
1 0
L L
One of the main advantages of using fiber based element model is to allow the neutral
axis to shift along the RC wall section during the analyses. Thus, calculation of the cracked
moment of inertia of the cross section is not required. Moreover, by defining the material
properties of the reinforcement, the confined and the unconfined concrete in the cross
section, the complex nonlinear stress–strain behavior is taken into account. The fiber
element model is also compatible with the complex wall structures with irregular openings
(Wallace 2007). On the other hand, effect of the compression stress in the shear strength
due to the internal friction, is not considered in fiber element modeling approach. Another
limitation of the fiber model is that the plane cross sections remain plane which may not be
true at all times, especially when significant amount of shear deformations and shear cracks
are encountered. In order to handle the effect of these deficiencies in the mathematical
model, the two large-scale experiments are selected such that the flexural effects govern
the behavior of the slender RC walls and no shear failure is observed during the cyclic
loading of the experiments. After all, the fiber element modelling is not a flawless
simulation technique for the modeling of shear walls but it offers a reasonable approach for
reliable analysis and engineering practice.
During the numerical analysis of a structural model, concrete infill walls can either be
modeled by the shear wall element, which consists of a vertical fiber and a concrete shear
layer, or the general wall element that is composed of all components including the
diagonal compression layers as illustrated in Fig. 5. Vertical fiber layers are used to model
the bending/axial behavior, while concrete shear layers are employed for defining the
contribution of concrete to the sectional shear strength (conventional shear behavior).
Diagonal layers in general wall element, transmit shear force by considering the contri-
bution of reinforcing steel to the shear strength through the interaction with the axial-

123
Bull Earthquake Eng

Fig. 5 Parallel layers to be employed for analytical modelling (Jiang and Liu 2011)

Fig. 6 Main aspect of the inelastic behavior (taken from Perform-3D user manual)

bending layers, however it may overestimate the shear strength of a wall. Connecting by
the nodes, all layers are interacting with each other and the neutral axis shift can be
simulated since the cross-section is based on fiber element layer (Perform 3D Components
& Elements 2006). Therefore, the mathematical model of RC infill walls and frame ele-
ments are represented by using the general wall element to consider both the vertical and
horizontal fiber layers. However, in order to avoid the above mentioned disadvantage of
diagonal layers, only the concrete shear layer (conventional shear layer) is used for the
shear behavior of elements.
The skeleton curve of the material constitutive model shown in Fig. 6 is defined by the
points Y, U, L and R, representing the initial stiffness, strain hardening, ultimate strength
and strength loss, respectively.

123
Bull Earthquake Eng

Fig. 7 Energy degradation factor: a EDF = 1.00 and b EDF \ 1.0 (taken from Perform-3D user manual)

Fig. 8 Unloading stiffness


factor (taken from Perform-3D
user manual)

For the implementation of hysteretic rules, the main aim is to obtain the dissipated
energy, which is affected by the energy degradation factor (EDF) and the unloading
stiffness factor (USF). The method for establishing a relation between the hysteretic loop
and the EDF is based on the max–min deformation of the component and the corre-
sponding EDF. In this sense, EDF for the hysteretic loop as a whole is defined by Eq. (3),
where w is the weighting factor, emin and emax are the positive and the negative energy
degradation at extreme deformation of the component. Thus, EDF is defined as the areal
ratio of the degraded and the non-degraded hysteresis loop within a range from 0.00 to
?1.00, while the USF controls the unloading behavior of the hysteretic curve within a
range from ?1.00 (maximum stiffness-minimum elastic range) to -1.00 (minimum
stiffness - maximum elastic range) as shown in Figs. 7 and 8.
e ¼ wemin þ ð1  wÞemax ð3Þ

Precision of the mathematical model gets better by increasing the number of fiber and
the elements along the RC wall. (Orakcal et al. 2004) have conducted analytical research to
investigate the effect of the number of fiber and the elements in modeling of RC shear
walls by using the multiple-vertical-line-elements (MVLEM), which has similar concept
with the fiber modeling approach in the software used here (Wallace 2007). According to
this research, top displacement-base shear cyclic curve response does not significantly
change under different number of fibers and the elements. On the other hand, using coarse
mesh in mathematical model adversely affects the strain response. Therefore, fine mesh
with an adequate number of the fiber element is used for the numerical analyses of RC infill
wall. For this purpose, mathematical model of the RC infill walls are implemented into the
software by using at least 16 elements with 18 fibers per story, which is nearly same with

123
Bull Earthquake Eng

the proposed number of fibers by Orakcal et al. (2004). Thus, the sensitivity of analytic
response to section discretization is eliminated in terms of the EDF and the USF during the
nonlinear constitutive numerical analyses.
Connection between the existing frame and the RC infill was achieved by using epoxy
grouted steel dowels to anchor the inner face of the beam and the columns in each of the
numerically calibrated experiments. However analytical modelling of the interaction be-
tween the frame and the RC infill wall with dowels is a complex task and should be
investigated in details. Therefore, experiments, which have negligible deformations at the
interface of frame and the RC infill wall due to the conservative number of dowels, are
preferably selected herein and numerically calibrated. In this regard, the steel dowels and
their effect on the seismic response are not considered during the numerical calibration of
the experiments. If the applicable number of anchored dowels is limited, then a comparison
of the results of such experiments, the fiber modelling and one of the microscopic mod-
eling approaches, such as finite element model, is recommended for investigating the
transfer of shear stresses throughout the interface.

4 Experimental studies and their analytical modeling

It is aimed to propose an analytical model for RC infill walls subjected to cyclic loads,
which will reflect the real behavior better and also can be handily used during the nu-
merical analysis of strengthened systems by the addition of shear walls. Henceforth, two
well-known experimental studies are considered for the calibration purpose of the analy-
tical model to be proposed in this paper.

4.1 Analytical investigation of the first experiment

The first of the two experiments involving the wall specimens that is selected for numerical
investigation was conducted by Strepelias et al. (2012) to examine the cyclic behavior of
frames strengthened with RC infill walls. In this experiment, single bay, four story test
specimen, called SW1, were prepared with the scale of 3:4 and then converted to the RC
infill wall with a thickness equal to the frame members. The steel dowels, which were used
for connecting the frame and the infill RC wall, were anchored into frame members by
using epoxy grout. The construction phase of test specimen SW1 is illustrated in Fig. 9.
The height of the test specimen was 9000 mm with a 190 mm thickness and 2200 mm
width.
The concrete strength of the test specimen was approximately 27 MPa. During the
construction, plain bars denoted by ST1 were used for the stirrups, while deformed bars
denoted by B500c were preferred for the longitudinal and web reinforcement, as indicated
in Fig. 10.
Initially, the test specimen was subjected to the cyclic loading and after the damage had
been repaired, pseudo-dynamic loading was applied. Displacement controlled cyclic
loading was subjected to the test specimen with an inverted triangular distributed lateral
load and displacement of the top floor was adjusted to ±5, ±60 and ±70 mm, respectively
as exhibited in Fig. 11. An axial force of 465 kN, was applied to the test specimen
constantly during the test. Mathematical model calibration is performed for only the dis-
placement controlled loading step of the experiment and pseudo-dynamic loading phase is
not considered within the contents of this study to create a more simplistic loading for
consecutive analyses.

123
Bull Earthquake Eng

Fig. 9 Construction phase of test specimen SW1 (Strepelias 2012)

Fig. 10 Reinforcement pattern of test specimen SW1: a infill wall; b beam and c column (Strepelias 2012)

(a) (b) (c)


80
Top Displacement (mm)

60

40

20

- 20

- 40

- 60

- 80

Fig. 11 Geometry (a); experiment setup (b) and displacement controlled cyclic loading of test specimen
SW1 (c) (Strepelias 2012)

123
Bull Earthquake Eng

During the numerical simulation of the test specimen, columns, beams and RC infill
walls are modeled separately by using general wall element in the software Perform-3D but
as mentioned in the previous chapter, the diagonal layers are ignored in order to consider
the conventional shear behavior only. Within the model, stirrup reinforcement of the frame
and horizontal bars of the RC infill wall are defined into the fiber section in terms of the
area ratio which is described as the rebar area divided by the concrete area. The longi-
tudinal bars of the frame and RC infill wall element are implemented in their exact
locations for each fiber in the program. Rotation gage elements are used to compute the RC
infill wall’s sectional rotations in the analytic model. Since the gage elements introduced in
the mathematical model can only be defined between the nodes, the rotation gage element
is placed 630 mm above the base of the test specimen, which is nearly the midpoint of the
two experimentally measured locations, namely 450 and 900 mm. The non-buckling
inelastic steel material and inelastic 1D concrete material component is used for the
definition of reinforcement and concrete material, respectively. The skeleton of the stress–
strain curve for concrete in compression was represented by the curve that includes
strength loss and tensional strength, while the reinforcement constitutive model was im-
plemented into the program as a tri-linear curve without strength degradation as shown in
Fig. 12. The relationship between shear stress and strain for the RC infill wall is defined
with bilinear model by using Eq. (1) defined in Requirements for Design and Construction
of RC Structures (TS500, 2000) for cracking shear strength (Vcr).
Vcr ¼ 0:65fctk Ach ð4Þ

In this expression, cracking shear depends on the cross-section area of wall (Ach) and
characteristic tensile strength of concrete (fctk). Since the study for the test specimen is
focused on modeling flexural failure mode without shear distress of the web, the non-linear
shear strain behavior is not foreseen for the RC infill wall in the mathematical model.
Therefore, elastic-perfectly-plastic (e-p-p) model is selected to represent the hysteretic
behavior of the shear stress–strain relation. In this model, stiffness and strength range stay
constant during cyclic loading, as shown in Fig. 13.
The proposed numerical model configuration of the test specimen with an adequate
number of the section discretization and component is illustrated in Fig. 14. Due to the test
having been carried out in a horizontal position, self-weight of the system is ignored during
analyses, however a total 450 kN vertical axial load is applied to the top nodes of each
column, as similar to the test procedure. Displacement controlled cyclic loading com-
patible with the experiment is performed with inverted triangular distributed lateral loads,
which are defined at each node corresponding to the floor level of the mathematical model.
The calibration philosophy of the EDF and the USF is based on the comparison of the
experimentally measured and analytically calculated base shear-top displacement curve. In

(a) Concrete Constuve Model (b) Steel Constuve Model (c) Shear Stress - Strain Model
30 700
Shear Stress (N/mm2)

1.4
Stress (N/mm2)

25 600
1.2
Stress (N/mm2)

20
500
1
400
15 0.8
300
0.6
10
200
0.4
5
100
0.2
0 0 0
-0.002 0.001 0.004 0.007 0.01 0.013 0 0.05 0.1 0.15
-5 0 0.001 0.002 0.003
Strain Strain Shear Strain

Fig. 12 Implemented constitutive material models for: a concrete; b steel and c shear stress–strain curve
for the test specimen SW1

123
Bull Earthquake Eng

Fig. 13 Elastic-perfectly-plastic material model

30

25

20

15

10

0
-0.002 0.001 0.004 0.007 0.01 0.013
-5
S

700
600
500
400
300
200
100
0
0 0.05 0.1 0.15

Fig. 14 The established analytical model of the test specimen SW1

this sense, different sets of EDF and USF are implemented into the program and the
obtained results are investigated for each set, respectively. According to the results, the
EDF and USF are readjusted by considering the numerically obtained base shear-top
displacement curve. For instance, if the curve obtained from the analysis result does not
match with the measured one by means of elastic range, USF is recalibrated to reflect the
experiment behavior properly. As for the dissipated energy, which is the area under the
loop, EDF is recalibrated for the next step. Thus, after consecutive analyses of the nu-
merical model under reversed cyclic loading, the EDF and the USF for hysteretic rules
compatible with the experimental measurements are obtained. Since the EDF and USF
controls the different response properties of the hysteretic curve, the obtained calibrated
sets are unique and not possible to arrive at the same prediction with entirely different sets
of the EDF and USF. Calibrated EDF corresponding to the steel and concrete material
hysteretic model is given in Fig. 15. The calculated base shear-top displacement and base
moment-rotation curves obtained from the analytical modelling are compared with the

123
Bull Earthquake Eng

Calibrated EDF for Concrete Constuve Model Calibrated EDF for Steel Constuve Model
(a) (b)
0.75 0.7 0.35 0.1
U L X
U

Stress
Stress

1.0 1.0
Y Y
0.5 0.3
R X

Strain Strain

Fig. 15 Calibrated EDF for: a concrete and b steel constitutive models

2000
300
(a) 250 (b) 1500
200
Base Moment (kN-m)
1000
150
Base Shear (kN)

100 500
50
0
0
-50 -500
-100
Proposed Model -1000
-150 @450 mm (Measured)
-200 Measured
-1500 @630 mm (Calculated)
-250 @900 mm (Measured)
-2000
-300
-80 -60 -40 -20 0 20 40 60 80 -0.01 -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 0.01

Top Displacement (mm) Rotaon (rad)

Fig. 16 Comparison of experimentally measured and analytically calculated base shear-top displacement
variation (a) and base moment-rotation curve for SW1 specimen (b)

measured experimental results, which indicate that the calibrated analytic model has
successfully reflected the measured response of the test specimen by means of not only
force–deformation but also base-rotation relationship (Fig. 16). Since the maximum ob-
tained base shear from the calibrated mathematical model (Vmax = 236 kN) is smaller than
the calculated cracking shear strength (Vcr = 350 kN) of the wall, non-linear shear de-
formation has not occurred similar as the test specimen’s observed behavior.

4.2 Analytical investigation of the second experiment

The second experimental program, which is considered for the numerical investigation,
was performed over two consecutive steps by Canbay et al. (2003). Initially, reversed
cyclic loading was applied to a three-bay, two-story, 1:3 scale bare frames, which included
common deficiencies like low concrete strength, inadequate stiffness and insufficient
confinement in the building stock of Turkey. Later, the damaged frame was rehabilitated
by adding an RC infill wall at the central span of the specimen. Without repairing the
damage of the frame elements, a second reversed cyclic loading was applied to the
strengthened RC frame, named R1. Detailed dimensions of the test specimen, including the
frame element section properties are illustrated in Fig. 17. The specimen was constructed
approximately to have a 2500 mm total height and the RC infill wall was generated with a
70 mm thickness. Connection between the frame and the RC infill was achieved by using
epoxy grouted dowels to anchor the inner face of the beam and columns. The measured

123
Bull Earthquake Eng

Fig. 17 a Dimension details, b beam and c column reinforcement pattern of R1 specimen (Canbay 2001)

(a) (b) 60 52.6 53.1 52.5 51.6 53.2


49.6
45.4
40
40 34.9
30
25
Lateral Load (kN)

20 20.2
15 12.6
10.7

0 0

-9.8
-20 -15.4
-20.1
-24.9
-30.2
-40 -35.2
-40.1
-42.6
-45.2 -47.1
-49.7 -51.8
-52.7-53.2
-60
Cycles

Fig. 18 General view of the test setup (a) and cyclic loading pattern for R1 specimen (b) (Canbay et al.
2002)

concrete compressive strength was 13.8 MPa for the first story; 16.7 MPa for the second
story and 30.8 MPa for the infill wall. Therefore, concrete material models are imple-
mented into the model by considering the differences in the concrete compressive strength
for each floor level, respectively as in Fig. 20a.
Longitudinal and transverse reinforcements of the frames were detailed by plain bars
that have individual yield strengths of approximately 400 and 322 MPa, respectively. The
mesh reinforcement pattern of the RC infill wall was constituted with plain bars (2U6/150)
that have 378 MPa yield strength. The general view of the test setup for the frame with the
RC infill wall and the reversed cyclic loading, which was subjected to the test specimen at
the second story, is given in Fig. 18 (Canbay et al. 2002). Within the scope of this
numerical study, only simulation of the frame rehabilitated with the RC infill wall (R1) is
investigated to verify the calibrated analytic model parameters from the experiment (SW1)
achieved by Strepelias et al. (2012).
The mathematical model of the test specimen R1 is evaluated under the same analytic
model principle of the first calibrated specimen SW1. The general wall element, which
includes similar number of sectional discretization and components used in the analytical

123
Bull Earthquake Eng

35
30
25
20
15
10
5
0
0 0.003 0.006 0.009 0.012

600

500

400

300

200

100

0
0 0.05 0.1 0.15

Fig. 19 Mathematical model established for R1 test specimen

(a) Concrete Constuve Model (b) Steel Constuve Model (c) Shear Stress - Strain Model
35 600

Shear Stress (N/mm 2)


C13.8 1.4
Stress (N/mm2)

30
Stress (N/mm2)

C16.7 500
1.2
25
C30.8 400 1
20
300 0.8
15
0.6
10 200
0.4
5 100
0.2
0
0 0.003 0.006 0.009 0.012 0 0
0 0.05 0.1 0.15 0 0.001 0.002 0.003
Strain Strain Shear Strain

Fig. 20 Implemented constitutive material models in Perform-3D: a concrete; b steel and c shear stress–
strain curve for the test specimen R1

model of SW1, is adopted as the mathematical model of the R1 specimen (Fig. 19).
Although the stress–strain relationship for the concrete and reinforcement is adapted by
using a tri-linear curve similar to the SW1 mathematical model, the tensile strength of
concrete is ignored due to pre-damage of the frame elements (Fig. 20). Calibrated hys-
teretic rules for the steel and concrete constitutive models from the SW1 analyses are
implemented into the mathematical model of R1. Shear stress–strain curve is defined as a
bilinear model employing the cracking shear strength calculated as approximately
Vcr = 78.5 kN by Eq. (4). Since the test specimen R1 has also flexure-controlled failure
mode, the hysteretic behavior of the shear stress–strain curve is implemented using elastic-
perfectly-plastic material model.
Due to the bond/anchorage problem in the longitudinal reinforcement of the members,
capacity reduction, which is explained in details by Canbay et al. (2003), was applied to the
frame elements in the mathematical model. Other than the self-weight of the system, which
is calculated automatically by the software, a constant axial force (9 kN) corresponding to
the external weights shown in Fig. 18a is applied to the top nodal points of each column.
Cyclic loading compatible with the experiment is realized as a lateral load, which is
defined at the top node of the analytical model. The calculated base shear-top displacement
variation from the calibrated mathematical model shown in Fig. 21 is compared suc-
cessfully with the measured experimental results reflecting the measured behavior of the
test specimen under cyclic loading. Since the computed maximum base shear for the
mathematical model Vmax = 52 kN does not exceed the cracking shear strength of the
wall, nonlinear behavior due to shear force is not observed with the test results.

123
Bull Earthquake Eng

60.00

40.00

Base Shear (kN) 20.00

0.00

-20.00
Proposed Model
Measured
-40.00

-60.00
-40.00 -30.00 -20.00 -10.00 0.00 10.00 20.00 30.00 40.00
Top Displacement (mm)

Fig. 21 Comparison of the experimentally measured and analytically calculated base shear-top displace-
ment cyclic curve for R1 specimen

4.3 Comparison of the proposed and widely used hysteretic models

As the third step of the evaluation of the accuracy of the proposed mathematical modelling
for RC infill walls, each of the experimental structures are modelled employing commonly
used hysteretic models. Among the previously described models, Mander model for
concrete and the modified version of Menegotto-Pinto model for steel are selected for
comparison, considering their being the two most accepted and widely used constitutive
models during the analytical research and investigation of RC structures and structural
systems. Therefore, these two material models are selected and adapted to the configured
mathematical model of test specimens.
Concrete’s hysteretic behavior defined by Mander model is implemented into the nu-
merical modelling by using the adjusted energy degradation factor defined by Xuewei et al.
(2011), as given in Fig. 22a. The values in the parenthesis denote the calibrated proposed
model in the same figure. For the modified Menegotto-Pinto steel constitutive model, EDF
and USF are revised according to parameters (R0 = 20; a1 = 18.5; a2 = 0.15; a3 = 0.01;

(a) 0.90 0.7 (b)


(0.75) (0.7)
U L
1.0
Stress
Stress

(1.0)
Y 0.5 0.3
(0.5) (0.3)
R X
Menegoo-Pinto

Perform-3D
Strain
Strain

Fig. 22 Energy degradation factor corresponding to the Mander concrete model (a) and selected original
steel model and its implementation in Perform-3D (b)

123
Bull Earthquake Eng

300 60
(a) 250 (b)
200 40

150
100

Base Shear (kN)


Base Shear (kN)

20
50
0 0
-50
-100 -20
-150
-200 Proposed Model -40
Proposed Model
-250 Widely Used Hys. Model
Widely Used Hyst. Model
Measured (Exp. SW1)
-300 Measured (Exp. R1)
-80 -60 -40 -20 0 20 40 60 80 -60
-40 -30 -20 -10 0 10 20 30 40
Top Displacement (mm) Top Displacement (mm)

Fig. 23 Comparison of experimentally measured and analytically calculated base shear-top displacement
for the test specimens: a SW1 and b R1

a4 = 7) given by Filippou et al. (1983). The modified hysteretic model and the proposed
original steel constitutive model are compared in Fig. 22b. As it can be seen from the
figures, although the adopted mathematical model has to be linear due to the limitation of
the computer program, no significant difference appears between implemented hysteretic
model and original proposed model. Indeed the amount of energy dissipation under cyclic
loading is almost the same as the original proposed model. The other parameters, such as
section discretization, assigned element fiber type, loading phase, are likewise preserved in
the mathematical model.
Figure 23 shows the base shear-top displacement variations for both experiments’
analytically obtained results by employing Mander and Menegotto-Pinto models for
concrete and steel, respectively. The proposed model and the corresponding test mea-
surements are also superimposed into the graphs for comparative purposes. When the plots
are comprehensively inspected by means of the adequacy of encountering the cyclic re-
sponse of the mathematical models and the experimental results, significant divergence for
capturing the experimental pattern by utilizing the widely used analytical model is ob-
served compared to the proposed model, particularly for R1 specimen. Moreover, this
observation is also viable for the amount of the dissipated energy for the same specimen.
When SW1 test specimen is considered, the numerical differences seem to decrease,
however the proposed model matches the experimental measurement results much better.
One of the main reasons for the diversity observed for specimen R1 with three bays seems
to be the pinching in the cyclic behavior of the system due to the bond-slip effect caused by
the insufficient amount of development lengths in the beam-column connections of the
existing frame. On the other hand, this phenomenon is not clearly observed in SW1 since
the system is consisting of a single bay with adequate detailing of reinforcing bars. Another
reason depends on the differences in the variation in the unloading and reloading braches
of the cyclic behavior mostly based on low concrete compressive strength, which is
calibrated by introducing the energy degradation and unloading stiffness factors in the
proposed model.

5 Conclusion

The numerical simulation of two different well-known large-scale experiments, in which


RC frames rehabilitated with RC infill walls are constructed and response measurements

123
Bull Earthquake Eng

are carried out subjected to reversed cyclic loading, is carried out by using the software
Perform-3D. Columns, beams and RC infill walls are modeled with general wall element
based on fiber layers. The mathematical model configuration of RC infill walls in accor-
dance with the test results is also explained in details. Consecutive analyses indicated that
the developed fiber element model is successful in reflecting the nonlinear behavior of RC
infill walls subjected to cyclic loading. Indeed, with an adequate number of components
and section discretization, the general wall element without diagonal layers can simulate
the nonlinear behavior of RC infill walls sufficiently, as shown in the present study. Thus, a
realistic and a representative numerical model, which is compatible with the experimental
evaluation and also practical for extensive use, is obtained for numerical simulation of RC
infill walls.
Another major result of the numerical simulation is that, the obtained model including
the calibrated hysteretic rules is compatible with the experimentally measured results by
means of seismic behavior of not only undamaged single bay RC infill walls but also pre-
damaged multi-bay strengthened frames, which include structural deficiencies encountered
in many earthquake prone countries. On the other hand, analyses presented herein indicate
that the widely used hysteretic constitutive models for the numerical investigation of the
nonlinear seismic behavior of RC structures might occasionally be insufficient to capture
the behavior of existing poor quality frames retrofitted by RC infill wall.
Consequently, while the established fiber layered analytical model with proposed
calibrated hysteretic constitutive rules might need some refinements in the light of the
results of new verifications; it gives a more accurate result than the widely used hysteretic
material models for the specific two cases handled during this research. This is because the
cyclic behavior of the entire structural system is calibrated in terms of energy degradation
and unloading stiffness factors. Moreover, the usage of the macroscopic modeling based on
fiber elements approach is more practical and needs less computational efforts during the
applications for representing the nonlinear behavior of RC infill walls.
Based on the measurements realized at the interface of the existing frame and the RC
infill of the selected experiments, it is assumed during the analytical modelling that the
anchoring steel dowels are sufficient to transfer the stresses and preserve the bond so that
they are not included in the model. On the other hand, since the number of the investigated
experiments is limited in this numerical research, the further investigation should be un-
dertaken by the new numerical analyses to increase the credibility of the proposed model.

Acknowledgments This research was financially supported by the Scientific and Research Council of
Turkey (TUBITAK) with the Grant No. 2211.

References

Anil O, Altin S (2006) An experimental study on reinforced concrete partially infilled frames. Eng Struct
29(2007):449–460
Bhaumik L, Raychowdhury P (2012) Seismic response of nuclear reactor buildings incorporating nonlinear
soil-structure interaction. Proceedings of the15th WCEE, Lisbon
Canbay E (2001) Contribution of RC infills to the seismic behavior of structural systems. PhD Dissertation,
Middle East Technical University, Ankara
Canbay E, Ersoy U, Özcebe G (2002) Betonarme dolgu duvarların yapıların sismik davranışı üzerine
etkileri. Research Report No: INTAG 563, Scientific and Technical Research Council of Turkey-
TUBITAK, Ankara
Canbay E, Ersoy U, Ozcebe G (2003) Contribution of reinforced concrete infills to seismic behavior of
structural systems. ACI Struct J 100(5):637–643

123
Bull Earthquake Eng

Clough R, Johnston S (1966) Effect of stiffness degradation on earthquake ductility requirements. Pro-
ceedings of the 2nd Japan Earthquake Engineering Symposium, Tokyo
Perform 3D Components & Elements (2006) For Perform-3D and Perform-Collapse. Computer & Structures
Inc
Darwish MSE (2006) Numerical modeling of infill RC walls in seismic retrofit of RC frames. MSc Dis-
sertation, the American University, Cairo
Esquivel JC (1992) Influence of some parameters on the inelastic earthquake response using different
hysteretic models for reinforced concrete. Proceedings of the 10th WCEE, Rotterdam
Fardis MN, Schetakis A, Strepelias E (2013) RC buildings retrofitted by converting frame bays into RC
walls. Bull Earthq Eng 11:1541–1561
Filippou FC, Popov EG, Bertero VV (1983) Effects of bond deterioration on hysteretic behavior of rein-
forced concrete joints. EERC Report No. UCB/EERC-83/19, University of California, Berkeley
Guedes J, Pegon P, Pinto AV (1994) A Fibre/Timoshenko beam element in CASTEM, Special Publication
No. I.96.31, Joint Research Centre, Ispra, Italy
Jalali A, Dashti F (2010) Nonlinear behavior of reinforced concrete shear walls using macroscopic and
microscopic models. Eng Struct 32:2959–2968
Jiang H, Liu L (2011) Numerical analysis of RC shear walls under cyclic loading by Perform-3D. Adv Mater
Res 250:2253–2257
Kabeyasawa T, Shiohara H, Otani S, Aoyama H (1983) Analysis of the full scale seven storey reinforced
concrete test structure. J Fac Eng 37(2):431–478
Kent DC, Park R (1971) Flexural members with confined concrete. J Struct Div 97(7):1969–1990
Lizuka H, Sakai Y, Koketsu K (2011) Strong ground motions and damage conditions associated with
seismic stations in the February 2011 Christchurch, New Zealand Earthquake: Flexural members with
confined concrete. Seismol Res Lett 82(6):875–881
Mander JB, Priestley MJN, Park R (1988) Theoretical stress strain model for confined concrete. J Struct Div
114(8):1804–1826
Martinelli P, Filippou FC (2009) Simulation of the shaking table test of a seven-story shear wall building.
Earthq Eng Struct D 38:587–607
Menegotto M, Pinto E (1973) Method of analysis for cyclically loaded reinforced concrete plane frames
including changes in geometry and non-elastic behavior of elements under combined normal force and
bending. Proceedings of the IABSE Symposium, Lisbon
Orakcal K, Wallace JW, Conte JP (2004) Flexural modeling of reinforced concrete walls: model attributes.
ACI Struct J 13(7):688–698
Orakcal K, Massone LM, Wallace JW (2006) Analytical modeling of reinforced concrete walls for pre-
dicting flexural and coupled–shear-flexural responses. PEER Report 2006/07, University of California,
Los Angeles
Otani S (1981) Hysteresis models of reinforced concrete for earthquake response analysis. J Fac Eng
36(2):407–441
Scott BD, Park R, Priestley MJN (1982) Stress-strain behavior of concrete confined by overlapping hoops at
low and high strain rates. ACI Struct J 79(1):13–27
Higashi Y, Endo T, Okhubo M, Shimuzu, Y (1980) Experimental study on strengthening reinforced concrete
structures by adding shear walls. Proceedings of the 7th WCEE, İstanbul
Sivri M (2011) Nonlinear finite element analysis of reinforced concrete strengthening shear walls. PhD
Dissertation, Suleyman Demirel University, Isparta (in Turkish)
Spyrakos CC, Maniatakis CA, Smyrou E, Psycharis IN (2012) FRP Strengthened brick-infilled RC frames:
an approach for their proper consideration in design. Open Constr Build J 6:306–324
Strepelias T (2012) Retrofitting of existing frame structures through RC infilling: experimental and ana-
lytical Investigation. PhD Dissertation, University of Patras, Patras (in Greek)
Strepelias T, Fardis MN, Bousias S, Palios X, Biskinis D (2012) RC Frames infilled into RC walls for
seismic retrofitting: Design, experimental behavior and modeling. Report Series in Structural and
Earthquake Engineering, University of Patras, Patras
Strepelias E, Palios X, Bousias S, Fardis MN (2014) Experimental investigation of concrete frames infilled
with RC for seismic rehabilitation. J Struct Eng 140(1)
Takeda T, Sozen MA, Nielsen NN (1970) Reinforced concrete response to simulated earthquakes. J Struct
Eng 96(12):2557–2573
Taucer F F, Spacone E, Filippou F C (1991) A Fiber beam-column element for seismic response analysis of
reinforced concrete structures. UCB/EERC Report 91/17, Earthquake Engineering Research Center,
University of California, Los Angeles
Thermou GE, Papanikolaou VK, Kappas AJ (2011). Analytical model for predicting the response of old type
columns rehabilitated with concrete jacketing under reversed cyclic loading. Proceedings of the

123
Bull Earthquake Eng

Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering,


Greece
TS500 (2000) Requirements for Design and Construction of Reinforced Concrete Structures-Turkish
Standards Institution. Ministry of Public Works and Settlement, Ankara
Varum H, Dias FT, Marques P, Pinto AV, Bhatti AQ (2013) Performance evaluation of retrofitting strategies
for non-seismically designed RC buildings using steel braces. B Earthq Eng 11:1129–1156
Wallace JW (2007) Modelling issues for tall reinforced concrete core wall buildings. Struct Des Tall Spec
Build 16:615–632
Xiaolei H, Xuewei C, Cheang J, Guiniu M, Peifeng W (2008) Numerical analysis of cyclic loading test of
shear walls based on open SEES. Proceedings of the 14th WCEE, China
Xuewei C, Xiaolei H, Fan L, Shuang W (2011) Fiber element based elastic-plastic analysis procedure and
engineering application. Procedia Eng 14:1807–1815
Yang J, Wallace JW, Lu X (2013) Nonlinear analysis of reinforced concrete walls using new concrete model
in opensees. Proceedings of the 5th Conference on Measuring Technology and Mechatronics Au-
tomation, China
Yin XW, Jiang HJ, Lu WS, Lu XL (2012) Shaking table model test of a super high-rise building with CFT
frame and core wall. Proceedings of the 15th WCEE, Lisbon
Zhou Y, Ye M, Yang Z, Ou L (2011) Hysteresis analysis of RC columns based on fiber cross section model.
Adv Mater Res 255:3770–3774

123
View publication stats

You might also like