Engineering Structures: Sciencedirect

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Engineering Structures 181 (2019) 310–323

Contents lists available at ScienceDirect

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

Numerical modeling of a two story third-scale reinforced masonry shear wall T


building subjected to quasi-static lateral loading
Hassan Bedeir, Marwan Shedid , Hussein Okail, Osama Hamdy

Structural Engineering Department, Ain Shams University, Cairo, Egypt

ARTICLE INFO ABSTRACT

Keywords: This paper presents numerical modeling of a two-story third-scale reinforced concrete masonry building com-
Concrete masonry posed of walls in orthogonal directions and subjected to quasi-static cyclic lateral loading up to failure. OpenSees
Structural walls and Response-2000 were used to capture the behavior at the section level as well as that at the element and
Ductility system levels. The results of ten experimentally tested individual walls were used to verify the modeling ap-
System level
proach that was used to generate the building model. The models were able to accurately capture the response of
Numerical models
OpenSees
individual walls and that of the tested building. Force based elements were used to represent the walls and values
Response-2000 for the factors controlling the cyclic behavior of masonry and steel material models used were recommended.
Eccentric lateral loading The maximum error obtained for the strength of all wall models and for the building model was 4% and 12%
respectively. Numerical results showed that overall ductility of the building was higher than that of its con-
stituent walls especially under eccentric load which was consistent with experimental results. The torsional
response of flanged walls was found to be essential for the accurate representation of the building behavior
under eccentric lateral loading.

1. Introduction loading. Different elements typically have different ductility capacities


even if they are the same type of lateral load resisting system. For in-
The concepts of seismic design have been in continuous develop- stance, the ductility of reinforced masonry shear walls can differ based
ment over the course of the past century. Having a force based design on the compressive stress, the wall aspect ratio as well as reinforcement
philosophy and realizing that the earthquake is basically the applica- ratio and wall coupling effects [3].
tion of ground acceleration to the building, it was natural to assume A building is normally composed of elements differing in their di-
that the forces induced by the seismic actions should be directly pro- mensions and reinforcement; it may also be composed of elements
portional to the mass of the structure [1]. Realizing the effect of the classified under different lateral load resisting systems. Consequently, it
structure’s dynamic properties on the seismic induced forces lead to the is obvious that the entire building (System Level) will have a ductility
development of the most widely used seismic design tool, the Response capacity different than, yet dependent on, its constituent elements. To
Spectrum, early introduced by Biot in 1932 [2]. date, there is a scarcity of research relating the behavior of the system
Buildings were then designed to behave in an elastic manner under level to the component level.
seismic loading and it was not until the 1960s that a building was found A simple illustration for the difference between the ductility of the
to possess an inherent property, ductility, which allows it to deform system and that of its components is discussed by Priestly et al. [3].
past its elastic limit without significant loss of strength [1]. Current design codes would assign a reduction factor for this building
Current design codes account for the building ductility by reducing similar to that of the least ductile component not taking into account
the seismic forces calculated on the basis of an elastic analysis of the the capacity provided by the more ductile elements [3].
structure. The reduction factor used depends on the lateral load re- The response of a building composed of shear walls and subjected to
sisting system such as the more ductile a structural system becomes, the lateral loading is complex due to the fact that multiple components with
more the design seismic forces can be reduced. different characteristics are incorporated in the behavior of the overall
The assigned factor is based on the ductility of individual elements structure as the eccentricity of lateral loads and coupling between
comprising the lateral force resisting system (Component Level) ob- walls.
tained from experimental testing of single elements under cyclic This paper presents an approach for the numerical modeling of mid-


Corresponding author.
E-mail addresses: hassan.bedeir@eng.au.edu.eg (H. Bedeir), marwan.shedid@eng.asu.edu.eg (M. Shedid).

https://doi.org/10.1016/j.engstruct.2018.12.010
Received 4 May 2018; Received in revised form 10 October 2018; Accepted 3 December 2018
Available online 14 December 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

rise buildings having reinforced concrete masonry (RCM) walls as their with an average cylinder compressive strength of 21.8 MPa was used for
main gravity and lateral load resisting systems and subjected to lateral the walls. The results from the rebar tension tests showed an average
loading. The study is divided into two main phases; in the first phase, yield strength of 495 MPa and 534 MPa for the M10 and D4 bars re-
numerical models were developed for experimentally tested RCM walls spectively with a Young’s modulus of 200.6 GPa [4].
by Shedid et al. [4] and Siyam et al. [5]. The experimental results were All wall specimens failed in flexure with yielding of the outermost
used to calibrate the generated models for individual walls (Component reinforcement and vertical cracks forming at both toes leading to their
Level). crushing. Diagonal cracks were observed over the lower third of the
For this calibration to be carried out, investigation for factors af- walls. The crack patterns for walls W1 and W2 are shown in Fig. 1.
fecting the behavior of RCM was required such as the plastic hinge
length, the shear deformation of walls and the tension stiffening be-
havior of RCM. To account for plastic hinging at the bottom of the 2.2. Siyam et al. [5]
cantilever walls and accurately represent the strength degradation, a
modification of the wall element length was needed based on the ma- Siyam et al. [5] experimentally tested six 1/3 scale fully grouted,
sonry stress-strain model. Shear deformation was applied in- reinforced concrete masonry (RCM) walls. The tested walls were two
dependently to the wall model since the used element considered story walls with a total height of 2.16 m and the test matrix contained
flexure deformation only. The tensile behavior of masonry needed to be three rectangular walls, one flanged wall and two slab coupled walls as
accurately modeled since it contributed to the stiffness of the wall. Each shown in [5]. All these walls were replicas of the walls composing the
of these factors along with their effects on the model results are dis- building tested by Heerema et al. [6] and modeled in this study. The
cussed in details in this paper. reported average compressive strength of tested masonry prisms (f′m)
In the second phase, a two story 1/3 scale RCM building with two was 16.4 MPa and the average yield strength of vertical reinforcement
reinforced concrete (RC) slabs, experimentally tested by Heerema et al. was 495 MPa. All walls were subjected to quasi-static cyclic loading
[6], was modeled using the same approach as in the first phase and until the outermost reinforcing bar fractured.
results were then verified at the System Level. All wall models were All walls fell in the special reinforced masonry shear walls SFRS
developed using OpenSees and Response-2000 was used to extract the classification of the MSJC (2013) standards [7] and they all failed in
shear behavior of RCM walls as will be explained in the coming sec- flexure. Idealized ductility capacity was calculated for each wall and a
tions. significant variability was noticed. The results showed that walls clas-
sified within the same SFRS could have different ductility and drift
2. Phase I: individual walls capacities [5].

In the first phase of this study, a modeling approach is developed for


the simulation of individual RCM walls subjected to quasi-static cyclic 2.3. Numerical verification
lateral loading using OpenSees. The experimental results for walls
tested by Shedid et al. [4] and Siyam et al. [5] were used for this Initial models for the six walls tested by Shedid et al. [4] were de-
purpose. veloped and subjected to a monotonic pushover displacement con-
trolled lateral load. The load-displacement curves of the models were
2.1. Shedid et al. [4] verified and compared to the envelopes of the experimental push and
pull cycles. These models were then subjected to cyclic loading similar
Shedid et al. [4] tested six fully grouted RCM walls under reversed to that applied in the experimental test and the numerical load-dis-
displacement-controlled quasi-static cyclic loading up to 50% de- placement loops were subsequently compared to those experimentally
gradation in strength in order to obtain enough information on their recorded. The same model was verified with the 1/3 scale walls tested
post peak behavior. by Siyam et al. [5] as explained later.
A series of three story and two story high half scale walls were
constructed using half scale replicas of the full scale (20-cm) concrete
blocks. Tested walls had a length of 1.8 m and a height of 4.0 m and 2.4. Modeling approach and element type
2.66 m for three-story and two-story walls, respectively as shown in
Table 1 and Fig. 1. Two approaches can be generally used for modeling RCM walls,
The reported average prism compressive strength (f′m) of the namely, micro-modeling and macro-modeling [8]. In case of micro-
grouted masonry prisms was 16.5 MPa at a strain of 0.0016. Fine grout modeling, the wall is discretized into a finite number of elements re-
presenting its constituent materials and their interface. This modeling
Table 1 strategy is used for investigating the local behavior of elements, and is
Wall Details, Shedid et al. [4]. time consuming and sometimes impractical when the global responses
Specimen Wall Dimensions Vertical Horizontal Axial at the element level and system level are of interest.
(Length × Height) Reinforcementa Reinforcementb Stress On the other hand, in case of macro-modeling, the wall is re-
(MPa) presented as a single frame element with two material models; one
representing the behavior of grouted masonry and the other re-
W1 1802 × 3990 mm 19M10 1 D4 at 95mmc 1.09
W2 11M10 0.89
presenting the reinforcement. This modeling technique requires much
W3 11M10 0.89 less computational time and is adequate for investigating the global
W4 1802 × 2660 mm 19M10 2 D4 at 95 mm 1.05
behavior of elements. Hence; macro-modeling was the method of choice
W5 11M10 0.88 for this study and the research platform OpenSees [9] was used to de-
W6 11M10 0.88 velop these macro-models for the walls.
Wall specimen is represented as a 2D force based (FB) beam-column
a
Area of M10 bars = 100 mm2. element with distributed plasticity. The FB formulation does not require
b
Area of D4 bars = 25.4 mm2.
c meshing to capture the highly inelastic curvature distribution of the
Reinforcement in the first story, for the rest of the wall 1 D4 at 190 mm is
wall, i.e. the wall can be represented as a single element. Each element
used.
is then assigned fiber sections modeling the masonry and reinforcement
parts of the wall cross section [10].

311
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Fig. 1. Wall Reinforcement Details (Top), Crack Pattern for walls W1 (Bottom Left) and W2 (Bottom Right), Shedid et al. [4].

Fig. 2. Grouted Masonry Material Model (Concrete06) (Left), Steel Material Model (Steel02) (Right).

2.5. Material models tangent modulus and other constants that control the transition from
elastic to plastic zone [9]. Table 2 shows the initial values used for the
Several material models are provided by OpenSees for concrete and materials and element definition, some of these values were changed
steel. In this study, Thorenfeldt concrete material (Concrete06) was during model refinement as shown in forthcoming sections.
used to model the behavior of grouted masonry and Giuffre-Menegotto-
Pinto steel material (Steel02) was used to model reinforcement.
Concrete 06 is defined by the compressive and tensile strengths, strain 2.6. Pushover analysis
at compressive and tensile strengths and other factors that control the
post peak curve of concrete; these factors were calibrated to fit the Initial models were developed for each of the six walls tested by
stress strain curve of tested masonry prisms as shown in Fig. 2. Steel 02 Shedid et al. [4] and then subjected to a pushover displacement con-
is defined by the yield strength, initial elastic modulus, post-yield trolled top lateral load. The results of the analysis were verified against
the experimental push and pull envelopes of the cyclic loading. Fig. 3(a)

312
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Table 2
Initial Model Parameters.
Element Parameter Definition Initial Value Notes

*
Wall Element Lip Weight of the bottom integration point for the wall element 5% of wall height Five point Gauss-Lobatto integration

Concrete06 n Compressive shape factor 3.2 Calibrated to fit the prism Stress-Strain Curve
k Post-Peak compressive shape factor 0.98
f’c Compressive strength −16.5 (MPa) Based on tested prisms
e0 Strain at f’c −0.0016
fcr Tensile strength 1.72* (MPa) Grout modulus of rupture (ACI 530–05)
b Exponent of the tension stiffening curve 0.4* Recommended by OpenSees

Steel 02 Fy Yield strength 495 (MPa) Based on experimental testing of bars


E Initial elastic tangent 200.6 (GPa)
b Strain-hardening ratio 0.01
R0 Control the transition from elastic to plastic branches. 20* Recommended by OpenSees
cR1 0.925*
cR2 0.15*

* Modified with model refinement.

shows the results of the initial model for W1, it showed very rapid Gfc 0.8f c'
strength degradation after reaching its peak; this behavior was the same 20 = + c
0.6fc' Lip Ec (1)
in all six wall models. Therefore, refinement of the model relying
mainly on the following three factors was conducted to capture the where Gfc
is the fracture energy, Lip is the weight of the integration
actual behavior. point, εc is the strain at compressive strength and ε20 is the strain at
80% strength degradation (see Fig. 4). The drawback of that regular-
ization method, as explained by its developers, is the need for post
2.6.1. Localization of plasticity processing of the curvatures, strains and stresses obtained from the
The solution of force based elements is conducted through solving model since the used material model is altered. The model would give
for sectional forces at a finite number of integration points along the correct global results (displacements and rotations) but incorrect local
element length. Each integration point is defined by its location along results (curvatures and strains) [10].
the element and its weight, which is the portion it represents of the The essence of this modification method was changing the material
element length. In distributed inelasticity elements which undergo a model based on the integration method used. Hence; post-processing
softening behavior, plastic strains tend to be concentrated at the section was needed. In order to avoid such step, it was proposed in this study to
that first plasticizes (bottom integration point in this case). As the use the regularization technique in reverse; i.e., keeping ε20 constant,
weight of this integration point decreases the element shows more rapid hence preserving the material model, while altering Lip and changing
strength degradation since all plastic strains and curvatures concentrate the integration weights using Eq. (1). Hence; the integration weight of
in a small portion of the element while the rest of it behaves in an the first integration point was changed to fit the material model.
elastic manner. The weight of this integration point can be physically These parameters were quantified as follows for the concrete model
related to, while not exactly equal to, the plastic hinge length. This used (Fig. 2), ε20 = 0.0057, Gfc = 40 N/mm giving Lip ≅ 800 mm. The
phenomenon is referred to as localization of plasticity and explains the value of Gfc was calculated by computing the area under the stress
abrupt loss of strength [10]. strain curve of the tested masonry prisms, the obtained values were in
Many regularization techniques are proposed in literature to over- agreement with those proposed in the literature for grouted masonry.
come this phenomenon, the technique of choice for this study was the The initial integration method used was a five-point Gauss Lobatto;
one developed by Coleman and Spacone [11]. They proposed modifying the weight of the first integration point (Lip) in this method is 5% of the
the concrete stress-strain relationship to maintain constant fracture wall height which caused the localization of plasticity at the base and
energy after the initiation of strain-softening by modifying the des- the rapid strength degradation. A 7-point user defined integration
cending branch of the Kent and Park concrete stress-strain model method was employed to apply the regularization. The weight of the
(Fig. 4(a)) based on the weight of the integration point using Eq. (1). bottom integration point was set to 800 mm, another point was added
at 1600 mm with a weight of 800 mm so that the bottom part is based
on a two-point Gauss Lobatto integration while the rest of the wall was

Fig. 3. Pushover analysis results of initial wall model (a) and regularized wall model (b) for wall W1.

313
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Fig. 4. Kent and Park concrete stress-strain model [11] (a); Schematic diagrams for the initial wall model (b) and regularized wall model (c).

divided based on the five-point Gauss Lobatto integration as shown in


Fig. 4.
The regularization greatly enhanced the wall results by solving the
rapid loss of strength, yet the model was still far from its experimental
counterpart and needed more investigation. Fig. 3(b) shows the result
of the application of the regularization on wall W1 of Shedid et al. [4]
experiment.

2.6.2. Tension stiffening Fig. 5. Calibration of grouted masonry tensile branch (ρ = 1.17%).
The regularized model, though considerably enhanced the results,
showed a stiffer behavior than that experimentally recorded. The ten- results of W1 is shown in Fig. 6(a). It is evident that the calibration of
sile behavior of grouted masonry was one of the aspects believed to this parameter had its effect on decreasing the stiffness of the wall
account for such discrepancy and needed further investigation. before the ultimate load, but had no effect on the descending branch of
As previously stated, the tension branch in the chosen concrete the wall load-displacement curve.
model (Concrete06) is defined by the tensile strength, strain at tensile Although the modification of the tensile behavior of masonry to
strength and a factor (b) that controls the post peak behavior. The ACI account for reinforcement enhanced the results, the model was still
530-05 [12] recommends a value of 1.72 MPa for the grout modulus of slightly stiffer than experimental results. It was evident that shear de-
rupture, this was used as an initial value for the tensile strength, and a formation component was not adequately modeled and needed further
value of 0.4 was assigned to the factor (b) as recommended by litera- investigation.
ture.
Sokolov [13] studied the effect of reinforcement on the tensile be-
havior of concrete. The results of his study related the tensile behavior 2.6.3. Shear deformation
of concrete to the reinforcement ratio using Eq. (2): The FB element used to represent the walls only accounts for flex-
ural deformation [9]. Therefore the final refinement of the model was
Ec ct then to take into account the shear deformation. Initially an elastic
shear force-shear strain relationship was applied to the wall model. This
ct cr , EC 2
ct =
ft , EC 2 1 0.27ln ( ct
cr , EC 2 ) 0.21 ct > cr , EC 2
(2) enhanced the elastic part of the wall load-displacement curve, however;
as the wall started going into the inelastic zone the effect diminished.
where ct is the tensile strength of concrete considering the effect of Many modeling techniques presented in literature propose applying
reinforcement, Ec is the concrete modulus of elasticity, is the re- shear deformations to cantilever walls by introducing an inelastic shear
inforcement ratio (%), ft , EC 2 and cr , EC 2 are the concrete tensile strength spring at a fraction of the wall height referred to as the wall center of
and cracking strain respectively according to EuroCode 2. rotation. In their attempt to evaluate the location of the center of ro-
This equation was used in this study to modify the concrete model tation using experimental results, various researchers suggested it was
tension branch. Concrete tensile strength, ft = 1.72 MPa was used in- located at a 40% of the wall height measured from its bottom [10]. It
stead of ft , EC 2 , and cr = ft / Em was used instead of cr , EC 2 . The results of was apparent then that inelastic shear behavior needed to be taken into
this calibration are shown in Table 3 and Fig. 5 and its effect on the account.
In this study, two steps were followed to define the inelastic shear
Table 3 behavior for each wall; the first was to define the location of the center
Modified Tension Branch Factors.
of rotation, the second was to extract the shear-shear strain relation for
Wall ρ (%) σct (MPa)a ɛcrb b the wall section at this location, idealize it into a bilinear relation and
apply it to the bottom part of the wall below the center of rotation. This
W1, W4 1.17 1.3 8.65E−05 0.86
W2, W3, W5, W6 0.55 1.52 1E−04 0.67
aimed at better representing the shear behavior by spreading it over the
height rather than concentrating it at a certain location by a spring.
a
Modified masonry tensile strength. For the first step, the full wall was modeled using Response-2000
b
ɛcr = σct/Em. [14] and subjected to the same loading conditions as the experiment;

314
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Fig. 6. Pushover analysis results for W1 after calibrating tension stiffening (a) and shear deformations (b).

the shear strain distribution over the wall height could be extracted for from all six models for yield load, maximum load, deformation at
each loading step until reaching the initiation of plastic shear de- maximum load and deformation at 20% strength degradation were
formation as shown in Fig. 7(a). 9.3%, 3.2%, 3.2% and 8.3% respectively.
It can be seen that the first section to undergo inelastic shear de-
formation was at a height of around 40% of the wall height and at a 2.7. Cyclic analysis
lateral load of approximately 55% of the wall ultimate capacity for the
analyzed walls with aspect ratios varying from 1.5 to 2.2. This height is The pushover analysis set the basis for developing a sound wall
consistent with the one proposed by literature for the center of rotation model which captured the behavior under monotonic lateral loading
as previously mentioned. The reduction in the shear stiffness of the wall accurately. The model was then calibrated to capture the wall behavior
is attributed to excessive cracking after yielding of vertical reinforce- under cyclic lateral loading. The models of all six walls were subjected
ment. to quasi-static cyclic lateral loading similar to that experimentally
For the second step, only the section at the center of rotation was performed. The results of the initial models were characterized by
modeled with Response-2000. Incremental shear force and bending wider loops than the experimental results and nearly no strength de-
moment were applied to the section depending on its location. Finally gradation as shown in Fig. 9 for W1. Studying the behavior of both
the shear-shear strain relation for this section was extracted and idea- masonry and steel components at each loading cycle and comparing
lized to a bilinear relation as shown in Fig. 7(b). them to the experimental observations, it was found that some para-
The bilinear shear-shear strain relation was applied to all integra- meters needed additional calibration.
tion points below the center of rotation using section aggregator in
OpenSees. This final step matched the model results to the experimental 2.7.1. Materials cyclic parameters
ones. The result of the application of shear deformation to wall W1 is The concrete model used (Concrete06) defines the cyclic behavior of
shown in Fig. 6(b). The effect of modifying shear behavior was mainly the material by two factors, α1 and α2 for behavior in compression and
on the wall stiffness before the ultimate load not on the descending tension, respectively. Initial values used for these factors were
branch since all walls were flexure dominated. α1 = 0.32, α2 = 0.08 as suggested in OpenSees manual [9], they were
then calibrated to enhance the results (Table 5). In addition, it was
2.6.4. Results of pushover analysis found that a crushing strain lesser than that used in the pushover
The discussed modeling method was applied to all six wall speci- analysis and an upper bound for the tensile strain were required to
mens, the results of such application are shown in Fig. 8 and table 4. better represent the experimental loops. The crushing strain and max-
The modeled wall specimens had aspect ratios of 1.5 and 2.2, and three imum tensile strain were applied using the “MinMax” material avail-
different cross sectional configurations, rectangular, flanged and end able in the OpenSees library.
confined walls. The modeling method had consistent effects on all wall For the steel model (Steel02), its cyclic behavior is defined by three
specimens which made it more reliable. The maximum error obtained factors, R0, cR1 and cR2. The initial values used as recommended by

Fig. 7. Shear Strain Distribution over W1 Height (Response-2000) (a), Bilinear Idealization of Shear Behaviour for W1 at a height of 1.8 m (b).

315
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Fig. 8. Modeling Results for Wall Specimens (W1–W6).

OpenSees [9] were R0 = 20, cR1 = 0.925, cR2 = 0.15. These factors having a percentage of the bar area; one of them is assigned the
were calibrated to modify the Pinching behavior of steel, hence; miti- buckling strain while the other is assigned only a fracture strain. The
gated the wide loops seen in the initial results. optimum area for the part assigned a buckling strain was 25–30% of the
Buckling and fracture strain limits were assigned to the steel model bar area based on a sensitivity analysis conducted. Table 5 shows the
using the “MinMax” material. By definition, the “MinMax” material final values used for the steel parameters.
defined in OpenSees applies a strain limit after which the material loses It is worth noting that two parameters highly affected the behavior
all strength. The buckling strain is assigned as a compressive of the walls. The first parameter is “α2” that controls the degradation of
(minimum) strain limit after which the steel is assumed to lose all of its the tensile branch in Concrete06 under cyclic loading. Using a relatively
resistance. When this was first applied it was found that the model loses large value for the parameter “α2” retained the resistance of
strength faster than the experimental results. It was hypothesized that Concrete06 in tension for more cycles leading to the hysteresis loops
this is due to the fact that when a bar buckles it retains a fraction of its being wider and stiffer. Hence; a smaller value than the initially re-
strength, opposed to fracture after which a bar has no contribution. commended (0.08) was needed for that parameter. Fig. 10 shows the
Hence; it was decided that a bar would be modeled as two fibers each effect of different values of “α2” on the cycle at maximum load for W1,

316
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Table 4 modeling of single cantilever RCM shear walls under cyclic lateral
Comparison Between Experimental and Numerical Results of pushover analysis loading. Macro-models representing the walls as FB elements were
(Walls tested by Shedid et al. [4]). found to be adequate and computationally efficient for capturing the
Specimen Pu (kN)a Δu (mm)b Δ0.8u (mm)c behavior of RCM walls. However; the FB element is sensitive to the
weight of the bottom integration point. A method was proposed to
W1 Experimental (Push) 177 25.1 5.3 regularize this weight to overcome localization of plasticity.
Experimental (Pull) 180 25.3 5.6
The effect of reinforcement on the tensile behavior of RCM was
Numerical 176 26 5.5
found to be essential for better prediction of the behavior. The FB
W2 Experimental (Push) 151 31.5 70
element considers flexure deformation only, hence; shear deformation
Experimental (Pull) 154 31.5 68
Numerical 155 31 67.5 should be separately aggregated into the element definition. The ma-
sonry behavior was represented accurately by a concrete stress-strain
W3 Experimental (Push) 152 36 93
Experimental (Pull) 147 36.1 95
model calibrated to match the grouted masonry prism stress-strain
Numerical 153 36 91.5 curve. However; this is true if the global behavior of the wall is the
W4 Experimental (Push) 265 14.1 27
point of interest. If local behavior of masonry constituents, i.e. masonry
Experimental (Pull) 267 12.5 28 block and grout, is the target of the model, they should be modeled
Numerical 263.5 16 30.5 separately.
W5 Experimental (Push) 245 14.6 42 Applying Cyclic loading to the wall models it was found that the
Experimental (Pull) 239 25.1 45 factors controlling the cyclic behavior of materials needed calibration
Numerical 235 20 42.5 to capture the actual wall response. Values were recommended for each
W6 Experimental (Push) 241 24.1 52 factor using a sensitivity analysis. Buckling and fracture of reinforce-
Experimental (Pull) 234 24 57 ment were modeled by applying strain limits over the steel material
Numerical 230 24 59 model.
a As previously stated, little research is available to date relating the
Lateral Load at Maximum Wall Resistance.
b
Top Displacement at Maximum Lateral Load.
behavior of the building to the behavior of individual walls. Phase II of
c
Top Displacement at 20% Strength Degradation. this study presents the modeling of a two story building experimentally
tested by Heerema et al. [6] under eccentric lateral loading. This model
the effect was consistent for all walls. Applying the buckling strain would serve as a prototype for a parametric study on numerical models
limit, which is the second parameter, mitigated the lack of strength of multiple buildings in an attempt to investigate the relation between
degradation post-peak in the initial model since the bars in compression the system and its components. The results of this parametric study are
lost a fraction of their resistance after buckling. All other factors in- presented by Bedeir et al. [15].
cluding those affecting the pinching of reinforcement had relatively
minor effects on the loops being less wide, but not on the wall re- 3. Phase II: building model
sistance.
The cyclic shear behavior of the walls was represented by a bilinear Through Phase I, a model was developed for the cyclic analysis of
SelfCentering material available in Opensees for the lower part of the single walls and was verified using the results of ten experimentally
walls, and an elastic material for the rest of the wall height. tested walls. The aim of phase II was to incorporate these wall models
into a single building model subjected to a quasi-static lateral load.
Heerema et al. [6], conducted an experimental study to investigate
2.7.2. Results of cyclic analysis the system behavior of an asymmetric two story 1/3-scale building with
Using the above mentioned modifications greatly enhanced the four walls aligned in the loading direction and four orthogonal walls.
model results and nearly matched the experimental results as shown in The walls used in the building were similar to the ones tested by Siyam
Fig. 11 for W1. et al. [5]. Torsional response of the building was permitted while wall
Four walls tested by Siyam et al. [5], were used to further verify the coupling by the slab was prevented. To avoid coupling, the RC slabs
modeling technique developed in this study, namely W1, W2, W5 and connecting the walls were designed with hinge lines in order to allow
W6. The models were subjected to the same load cycles as the experi- the slab to deform out of its plane, while not losing its diaphragm ac-
ment and the results of this simulation are presented in Fig. 12 and tion. The building was subjected to displacement-controlled quasi-static
Tables 6 and 7. cyclic lateral loading and cycled up to around 40% strength degrada-
Phase I gave some insight on the factors affecting the numerical tion.

Fig. 9. Initial Cyclic Behaviour of W1. Hysteresis Loops (a) Envelope (b).

317
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Table 5
Cyclic Parameters for Concrete06 and Steel02 Material Models.
Material Concrete06 Steel02

Parameter α1 α2 Crushing Maximum R0 cR1 cR2

Definition Parameter for compressive Parameter for tensile Strain Tensile Strain Parameters to control the transition Buckling Maximum Tensile
plastic strain definition plastic strain definition from elastic to plastic branches. Strain Strain
Value 0.03 1e−5 −0.004 0.015–0.025a 10 0.95 1 −0.005 0.1

a
Maximum values for flanged walls.

The building height was 2.16 m with two floors 1 m high each. Four The orthogonal walls were engaged in the building response due to
walls were aligned in the actuator direction, and four others were twist, this let the building maintain its lateral load carrying capacity at
aligned in the orthogonal direction to restrain the building twist. An displacement levels beyond its strength compared to the predicted re-
eccentric center of rigidity was achieved by choosing an asymmetric sponse (Fig. 13(a)). This showed the importance and effectiveness of
wall configuration; hence, twisting of the building was guaranteed. The orthogonal walls in enhancing the post peak response of the building
walls incorporated in this building were analogous to those tested by and it was captured by the numerical model as well.
Siyam et al. [5]. It is theoretically accepted that the response of the building under
The overall response of the building was compared to the response lateral load is the result of the superposition of its constituent walls in
of its constituent walls obtained from Siyam’s [5] experiment. As- the loading direction. In the tested building, it was found that the
suming that walls (W1, W2, W5, W8) were the only contributors to the building response was equal to the sum of walls W1, W2, W5 and W8
building response in the loading direction, and that there is no coupling until it reached the maximum load. However; the building showed
between W1 and W2; building response could be predicted by adding much lesser strength degradation than the expected by superposition.
the individual walls capacities corresponding to their displacement at This behavior was attributed to the presence of the orthogonal walls
each loading step of the building. The slab was designed and con- which came into action after the main lateral load resisting walls lost
structed with weak joints in order to avoid coupling between the walls their strength and the building started twisting significantly [6].
[6]. As previously stated, the overall ductility of a building is not
It can be seen from the experimental results, presented in Fig. 13(a), equivalent to that of its constituent elements. This is due to the fact that
that the system behavior nearly matches the predicted response up to the elements comprising the lateral load resistance of the building do
maximum load. Hence; it was deduced that walls W1, W2, W5, and W8 not lose strength with the same rate, giving the whole system an overall
were the main components of the Lateral Load Resisting System of the response different from that of the components.
building as designed. The building ductility was calculated based on idealized bilinear

Fig. 10. Effect of “α2” on the Cyclic Response of W1.

318
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Fig. 11. Refined Cyclic Behavior of W1. Hysteresis Loops (a) Envelope (b).

elasto-plastic response and it was found that the actual building duc- verification against its experimental counterpart are presented in this
tility was 30% higher than that calculated based on the superposition of section.
individual wall response. This is mainly due to the action of the or- For the initial model all walls were arranged as the experimental
thogonal walls which prevented the rapid loss of strength of the study and joined by a diaphragm at their top nodes using the
building at higher loads [6]. “rigidDiaphragm” constraint available in OpenSees [9]. The “ri-
gidDiaphragm” constraint links all nodes in a single plane however it
3.1. Verification of numerical model doesn’t take into account any coupling between linked walls, this was
consistent with the experimental program in which slab coupling of the
The individual wall models developed in the first phase were in- walls was prevented. The model was then subjected to the same cyclic
corporated into a 3D building model simulating the building experi- loading performed experimentally.
mentally tested by Heerema et al. [6]. The results of this model and its The results of the initial model, though showed the same trend of

Fig. 12. Envelope of Cyclic Loading (W1–W6) Numerical Vs. Experimental.

319
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Table 6
Comparison between Experimental and Numerical Wall Capacities.
Specimen Loading Direction Experimental Numerical

a b
Py (kN) Pu (kN) Py (kN)a Pu (kN)b

W1 Push 65.9 90.5 69 87


Pull 68 81.2 69.4 79

W2 Push 90.1 118.5 95.3 119


Pull 98.4 116 92 116

W5 Push 9.9 15.5 10.1 14


Pull 8.6 13.7 10 13

W6 Push 6.8 9.1 6.6 8.5


Pull 7.7 8.8 6.7 8

a
Lateral Load at First Yield.
b
Lateral Load at Maximum Wall Resistance.

Table 7
Comparison between Experimental and Numerical Wall Displacements and
Ductilities.
Specimen Loading Experimental Numerical
Direction
a
Δy (mm) Δ0.8u μep0.8u Δy (mm)a Δ0.8u μep0.8u
(mm)b (mm)b

W1 Push 6 35.2 4.6 6 45 4.82


Pull 6.4 29.1 4.3 45.2 4.22

W2 Push 6.8 40.5 4.7 7 42.2 4.88


Pull 7.1 30.1 3.9 34.4 3.97

W5 Push 9.5 54 4.1 9.5 60 4.5


Pull 10.1 56.4 3.9 50 3.88

W6 Push 13.8 59.7 3.2 14 79.3 4.62


Pull 15.7 78.6 5.2 70.2 4.39

a
Displacement at First Yield.
b
Displacement at 20% strength Degradation.

strength degradation, had 30% less strength than the experimental


(Fig. 13(b)). In order to spot the reason for that error, the displacement
of each wall within the model was extracted at each loading step and
compared to their experimental counterparts. It was found that W8
(Flanged wall) attained displacement levels approximately equal to
30% of the experimental, while walls W1 and W2 attained displacement
levels equal to 125% of the experimental. This showed that the center
of rigidity shifted towards W8 consequently applying less displacement
demand on it.
It was hypothesized that this error was due to ignoring the torsional
stiffness of the flanged wall (W8) in the numerical model. It was then
decided to apply the torsional stiffness as a bilinear curve similar to the
shear-shear strain relationship. A non-linear model for W8 was devel-
oped using SAP2000 in order to extract its elastic torsional stiffness,
since torsional stiffness is governed by the shear modulus (G); it was
decided to take the ratio between the inelastic and elastic torsional
stiffness equal to the ratio between the elastic and inelastic shear
stiffness extracted from Response-2000. The cyclic torsion behavior of
W8 was represented by a bilinear SelfCentering material. Applying this
parameter to the building model greatly enhanced the results; still the
model strength was approximately 12% less than the experimental re-
sults (Fig. 13(c)). Again, the displacement of each wall was extracted
and compared to the experimental and, nearly no error was found be- Fig. 13. (a) Predicted Building Response Vs. Experimental Results (Heerema
tween these displacements (Table 8). et al. [6]). (b) Envelope of Cyclic Loading for Building tested by Heerema et al.,
Since wall displacements of the numerical model were nearly Numerical vs. Experimental - Initial Model. (c) Envelope of Cyclic Loading for
identical to the experimental, it was evident that the model captured Building tested by Heerema et al., Numerical vs. Experimental – Refined Model
the real building behavior. In the experimental study by Heerema et al. (Torsion W8).
[6], each wall contribution to the total building resistance was calcu-
lated at each drift level. It was found that the sum of contributions of
the walls in the loading direction (W1, W2, W5 and W8) was about

320
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Table 8
Individual Wall Displacements in each Loading Cycle (Experimental Vs.
Numerical).
Building Wall Displacement (mm) (Average of Push and Pull Cycles)
Displacement
(mm) Experimental Numerical

W5 W8 W1, W5 W8 W1,
W2 W2

2.5 2.48 1.12 3.84 2.50 1.00 4.05


3.5 3.46 1.67 5.27 3.50 1.74 5.32
6.1 6.12 2.95 9.41 6.10 3.49 8.79
9.8 9.78 4.98 14.83 9.80 5.75 13.97
14.9 14.90 7.08 23.12 14.90 7.20 22.83
21.5 21.50 8.93 34.54 21.50 8.81 34.58
28.5 28.51 11.00 46.63 28.50 10.58 46.97
35.5 35.49 12.58 59.05 35.50 12.51 59.19
42.6 42.64 13.52 71.43 42.60 13.17 72.93
53.1 53.07 15.58 91.28 53.10 15.30 92.06

12–15% less than the building resistance as shown in Fig. 14(a). For
comparison, the results of the walls were extracted from the numerical
model and compared to the experimental building resistance
(Fig. 14(b)) [6].
The explanation of this phenomenon was that, although the slab
was designed in order to prevent coupling of the walls, it is hypothe-
sized that at higher drift levels orthogonal walls developed partial
coupling with the walls in the loading direction through the RC slab.
This effect would result in increasing the building resistance. This was Fig. 15. Hysteresis Loops for Building tested by Heerema et al., and cycle at
similar to the observations reported by Stavridis et al. [16] regarding maximum load (21.5 mm), Numerical vs. Experimental.
wall coupling [6]. This phenomenon is not incorporated in the Nu-
merical model and this is expected to account for the 12% difference in reinforcement. After yielding the stiffness degradation was more rapid
strength compared to the experimental results. reaching a loss of 80% at failure.
The building hysteresis loops were extracted from the numerical It is apparent from the results that the system response is sig-
model and compared to the experimental results. The numerical model nificantly different than its constituent components. This implies that
reached an average peak strength of 214 kN, (between push and pull relying on testing or modeling of individual components to evaluate
cycles), at 1% drift, compared to 244 kN experimentally determined at building behavior, in terms of strength and ductility, may not be ac-
the same drift level. Fig. 15 shows the hysteresis loops and the cycle at curate and would need refinement.
peak load. The response of each wall was extracted and compared to the This study presents a step into better modeling of structures which
experimental response. The results had a very close match as shown in takes into account the three dimensional interaction between the
Table 9. components leading to a realistic representation of the building re-
Finally, the stiffness degradation of each wall was captured and sponse. Additional redundancy in the system, provided by orthogonal
plotted against building drift, then compared to the experimentally walls, could be easily represented using this method. A further im-
measured stiffness. It was found that the wall stiffness degradation was provement for this approach would be taking the effect of the slab
almost identical to the experimental results. Fig. 16 shows these com- coupling into consideration, hence; making use of the extra ductility
parisons for the walls aligned along the loading direction. The walls lost provided by the slab while knowing which parts of it need special de-
about 20% of their stiffness due to cracking and up to yielding of tailing.

Fig. 14. Wall Contribution to the Building Resistance (a) Experimental (Heerema et al. [6]), (b) Numerical.

321
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

Table 9 reinforcement detailing could be enforced.


Individual Wall Reactions in each Loading Cycle (Experimental Vs. Numerical).
Building Wall Reaction (kN) (Average of Push and Pull Cycles) 4. Summary and conclusions
Displacement (mm)
Experimental Numerical This paper presents a modeling approach to capture the behavior of
a RCM shear wall building subjected to an eccentric Quasi-Static cyclic
W5 W8 W1, W5 W8 W1,
W2 W2
lateral load. Models of single RCM walls subjected to quasi-static cyclic
lateral load were developed using OpenSees research platform and
2.5 38.0 36.5 5.5 44.8 42.7 5.9 Response-2000 software and verified against their experimental coun-
3.5 46.0 43.5 6.7 51.8 48.6 7.0 terparts.
6.1 66.0 57.0 9.3 69.5 65.5 9.4
9.8 78.0 75.5 13.7 80.9 86.3 12.0
The modeling technique was applied to ten experimentally tested
14.9 85.0 95.5 14.5 81.4 96.7 13.5 wall specimens and a 3D building. The tested wall specimens covered a
21.5 82.0 104.0 14.5 81.9 105.1 13.4 wide range of wall aspect ratios and configurations which made the
28.5 70.0 109.0 13.2 72.0 110.2 12.6 technique more reliable. It was found that in order to accurately re-
35.5 60.5 113.0 12.1 59.2 109.7 10.4
present the lateral load behavior, wall models needed some refinement;
42.6 40.0 115.0 10.4 40.3 108.5 8.9
53.1 7.5 115.5 5.0 27.1 109.0 8.1 each refinement step had a consistent effect as follows:

• Regularizing the weight of the bottom integration point solved the


rapid strength degradation caused by localization of plasticity
• Modifying the tension model for masonry based on the reinforce-
ment ratio enhanced the initial stiffness of the model and better
captured the strength
• Applying elasto-plastic shear-shear strain relations to the walls
added the shear deformation component which resulted in a closer
match between numerical and experimental response

The modeling technique was proven adequate and computationally


efficient for capturing the cyclic behavior. Certain values were re-
commended for the factors controlling the cyclic behavior of masonry
and steel material models, and a method was proposed to model the
buckling of reinforcement. The maximum error obtained from all wall
models for yield load, maximum load and deformation at maximum
load were 9.3%, 3.9% and 8.8% respectively.
The results of an experimentally tested two story third scale
building were then used to verify that the technique is adequate for a
3D model. The macro-models developed for the individual walls were
incorporated into the full building model. All walls were linked using a
rigidDiaphragm constraint at their top joints to simulate the slab be-
havior.
The model results were compared to the experimental ones in terms
of full building response, individual wall responses and wall stiffness
degradation. It was found that the model is able to capture the building
behavior with high accuracy, however; it was not able to capture the
coupling between walls at higher drift levels, a phenomenon reported
by the experimental study.
An interesting conclusion was that the nonlinear torsional behavior
of flanged walls has a considerable effect on the behavior of a building
subjected to rotation under lateral load. A simple procedure to idealize
the nonlinear torsional behavior of the wall to a bilinear behavior was
presented and proven to yield accurate results.
This study presents a realistic approach for modeling of structures
under lateral loads, which takes into account the global behavior of the
building and the interaction between its components. It is the authors’
belief that following such approach and performing further studies on
the global behavior of structures could result in more efficient design
and could affect codified values for force reduction factors which, to
date, are based on the ductility of individual components.

References

Fig. 16. Stiffness Degradation for Walls along the Loading Direction.
[1] Calvi G, Priestley M, Kowalsky M. Displacement-based seismic design of structures.
National conference on earthquake engineering and engineering seismology,
Athens, Greece. 2008.
This approach can provide more efficient and economic design by
[2] Biot MA. Vibrations of buildings during earthquake. PhD thesis no. 259. Pasadena
considering the higher ductility capacity present in the building. In (CA): California Institute of Technology; 1932.
addition, the analysis would identify which parts of the structure are [3] Priestley M, Calvi G, Kowalsky M. Displacement-based seismic design of structures.
Pavia (Italy): IUSS Press; 2007.
the major contributors to such ductility and therefore extra attention in

322
H. Bedeir et al. Engineering Structures 181 (2019) 310–323

[4] Shedid M, El-Dakhakhni W, Drysdale R. Alternative strategies to enhance the [10] Calabrese A. Numerical issues in distributed inelasticity modelling of RC frame
seismic performance of reinforced concrete-block shear wall systems. J Struct Eng elements for seismic analysis. Pavia (Italy): University of Pavia; 2008. M.Sc. thesis.
ASCE 2010;136(6):676–89. [11] Coleman J, Spacone E. Localization issues in force-based frame elements. J. Struct.
[5] Siyam M, El-Dakhakhni W, Shedid M, Drysdale R. Seismic response evaluation of Eng. 2001;127(11):1257–65.
ductile reinforced concrete block structural walls. I: experimental results and force- [12] Building code requirements for masonry structures, (ACI 530-05/ASCE 5-05/TMS
based design parameters. J Perform Constr Facil ASCE 2016. 402-05).
[6] Heerema P, Shedid M, Konstantinidis D, El-Dakhakhni W. Behavior of asymmetric [13] Sokolov A. Tension stiffening model for reinforced concrete beams. Lithuania:
reinforced masonry shear wall building under quasi-static loading. J Struct Eng Vilnius Gediminas Technical University; 2010. Ph.D. thesis.
2014. [14] Bentz E, Collins M. Response 2000. Ontario (Canada): University of Toronto.
[7] Building Code Requirements and Specification for Masonry Structures. MSJC Available from: http://www.ecf.utoronto.ca/~bentz/r2k.htm, Version 1.0.5.
(Masonry Standards Joint Committee); 2013. February 16, 2015.
[8] Ezzeldin M, Wiebe L, Shedid M, El-Dakhakhni W. Numerical modelling of reinforced [15] Bedeir H. Assessment of the seismic performance of reinforced masonry and re-
concrete block structural walls under seismic loading. 9th International masonry inforced concrete shear wall buildings. Cairo (Egypt): Ain Shams University; 2016.
conference 2014 in Guimarães. 2014. M.Sc. thesis.
[9] McKenna F, Fenves GL, Scott MH. Open system for earthquake engineering simu- [16] Stavridis A, Ahmadi F, Mavros M, Koutromanos I, Hernández J, Rodríguez J, Shing
lation. Berkeley (CA): University of California; 2000. Available from: < http:// PB, Klingner RE. Shake-table tests of a 3-story, full-scale masonry wall system.
OpenSees.berkeley.edu > , Version 2.4.5. August 20, 2014. Dallas, Texas: ACI Masonry Seminar; 2011.

323

You might also like