Professional Documents
Culture Documents
Finite Element Interface Modelling and Experimental Verification of Masonry Infilled RC Frame - Al Chaar - 2008
Finite Element Interface Modelling and Experimental Verification of Masonry Infilled RC Frame - Al Chaar - 2008
net/publication/284415010
CITATIONS READS
23 1,946
3 authors, including:
Ghassan Al-Chaar
US Army Corps of Engineers
39 PUBLICATIONS 688 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Ghassan Al-Chaar on 26 July 2017.
Masonry walls have been used as both load-bearing commercial finite element program is identified with similar
structural elements and architectural non-structural ele- capabilities and is tested for its capabilities.
ments in single- and multi-story buildings. Both reinforced
and unreinforced masonry partitions have been used, The finite element models identified in this study for
sometimes filling the space within structural frames and modeling of unreinforced concrete masonry infill walls
other times not bounded by any confining structure. In and R/C frames are (1) a cohesive interface model to
load-bearing walls of the latter type, ties and columns simulate the behavior of mortar joints between masonry
are generally used to increase the structural integrity. In units and the behavior of the frame/panel interface, and
cases where the masonry infill interacts with the bounding (2) a smeared crack finite element formulation to model
frame, that interaction must be accounted for in design concrete in R/C frames and masonry units. The interface
and evaluation of structures subjected to lateral loading model is used to analyze a simple combination of concrete
such as an earthquake. blocks and mortar joints, and it also can account for the
shearing, residual shear strength, and opening and closing
Although there has been much previous work to of joints under cyclic shear loads.
develop analytical models that realistically capture
the behavior characteristics of experimentally tested The behavior of infilled frames is briefly discussed
masonry-infilled R/C frame prototypes, many technical to determine important aspects that need to be considered
difficulties in analytical modeling remain unresolved. and accommodated by the FE modeling approaches. Then,
A simplified analytical model that captures the salient a brief review of the modeling approaches and models
characteristics of infilled masonry structures has not yet used specifically for masonry-infilled frames is presented.
been developed. Success toward that end will require From the successful experiences of other investigators, as-
understanding masonry-infilled frame behavior in much semblies and infilled frame samples for which laboratory
more detail than offered by a strut or beam model. The test and numerical results are available are selected. A
required understanding can be accomplished only through summary of the results of verification studies conducted at
experimental investigation followed by numerical simu- the constitutive model level using the selected commercial
lation and parametric studies. The correct approach will software program is presented. The models in the com-
facilitate the introduction and calibration of a simple yet mercial program were first used to analyze masonry prism
accurate model for infill walls. subassemblies for validation and possible calibration. Then
the program was used to develop a sophisticated model and
Recent research has shed new light on the behavior analysis of the behavior of two types of infill frames, with
of infill frames and has produced advanced analytic tools. two distinctive failure mechanisms, subjected to in-plane
Classical diagonal strut models have been subjected to lateral loading. The numerical results are compared with
more thorough evaluations with new experimental data, available experimental and other numerical results.
and various limit analysis methods have been developed
to account for the different load-resisting mechanisms of Current State of Knowledge
infilled frames. Sophisticated finite element models have
been developed to analyze the nonlinear behavior of infilled Behavior of Masonry-infilled Frames
frames in a detailed manner. This paper summarizes some
of those findings and developments, identifies numerical Subjected to In-plane Loads
models, and demonstrates their capabilities. Also, to facili-
tate the use of these models by researchers and designers, a The behavior of masonry-infilled steel and reinforced
concrete frames subjected to in-plane lateral loads has
been investigated by a number of researchers. Fiorato et
1
Structural Engineer, U.S. Army Engineer Research and al. (1970) tested 1/8-scale non-ductile R/C frames infilled
Development Center – Construction Engineering Labo- with brick masonry under both monotonically increasing
ratory, 2902 Newmark Dr., Champaign, IL 61822. and cyclic lateral loads. That work was followed by the
2
President, Bridge Engineering Solutions, Inc., 1706 studies of Klingner and Bertero (1976), Bertero and Brok-
Aralia Dr., Mt. Prospect, IL 60056. ken (1983), Zarnic and Tomazevic (1985), and Schmidt
3
President, Trilogy Consultants, Inc., 440 Huehl Rd., (1989). More recently, single-story reinforced concrete
Northbrook, IL 60062. frames with masonry infills were studied by Mehrabi et al.
FE Modeling Studies of Masonry-infilled The literature review strongly suggests that the most
Frames realistic approach to modeling masonry-infilled R/C frames
would combine both continuum and interface constitutive
Dhanasekar and Page (1986) and Liauw and Lo (1988) models: the continuum model captures the behavior of the
have used linear and nonlinear beam elements to model the reinforced concrete in frame and masonry units in infill, and
behavior of steel frames, and interface elements to model the interface model captures the behavior of mortar joints
the interaction between the infill and the frame. Dhanasekar between individual masonry units and between the infill and
and Page used a nonlinear orthotropic model to simulate the frame. Mehrabi and Shing (1997) have tested the validity
behavior of brick infills, and Liauw and Lo used a simple of this approach and the capability of related constitutive
smeared crack model to simulate the behavior of micro- models. The use of a plasticity-based total strain, rotating
concrete infills. Schmidt (1989) used smeared crack ele- smeared crack model with tension softening and shear re-
ments to model both reinforced concrete frames and brick tention for modeling the concrete continuum in concert with
infills. In these three analyses, the infill panels have been a combined Coulomb friction/tension cutoff/compression
modeled as a homogenous material before fracture and the cap interface model for masonry joints was recognized to
effects of mortar joints have been smeared out. be in good agreement with the requirements established by
the modeling approach [Lofti and Shing (1994); Mehrabi
Mehrabi et al. (1994) demonstrated through experi- and Shing (1997)]. In general, the models were shown to
mental and analytical studies that diagonal and horizontal be able to represent the important behavioral aspects of the
cracking within the infill and slip at cracked joints is the materials and elements used in infilled-frame structures.
dominant failure mechanism for R/C frames infilled with
unreinforced masonry. Although cracking, crushing, and Review of Commercial FE Programs
accumulation of damage occur also in the masonry units,
it is the degradation of shear resistance in cracked masonry The approach and constitutive models described above
joints that defines loss of lateral stiffness and resistance of have thus far been implemented mostly by researchers in
a masonry-infilled frame. It can be seen that smeared crack an academic setting. Those implementations have been
models have a deficiency in modeling unreinforced ma- developed using analytical software programs that are
sonry infills because they alone cannot realistically capture not necessarily available in the public domain and typi-
diagonal shear cracking and the shear sliding of cracked cally have limited scope and applicability. For effective
concrete or masonry mortar joints. Those deficiencies are modeling and simulation of infilled frames in the design
inherent in the kinematic constraints related to trying to and engineering community, a commercial finite element
account for cracking in a continuum (i.e., a homogeneous program must be employed. The advantages of commercial
material). Consequently, the use of smeared crack ele- modeling programs include ease of model construction,
ments alone will lead to non-conservative design results large element and model libraries, user-friendly input and
for unreinforced masonry infill. In order to realistically output formats, and integrated graphics capabilities. Four
account for the natural planes of weakness in unreinforced available FE programs capable of modeling concrete and
masonry infill, interface elements must be incorporated interfaces were selected for review. These were Ansys,
into the model. Adina, Abaqus, and Diana. The programs were reviewed
for their capabilities in modeling structural discontinuities,
A number of plasticity-based continuous interface such as mortar joints, in otherwise heterogeneous materi-
models have been developed to model the tension and als such as concrete and masonry. For the type of infilled
shear behavior of masonry mortar joints [Rots (1991); frames considered in this study, i.e., R/C frames infilled
Lotfi and Shing (1994); Lourenco (1996)]. These models with unreinforced concrete masonry (UCM) blocks, the
Cap Mode
Tension Mode
Intermediate Yield
surface Residual Yield Surface
A comparison between the major features and capa- Figure 2. Comparative Interface Model Yield
bilities of the Diana interface model and the comparative Criterion [Mehrabi and Shing (1997)].
Other Notations:
ν = Poisson’s ratio, Diana smeared crack model
E = Modulus of elasticity, psi, Diana smeared crack model
f cm′ = compressive strength of masonry mortar cubes, psi
f cu′ = compressive strength of masonry units, psi (with respect to net cross-sectional area)
f m′ = compressive strength of masonry prisms, psi (with respect to net cross-sectional area)
so = joint tensile strength, psi, comparative interface model
εu = strain at maximum strength for axial testing of masonry prisms (in./in.)
Table 2. Comparison between the Diana and Comparative Interface Models, Features, and Relevant Parameters.
Diana Model Comparative Interface Model
Model Feature
Shape Parameters Shape Parameters
Shear slipping Coulomb φ, co Hyperbola µ,s,r
Surface
Investigating Diana Interface Model Performance loading and under 100 psi (689 kPa) constant normal
pressure. A full shear loading cycle was applied with 1.2
The Diana interface model was investigated in terms in. (30 mm) shear displacement in both directions, a total
of some important behavioral modes. The model for this of 2.4 in. (61 mm) shear displacement. As for the shear
study is a single joint, 8 in. (203 mm) long, 3/8 in. (10 mm) stress, the joint reached a maximum stress of 123.8 psi
thick, and unit depth. This joint is placed between two (854 kPa), after which the shear resistance dropped with
elastic masonry blocks and is analyzed under tension and softening toward residual shear strength. The model was
direct shear displacement loading one at a time. The case able to trace the full cycle, and pass unloading and loading
of shear loading is analyzed first without any normal stress successfully. In return, after 1.7 in. (43 mm) cumulative
acting on the interface, then with normal pressure. A com- shear displacement, the shear strength reached the residual
bination of the Modified Newton-Raphson and the BFGS4 shear strength defined by the corresponding material pa-
secant solution methods was used for analysis at the mate- rameter (φr = 0.65).
rial and structural levels, respectively. The direct solver is
employed, and the convergence criterion is set for a small Verification and Comparison
value of the norm of the residual energy at each iteration.
The step size is varied for each run in order to find and fit To validate the performance of the Diana interface
the parameters to their best performance. Table 3 shows model, the model was used to simulate two direct shear
the parameters used for the Diana interface model for these laboratory tests reported by Mehrabi et al. (1994). For those
analyses. These parameters are adopted representing a bed tests, numerical results using the comparative interface
joint for clay brick masonry [Mehrabi et al. (1994)]. model were also available. The tests were performed on
two mortar bed joints—one between two concrete hollow
The first analysis was performed to check the tensile ca- blocks and the other between two solid concrete blocks.
pacity and after-peak behavior of the interface element. The Each joint was subjected to 100 psi (689 kPa) normal pres-
interface was subjected to incremental normal displacement. sure in Test 1, and 150 psi (1,034 kPa) normal pressure in
The interface element reached a maximum tensile stress of Test 2. Cyclic shear displacement loading was used.
38.2 psi (263 kPa) (compared with 40 psi (276 kPa) defined
by the input parameters), after which significant joint open- For numerical modeling of these tests, Mehrabi et al
ing and tension softening occurred. The second analysis was (1994) used the comparative interface model with parameters
carried out on the same element, this time subjected to shear reported in Table 4 for Tests 1 and 2. The Diana interface
displacement loading without normal pressure. The maxi- model parameters were selected to match those of the com-
mum shear strength was approximately 40 psi (276 kPa), parative interface model as much as possible. Both tests were
the same as defined for the model by cohesion factor. The modeled using the same material parameters because in the
residual shear became zero in absence of normal pressure. comparative models, those parameters defined differently be-
tween Tests 1 and 2 do not have equivalents in the Diana model
The third analysis was also carried out under shear and/or they have no influence in the analysis performed here
displacement loading, but with an applied constant normal (e.g., initial asperity angle). The parameters used for analysis
pressure of 100 psi (689 kPa). As expected, the normal with the Diana interface model are shown in Table 5. Some
compressive pressure increased the shear capacity, yielding of the parameters had to be specified independently because
a maximum shear stress of 123.8 psi (854 kPa), which is there was no equivalent for them in the comparative interface
comparable by the cohesion resistance (40 psi (276 kPa)) model. Parameters Ψi , σu , δ, f c′ , Cs , Gfc , and κp were taken
plus friction that resulted from 100 psi (689 kPa) normal either in the range recommended by the Diana manual or
pressure (0.79 × 100 = 79 psi (544 kPa)). The model suc- assigned values within a reasonable range. For example, the
cessfully followed the sharp drop after loss of cohesion compressive strength, f c′ , which defines the limit of the com-
and approached the residual shear resistance defined by pression cap on the yield surface for Diana interface model, is
model parameters. not reflected or used in the comparative interface model but
is assumed to be 1,500 psi 10.3 MPa, which falls within the
The last verification analysis was carried out on the customary range of masonry strength. The cohesion factor, c,
same interface element subjected to shear displacement was selected to be 40 psi (276 kPa)to approximate the cohesion
factor calculated for the comparative interface model, based
4
BFGS: Broyden-Fletcher-Goldfarb-Shanno algorithm. on the other input material parameters.
1 2.8E+4 3.5E+4 40 1.61 0.9 0.75 4 0.2 16.1 2 10 5 0.45 3E-4 3 15.7
Table 5. Material Parameters for Diana Interface Model in Comparison Analyses (1 psi = 0.00689 MPa,
1 in. = 25.4 mm)
Knn Kss ft GfI co σu GfII f c′
tgφi tgΨ tgφr δ Cs Gfc κp
(psi) (psi) (psi) (psi-in) (psi) (psi) (psi-in) (psi)
2.8E+4 3.5E+4 40 1.61 40 0.9 0.005 0.75 150 2.3 16.1 1500 1.0 2 0.006
Figure 3 shows the shear stress/shear displacement described previously because the two models will be used
curve generated by the simulation of Test 1 using the Di- together to model masonry assemblages. Therefore, the
ana interface model. Figure 4 shows the shear stress/shear selected Diana smeared crack model was tested for its
displacement curves generated by the laboratory test, and overall behavior.
the numerical simulation of Test 1 using the comparative
interface model. These simulation results clearly show A single block was tested under compression and
the capability of the Diana interface model to accurately tension. The material parameters, type of tension and
simulate the test and other numerical results. compression curves, and assumed for them are shown in
Table 6. Values for parameters such as E, ν, ft, and f c′ were
Investigating the Smeared Crack Model selected to represent a normal weight concrete material.
These parameters usually define only the strength limits
Diana finite element program has a wide range of and slope of the initial ascending branch of the stress/strain
continuum models for concrete and masonry materials. All curves. Parameters GfI, and Gfc, on the other hand, define the
of these models follow well established and verified formu- curvature and descending (softening) branch of the stress/
lations, so a full validation investigation is not necessary. strain curve. The initial values reflected in Table 6 for these
The purpose of this section is to demonstrate the capability parameters were estimated by the area under the expected
of the selected model, which is a plasticity-based, total- stress/strain curves. By varying the latter parameters, a
strain, rotating smeared crack model intended to provide limited calibration of the stress/strain curves was performed
an acceptable form of stress/strain relationship for concrete to provide numerical robustness to the model and also to
used in R/C frames and masonry blocks. Without loss of generate desirable shapes for stress-strain curves. The first
generality, a plane stress condition was considered in this objective is especially important for implementation of the
evaluation to match the two-dimensional interface model constitutive model in a large-scale model for analysis.
Shear Stress (psi)
Figure 3. Test 1 Simulated using Diana Interface Figure 4. Laboratory Result for Test 1 [Mehrabi et al.
Model (1 psi = 0.00689 MPa, 1 in. = 25.4 mm) (1994)]. (1 psi = 0.00689 MPa, 1 in. = 25.4 mm)
A four-node finite element representing a concrete solid and hollow blocks are assumed to be the same, and
block was considered for these analyses. The two- minor differences introduced by Mehrabi et al. in their
dimensional analysis focused on the fracture energy in calibration process were ignored. The tensile strength of
compression, Gfc, and its effect on the parabolic curve of concrete for which no test result was reported is assumed
the compressive failure. The model was subjected to in- to be 10% of the compressive strength. The fracture energy
creasing uniaxial compression in a series of analyses with in compression, Gfc , was increased to 22 psi-in. (3.85 MPa-
Gfc as the variable. mm) to provide strain at maximum strength and a shape of
descending branch (softening) in the normal stress/normal
Modeling of Masonry Prisms strain curve similar to that obtained from the test results.
Experimental Results
2500
2000
Normal Stress (psi)
1500
1000
0
0.000 0.001 0.002 0.003 0.004 0.005 0.006
Normal Strain (in/in)
Figure 6—Normal Stress/Normal Strain Curves for Hollow and Solid Block Masonry Prisms (1 psi = 0.00689
MPa, 1 in. = 25.4 mm)
This table includes the average compressive strength of in R/C infilled frame structures.
masonry units with respect to the net cross-sectional area
( f cu′ ), average compressive strength of masonry mortar Diana is here to model and analyze one frame with
tested on cubes ( f cm′ ), compressive strength of masonry a relatively weak infill to further calibrate the models for
obtaining agreement between analytical and experimental
prisms with respect to the net cross-sectional area ( f m′ ),
results. The failure of such specimen is expected to be gov-
and strain at maximum strength for masonry prisms (εu).
erned by shearing and slip along masonry bed joints. Then a
A sample stress contour for one of the prism tests is shown
second frame with a relatively strong infill will be analyzed
in Figure 5. The high stress concentration is evident near
with the calibrated models. The failure mechanism for that
the mortar joint between the top and the middle masonry
frame is expected to be governed by diagonal cracking of
units. The stresses at the ends of the prisms are also high
the infill and shear failure of columns. The goal is to verify
but attributed to the loading applied directly at both ends
that the models calibrated with the results of one test could
of the prism.
be applied to another frame.
Modeling and Analysis of
Experimental Results of Masonry-infilled
Masonry-Infilled R/C Frames
R/C Frames
The purpose of this section is to examine the capa-
The specimens tested by Mehrabi et al. (1994), as
bilities and to identify the limitations of the Diana finite
modeled in the current study, were half-scale frame physical
element program in simulating the behavior of masonry-
models representing the interior bay at the bottom story of a
infilled structures subjected to lateral loading and to estab-
prototype frame. The prototype frame was a six-story, three-
lish a framework for future modeling and analysis of these
bay moment resisting R/C frame with a 45 x 15 ft tributary
highly nonlinear structures. Two specimens among those
floor area at each story. Two types of frames were designed
tested by Mehrabi et al. (1994) are considered with distinc-
for the prototype structure with respect to the lateral loadings.
tively different load carrying and failure mechanisms. The
One was a “weak” frame, designed only for a strong wind
two specimens both have frames that are not designed for
load, and the other was a “strong” frame, designed to resist
high seismicity regions (referred to as weak frames) and
the equivalent static forces of strong seismic loading. For
therefore are susceptible to development of shear failure
infill panels, 4 x 4 x 8 in. (100 x 100 x 200 mm) (nominal)
in the frame columns—a failure that is of the most concern
Figure 7—A Sample Stress/Strain Curve for Axial Testing of Masonry Prisms
(1 psi = 0.00689 MPa and 1 in. = 25.4 mm)
P1
Figure 8—Geometry and Loading Details of Test Specimen [Mehrabi et al. (1994)] (1 in. = 25.4 mm)
hollow and solid concrete masonry blocks were used to loading. In this figure, P2 is 22 kips (97.9 kN), P3 is 11 kips
represent weak and strong infill panels, respectively. (48.8 kN), and d is equal to 16.5 in. (419 mm) Figure 9 shows
damage to the two specimens mapped after completion of
The accuracy and reliability of an analytical model the tests for weak and strong infill specimens.
for simulating the behavior of an infilled frame strongly
depend on the capability of the model to predict the load- Mehrabi et al. (1994) used finite element modeling to
carrying and failure mechanisms in addition to estimate analyze their test specimen. They used their experimental
strength and deformations. The specimens considered for results to calibrate the model and determine parameter
this study are one with a weak frame and weak infill, and values. A smeared-crack finite element formulation was
one with a weak frame and strong infill. These are common used to model concrete in the R/C frame and masonry
failures for frames with unreinforced masonry infill that are units; a nonlinear constitutive model was used for bond-
not designed for high seismicity in accordance with recent slip behavior between steel reinforcement and concrete,
design code revisions. and an interface model was used for the mortar joints.
For the specimen with strong infill, Mehrabi et al. used
Geometry and details of the selected specimens are interface elements at column ends to allow shear failure
shown in Figure 8. Each test specimen was subjected to con- of the columns that otherwise could not be modeled with
stant vertical loading and monotonically increasing lateral their smeared crack model for concrete.
Figure 9—Failure Pattern from Laboratory Tests on Weak Infill (Left) and Strong Infill (Right)
Modeling of Masonry-infilled R/C Frames to about 30 times those calibrated by their laboratory direct
shear test results. The need for adjustment was attributed to
The infilled frame with geometry and details shown the inaccuracy of interface normal and shear displacements
in Figure 8 was modeled with Diana. As before, two- in the elastic region of their laboratory test responses, which
dimensional, plain-stress, four-node elements with four was caused by deflections of the test fixture. They had
integration points were used to model the concrete in R/C differentiated among mortar bed joints, head joints, joints
frame and masonry units. Two-dimensional, four-node between wall and frame by introducing different interface
interface elements with two integration points were used to material parameters and thicknesses.
model the mortar bed joints, head joints, and joints between
the infill and the frame. The Diana rotating smeared crack Parameter Setting for Frame with Weak Infill
and interface element models were utilized as described
earlier in this paper. Reinforcement bars for the frame were For the first trial using Diana, material parameters
modeled using Diana elastic-hardening plastic, two-node were selected in agreement with those used by Mehrabi et
discrete bar elements. The loading plates were modeled al. (1994). The parameters for the bed joints were the same
with linear elastic, four-node plane-stress elements. The as those used for the prism analysis described previously
loading top beam was modeled with two-node beam ele- and reflected in Table 1. The use of the normal and shear
ments using linear-elastic material with steel properties stiffnesses shown in Table 1 resulted in divergence and lack
while the frame footing was modeled with four-node plane- of solution at initial stages of the analysis. This divergence
stress elements using linear-elastic material with concrete occurred with initiation of shear cracking and slipping
properties. The finite element mesh is shown in Figure 10 at interfaces. The numerical process for determining the
at initial stages of lateral loading. Each masonry unit was stress on the yield surface (i.e., return mapping) failed in
discretized into two elements with an aspect ratio of 1:1. tension-shear or compression-shear corner zones. Persistent
The loading scheme followed the one illustrated in Fig- efforts to prevent the divergence of the algorithm, both with
ure 20. The vertical load was first applied in increments to reduction of step sizes and tolerances in practical range and
its maximum and kept constant while lateral displacement the application of various available solution methods, did
loading was applied gradually. not resolve the issue. It was concluded that sharp corners,
especially at tension cutoff zone, must be repaired in the
Model Parameter Setting and Analysis yield surface of the interface model if this problem is to
be avoided. Mehrabi et al. (1994) used a hyperbolic yield
When analyzing infilled frames, Mehrabi et al. (1994) surface that avoided corners in the shear-tension zone, and
adjusted the normal and shear stiffnesses of mortar joints their model did not include a compression cap.
Table 10. Material Parameters for Head Joints, Hollow Block Infill, Diana Interface Model (1 psi = 0.00689 MPa
and 1 in. = 25.4 mm)
Parameter Width Knn Kss ft GfI co σu GfII f c′
tgφi tgΨ tgφr δ Cs Gfc κp
Set (in.) (psi) (psi) (psi) (psi-in) (psi) (psi) (psi-in) (psi)
Target
1.25 215,300 269,200 10 1.61 10 0.8 0.005 0.7 150 2.3 16.1 1,500 1.0 55 0.006
values
Actual
12.5 21,530 26,920 1 1.61 1 0.8 0.005 0.7 150 2.3 16.1 150 1.0 55 0.006
values used
Table 11. Material Parameters for Joints between Frame and Wall, Hollow Block Infill, Diana Interface Model
(1 psi = 0.00689 MPa and 1 in. = 25.4 mm)
Parameter Width Knn Kss ft GfI co σu GfII f c′
tgφi tgΨ tgφr δ Cs Gfc κp
Set (in.) (psi) (psi) (psi) (psi-in) (psi) (psi) (psi-in) (psi)
Target
1.4 215,300 269,200 20 1.61 20 0.8 0.005 0.7 150 2.3 16.1 1,500 1.0 55 0.006
values
Actual
14 21,530 26,920 2 1.61 2 0.8 0.005 0.7 150 2.3 16.1 150 1.0 55 0.006
values used
Figure 11—Deformed Shape and Stress Contour for Frame with Weak Infill, Analysis Using Diana
50
Lateral Load (kips)
40
30
20
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Lateral Displacement (in)
Figure 12—Experimental and Numerical (Diana) Lateral Load/Lateral Displacement Curves for Frame with Weak
Infill (1 kip = 4.45 kN and 1 in. = 25.4 mm)
signaling crushing at loaded corners under higher lateral of the masonry assembly made of solid blocks. The width
displacements. These results agree well with the experi- of the mortar joints had to be also adjusted to higher val-
mental and analytical results by Mehrabi et al. (1994) and ues. These changes are shown in Table 12. A thickness of
illustrate the capability of the models to predict the load- 3.625 in. (92 mm) was used for masonry units in the infill.
carrying and failure mechanisms of the R/C frame with The model was then analyzed with the same loading scheme
weak masonry infill. Figure 12 shows response curves used for the weak-infill frame.
obtained from this analysis compared with those from
the laboratory test. Stiffness, response trend (including Analysis Results for Frame with Strong Infill
initiation of nonlinear behavior due to mortar joint shear
cracking and separation), and lateral resistance obtained The analysis results indicated that the predicted behav-
numerically agree well with the experimental results. ior of the frame with strong infill using the abovementioned
models agreed well with the experimental behavior for
Parameter Setting for Frame with Strong Infill the initial portion of the response curve, i.e., for lateral
displacement smaller than 0.2 in. (5 mm). However, the
To examine the capability of the calibrated models numerical response did not flatten and the lateral resis-
to predict the behavior of another infilled frame with dif- tance continued to increase with increasing displacement
ferent characteristics, the Diana constructed and material far beyond the strength obtained in the experimental test
parameters calibrated for the weak-infill frame were used (see Figure 13). Reviewing the numerical results indicated
to analyze the frame with strong infill. The material param- that the shear failure at the top end of windward column,
eters for masonry units were changed to those of the solid expected to govern the failure of the infilled frame based
blocks (see Table 8). The only parameter in the interface on the experimental results, did not occur. Furthermore,
model needing adjustment was the compressive strength with stronger mortar joints in the strong infill, sliding of
(or the compression cap), f’c, to reflect the higher strength bed joints did not occur as it did in the weak-infill frame.
Table 12. Material Parameters for Mortar Joints in Solid Block Infill that are Different from Hollow Block Infill,
Diana Interface (1 psi = 0.00689 MPa and 1 in. = 25.4 mm)
Width of Bed Joints Width of Head Joints Width of Frame to f c′
Parameter Wall Joints
Set (in.) (in.) (psi)
(in.)
Target values 3.5 3 3.5 2,000
Actual values used 35 30 35 200
80.0
70.0
Lateral Load (kips)
60.0
50.0
40.0
30.0
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Lateral Displacement (in)
Figure 13—Experimental and Numerical (Diana) Lateral Load/Lateral Displacement Curves for Frame with Strong
Infill (1 kip = 4.45 kN and 1 in. = 25.4 mm)
Only some diagonal cracking and mortar joint separation In general, the Diana program performed well for
were observed. This behavior resulted in the development modeling of masonry-infilled R/C frames of various
of confined a diagonal compression strut in the infill, pro- characteristics. However, the capabilities and limitations
viding a much higher lateral strength than expected. This of the models should be recognized, and modifications or
problem was attributed to the inefficiency of the smeared improvements be applied for better performance.
crack model for concrete in the R/C frame for modeling
the shear failure. Conclusions
To overcome this modeling problem, a discontinuity This study identified a modeling approach that treats
in the form of an interface model was introduced at the the discontinuities in masonry by introducing interface
column ends. The purpose was to allow shear failure of the elements for the masonry joints. Furthermore, the study
columns along these interfaces, as was anticipated for the identified constitutive material models through an exten-
strong-infill frame. The material properties assigned to the sive literature review, and demonstrated their capabilities
interface element at the column ends were similar to those through verification analyses. It was concluded that an
of the interfaces in the masonry. However, some param- approach using both continuum and interface constitutive
eters, such as tensile strength, cohesion, and compressive models is most suitable for R/C frames infilled with UCM
strength, were adjusted to reflect properties of concrete in blocks. To facilitate the use of these models by researchers
the frame. To avoid problems with convergence, the inter- and designers, a commercial finite element program having
face thicknesses were increased here by the same method similar capabilities was identified. The constitutive mod-
as used for the interfaces in infill. els used by this program, Diana, were examined through
analysis of data available from earlier experimental physi-
With the above modification, the model was analyzed cal modeling tests to verify their strengths, weaknesses,
again. As can be seen in Figure 14, the response agreed capabilities and limitations.
well with the experimental results. Figure 14 shows the
deformed shape and displacement field contour at 0.795 in. Diana was first used to analyze masonry prisms made
(20 mm) lateral displacement. It also shows the shear fail- of hollow and solid blocks. Next, one frame with weak infill
ure of the top end of the windward column. A large part (hollow blocks) was modeled and analyzed, and further
of the lateral displacement has been absorbed by the shear calibration was performed to obtain agreement between
failure of the column and displacement of the upper right- the analytical and experimental results. The failure of the
portion of the frame and infill. Load-carrying and failure weak-infill frame was governed by shear cracking and slip
mechanisms of the modified model agree well with the along masonry bed joints. A second frame, with strong infill
experimental results. (solid blocks), was analyzed with the calibrated models.
The anticipated failure mechanism for this frame was di- results. Once calibrated, the models can be used reliably
agonal cracking in the infill and shear failure of windward for parametric study of the behavior of infilled frames
column at the top end. The goal was to investigate whether subjected to lateral loading.
the models calibrated with the results of one test could be
applied to another test as well. In order to obtain agreement References
in this second test, interface elements had to be introduced
into the columns to allow the shear expected failure. That Al-Chaar, G. K., “Evaluating Strength and Stiffness of
adjustment was necessary because the shear failure could Unreinforced Masonry Infill Structures,” ERDC/CERL
not be obtained using the smeared crack model applied to TR-02-1/ADA407072, Champaign, IL:Engineer Research
the concrete material in the frame. and Development Center–Construction Engineering Re-
search Laboratory, January, 2002.
The models in Diana showed good capabilities for
predicting load-carrying and failure mechanisms for such Al-Chaar, G. K., “Non-Ductile Behavior of Reinforced
complex structures as infilled frames. This capability is Concrete Frame with Masonry Infill Panels Subjected to In-
recognized to be the most crucial for reliable strength and Plane Loading,” USACERL Technical Manuscript 88/18.
ductility predictions. Among limitations of the models in Champaign, IL: U. S. Army Construction Engineering
Diana, the convergence problem with the numerical scheme Research Laboratory, 1988.
in return mapping on the yield surface for higher stiff-
nesses for interface elements is a significant impediment Bertero, V. V., and Brokken, S., “Infills in Seismic Resis-
to a robust and reliable solution. A short-term solution was tant Building,” Journal of Structural Engineering, ASCE,
implemented to overcome this problem, but is believed that U.S.A.: 109(6): pp. 1337-1361, 1983.
this numerical inefficiency in the models can be overcome
by improvements to the failure surface and return mapping Buonopane, S. G., and White, R. N., “Pseudodynamic
schemes of the interface constitutive model. For the case Testing of Masonry Infilled Reinforced Concrete Frame,”
of potential shear failure in columns, it was shown that the Journal of Structural Engineering, ASCE, U.S.A.: 125(6):
use of interface elements in the column ends is required to pp. 578-589, 1999.
compensate for inability of the smeared crack formulation
to model of shear failure. In any case, it is understood that Dhanasekar, M., and Page, A. W., “The Influence of Brick
for successful application of any FE program to infilled Masonry Infill Properties on the Behavior of Infilled
frames, the model parameters must be calibrated using ap- Frames,” Proceedings of the Institute of Civil Engineers,
propriate material-level and structural-level experimental UK: 81(2): pp. 593-606, 1986.