EXPERIMENTAL AND ANALYTICAL STUDY

OF MASONRY INFILLED FRAMES

By Yaw-Jeng Chiou,1 Jyh-Cherng Tzeng,2 and Yuh-Wehn Liou3

ABSTRACT: The structural behavior of a framed masonry wall subjected to in-plane monotonic loading is
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.

investigated by a full-scale test and the method of discontinuous deformation analysis. The concept of artificial
joints is adopted to refine discontinuous deformation analysis so that it can analyze both continuous and dis-
continuous behavior of the masonry structure. The numerical solutions are compared with experimental results.
A satisfactory agreement is obtained. The structural behavior and stress distributions of the framed masonry
wall show that the characteristics of the masonry structure is highly influenced by the failure of mortar. In
addition, the filled masonry wall affects dominantly the behavior of the framed masonry structure. The partially
filled masonry wall induces a short column effect and leads to a severe failure of the column. On the other
hand, the completely filled masonry wall increases the stiffness of the structure and the adjacent column fails
in the configuration of nearly uniform cracks.

INTRODUCTION
Framed masonry walls are constructed with brittle materials
and the failure of the walls is frequently initiated from the
cracking of mortar and separation of bricks. Structure failure
induced by cracking and separation usually causes discontin-
uous and nonlinear behaviors.
The behavior of masonry structures had been extensively
studied. Smith (1966) examined the behavior of infilled frames
by the finite-difference method and adopted a simplified equiv-
alent single strut model to replace the wall. Since then the
equivalent strut model has been widely used by engineers and
researchers. Liauw (1972) proposed an equivalent frame
method to analyze the infilled frames. Thiruvengadam (1985)
used the finite-element method, the equivalent single strut
model, and the equivalent multiple strut model to study the
natural frequencies of infilled frames with an opening. Ach-
yutha et al. (1986) proposed an iterative finite-element method
of analysis to simulate the elastic behavior of infilled frames
with and without an opening. Ali and Page (1988) studied the
stress distributions and failure behavior of masonry structures
by the finite-element method. Alternatively, Papia (1988) pro-
posed a coupled finite-element and boundary-element method
to analyze the infilled frames. The analysis was carried out by
using the boundary-element method for the infill and properly
dividing the frame into the finite elements. Dawe and Seah
(1989a,b) investigated the behavior of masonry infilled steel
frames experimentally by using large-scale specimens and
scale models. They compared the experimental dynamic re-
sults with those of the three analytical models—the single-
degree-of-freedom model, the braced frame model, and the
equivalent strut model. They pointed out that the analytical
result of the equivalent strut model was found to be unsatis-
factory in predicting the dynamic response of masonry infilled
frames. Gulkan et al. (1990) and Clough et al. (1990) studied
the earthquake response of masonry structures by the method
of seismic testing. Abrams and Paulson (1991) presented a
similar study. Lotfi and Shing (1991) studied the masonry
1
Prof., Dept. of Civ. Engrg., Nat. Cheng-Kung Univ., Tainan, Taiwan
701.
2
Grad. Students, Dept. of Civ. Engrg., Nat. Cheng-Kung Univ., Tainan,
Taiwan 701.
3
Assoc. Prof., Dept. of Industrial Safety and Hygiene, Chia-Nan Col-
lege of Pharmacy and Sci., Tainan, Taiwan 717.
Note. Associate Editor: Walter H. Gerstle. Discussion open until March
1, 2000. To extend the closing date one month, a written request must
be filed with the ASCE Manager of Journals. The manuscript for this
paper was submitted for review and possible publication on July 31, 1998.
This paper is part of the Journal of Structural Engineering, Vol. 125,
No. 10, October, 1999. 䉷ASCE, ISSN 0733-9445/99/0010-1109–1117/
$8.00 ⫹ $.50 per page. Paper No. 18906.
JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999 / 1109
shear wall by incorporating the smeared crack models into the
finite-element method. On the basis of the finite-element
method and fracture mechanics techniques, El Haddad (1991)
studied the cracking and stress redistribution of infilled frames.
May and Naji (1991) developed a nonlinear finite-element
method to simulate the behavior of steel frames infilled with
concrete panels subjected to monotonic or cyclic loading.
Mehrabi (1994) evaluated the safety of masonry infilled RC
structures under earthquake loadings by using the half-scale
specimen experiment and finite-element method. Saneinejad
and Hobbs (1995) developed an inelastic analysis and design
method for infilled steel frames subjected to in-plane forces.
Haider (1996) studied the in-plane cyclic response of RC
frames with unreinforced masonry infill by a full-scale test and
the equivalent strut model.
The cracking and separation phenomena occurring in the
masonry structures cause distinct block elements. As a result
the masonry structures characterize discontinuous and nonlin-
ear behavior. The nature and orientation of discrete blocks play
an important role in the performance of masonry structures.
To model the masonry as discrete blocks by the classical nu-
merical methods (finite-difference method, finite-element
method, and boundary-element method) is not an easy task.
El Shabrawi and Verdel (1995) applied the distinct element
method (DEM) (Cundall and Strack 1979), to study the be-
havior of ancient masonry structures under dynamic loads. The
DEM has proved to be indispensable to many engineers in
approaching rock analysis. Because the materials used in frac-
tured rock masses and masonry buildings are similar in nature,
the use of the DEM is justified. However, the DEM employs
an explicit central difference time-marching scheme to inte-
grate the equation of motion directly. Because a central dif-
ference procedure is conditionally stable, the time-step size
must not exceed a critical value. As a result, the computation
time used in DEM is dramatically large, even for a simple
problem. Moreover, a mathematical damping is used in DEM
to dissipate the extra kinetic energy, and the current block
kinematics in DEM cannot handle complex contact situations
such as corner-corner contacts. Because of its incomplete
block kinematics and mathematical damping, the explicit
scheme used in DEM cannot guarantee the dynamic equilib-
rium state of a system at any time. An alternative method for
analyzing the block system is the method of discontinuous
deformation analysis (DDA) (Shi 1988). DDA is an implicit
method, and it possesses some unique features such as com-
plete block kinematics, perfect first-order displacement ap-
proximation, strict postulate of equilibrium, and correct energy
consumption. These features make DDA a rigorous analysis
for discrete blocks. This method has been adopted and ex-

J. Struct. Eng., 1999, 125(10): 1109-1117

 tended by some researchers (Ke 1993; Shyu 1993; Chang
1994; Lin 1995).
The structure behaviors of framed masonry walls subjected
to in-plane monotonic loading are investigated by a full-scale
test and DDA in this study. However, the masonry wall be-
haves as a continuous structure if the brick and mortar do not
crack. The blocks in the prototype of DDA are independent
and the block kinematics fulfills no interpenetration and no
tension between blocks at any time. The prototype of DDA
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.
pression reinforcements for beam and columns are taken to be
4-#7 steels. The height of the masonry wall of the partially
filled frame is 110 cm, and there is a wooden window in the
opening area.
FUNDAMENTALS OF DDA
In the method of DDA (Shi 1988), the variables are dis-
placements and the equilibrium equations are solved in the

can efficiently analyze the discontinuous structural behavior,
whereas it is less efficient for continuous structure. Chiou et
al. (1998) developed an interface element to modify DDA so
that it can analyze both continuous and discontinuous behav-
iors of masonry structure. The modified DDA was verified to
be capable of simulating the behavior of a masonry structure
and RC accurately. But its required computer storage is much
higher than that of the prototype of DDA. To increase the
computational efficiency, the concept of an artificial joint (Ke
1993) is adopted to refine DDA in this study. The artificial
joints improve the prototype of DDA; however, they are not
completely satisfactory. On the failure analysis of a block sys-
tem, the failure directions of blocks are predetermined, pro-
vided the failures of blocks are assumed to be along the arti-
ficial joints.
In the following, the experimental system and fundamentals
of the refined DDA are presented first. Then the structural
behaviors of framed masonry walls are fully studied by a full-
scale test and DDA. Finally, the justification of an equivalent
strut is assessed by the distribution of principal stresses in the
masonry wall.
same way as the finite-element method. The blocks in DDA
are independent, and so connections exist only when the
blocks are in contact with one another. DDA incorporates a
complete block kinematic that fulfills no interpenetration and
no tension between blocks at any time. This kinematical theory
provides efficient rules for solving simultaneous equilibrium
equations with opening-closing iterations within each step un-
der the constraints of no interpenetration and no tension. The
interactions between blocks are simulated by contact springs.
DDA also has an efficient scheme for contact detection. The
first entrance rule is used to delete inadmissible contacts and
to deal with corner-corner contacts. With this and opening-
closing iterations within each step, DDA computes the correct
contact force in equilibrium with time. The Mohr-Coulomb
law is used to regulate contact behavior, in which friction loss
along contacts is the sole source of energy consumption, with-
out adding any artificial damping. Therefore, correct energy
consumption is undertaken in DDA.
A complete first-order polynomial is chosen as the displace-
ment function for a 2D block, and this displacement function
restricts the block to constant stress. Referring to Fig. 3, the
displacements (u, v ) of any point (x, y) in a representative

再冎
Block i (Shi 1988) are given as
EXPERIMENTAL ARRANGEMENT
u0
A full-scale test is arranged to study the behaviors of framed

再冎 冋 册
v0
masonry walls and to verify the numerical solutions. The ex-
perimental system is shown in Fig. 1. The lateral force is cre- u 1 0 ⫺( y⫺y0) (x⫺x0) 0 ( y⫺y0)/2 r0
=
ated by the jack, and the magnitudes of force and lateral dis- v 0 1 (x⫺x0) 0 ( y⫺y0) (x⫺x0)/2 εx
placement are measured by the load cell and the clip-on gauge, εy
respectively. Three specimens: (1) an RC frame; (2) an RC ␥xy
frame partially filled with masonry wall; and (3) an RC frame (1a)
completely filled with masonry wall are studied and illustrated or
in Fig. 2. The dimension of the specimen is 320 ⫻ 300 cm.
The cross sections of the beam and column elements are 35
⫻ 40 cm and 30 ⫻ 35 cm, respectively. The tension and com- 再冎
u
v
= [Ti][Di] (1b)

FIG. 1. Experimental System

1110 / JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999

J. Struct. Eng., 1999, 125(10): 1109-1117

 ⭸2 ⌸
(Krs)ij = , r, s = 1, 2, . . . , 6 (3)
⭸dn ⭸dsj
⭸⌸(0)
(Fr)i = ⫺ , r = 1, 2, . . . , 6 (4)
⭸dn
where ⌸ = total potential energy. The potential energy includes
the contributions of inertia forces, initial stresses, internal
strains, point loading, body forces, contact spring deforma-
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.

tions, etc. The inclusion of inertia term makes the global stiff-
ness matrix positively and diagonally dominant, without nu-
merical hazard.
DDA is refined by the concept of artificial joints (Ke 1993)
in this study. Fig. 4 shows two blocks (Blocks i and j) just in
contact at time t = 0. In DDA, two half-angle-half-angle con-
tacts are identified at the start. Contact A is formed by the
contact of Vertex 1 of Block i with Vertex 3 of Block j, and
Contact B is formed by the contact of Vertex 2 of Block i with
Vertex 4 of Block j. To model the joint between Blocks i and
j, one normal spring and one shear spring are added to Contact
A and Contact B. At the end of iteration, if neither the Mohr-
Coulomb test nor a test for separation of Contacts A and B
reveals failure, the springs will retain operation unchanged at
the contacts so that these two blocks are permanently tied to-
gether. Therefore, as the stiffness of the springs is increased,

FIG. 2. Test Specimens: (a) RC Frame; (b) RC Frame Partially

Filled with Masonry Wall; (c) RC Frame Completely Filled with
Masonry Wall
FIG. 3. Schematic Configuration of Block

where (x0 , y0) = coordinates of block centroid; [Ti] = first-

order displacement function; [Di]T = (u0, v0, r0, εx , εy , ␥xy ) =
displacement vector of block i; (u0, v0) = rigid body transla-
tion; r0 = rigid body rotation; and (εx , εy , ␥xy ) = strain com-
ponents in a 2D geometry. By minimizing the total potential

冋 册再 冎 再 冎
energy, the equilibrium equations for n blocks (Shi 1988) are
K11 K12 K13 ⭈⭈⭈ K1n D1 F1
K21 K22 K23 ⭈⭈⭈ K2n D2 F2
K31 K32 K33 ⭈⭈⭈ K3n D3 = F3 (2)
⭈⭈⭈ ⭈⭈⭈ ⭈⭈⭈ ⭈⭈
⭈ ⭈⭈⭈ ⭈⭈⭈ ⭈⭈⭈
Kn 1 Kn 2 Kn 3 ⭈⭈⭈ Kn n Dn Fn
where Di = displacement vector of Block i; Kij = component
of stiffness matrix; Kii depends on the material modulus and
inertia effect of Block i; Kij (i ≠ j) depends on the contacts or
bolt connection between Blocks i and j; and Fi = force vector
of Block i. The coefficients for Ki j and Fi are written as FIG. 4. Two Blocks Just in Contact at Time t = 0

JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999 / 1111

J. Struct. Eng., 1999, 125(10): 1109-1117

 these two blocks increasingly behave like a continuous body;
very strong springs will force the sides of the block in contact
with each other to displace compatibility. The characteristic of
contact springs initiates the concept of artificial joints. A con-
tinuous body can be cut into a number of subblocks by arti-
ficial joints. These subblocks are connected with one another
at contacts by very strong springs. Also, they are used to assure
continuity across any artificial joint up to a certain limit. How-
ever, on the failure analysis of structure, if the structure failure
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.
is assumed to be along the boundaries of prescribed artificial
joints, the artificial joints become incipient with the finite
strength of the block material.
MODELING OF FRAMED MASONRY WALLS
Masonry walls are built by using brittle bricks and mortar.
Mortar is usually the weak plane of the masonry structure;
therefore, cracking is frequently initiated from there. The
cracking of the mortar and separation of the bricks usually
cause discontinuous and nonlinear behavior.
The failure modes of mortar are classified into two types:
Tensile failure and shear failure. The mixed mode failure of
mortar is neglected in this study. The characteristic of the ten-
sile failure is similar to the condition of no tension between
blocks in DDA, whereas the shear failure is similar to the
friction behavior between blocks. The masonry walls are cut
into subblocks by the artificial joints, which become incipient
with the finite strength of mortar. The bricks are simulated by
the subblocks and these subblocks are connected to one an-
other by contact springs. The stiffness of the contact springs
is proportional to the strength of mortar and has the dimension
of force per length. The strength of mortar is represented by
its resultant forces, which are determined by the effective
length times of either the tensile strength or the shear strength
of the mortar. The contact conditions of blocks determine the
contact length. The contact vertices share the contact length.
The shared length is the effective length and is taken to be
one-half of the contact length. Four situations for the forma-
tion of a contact pair between two blocks in contact are
sketched in Fig. 5.
1. Two half-angle-edge contacts in the same block: As
shown in Fig. 5(a), Vertices 1 and 2 of the top block are
in contact with Edge 34 of the bottom block. The contact
length L is the distance between Vertices 1 and 2.
FIG. 5. Four Possible Contact Pairs between Two Blocks: (a)
Two Half-Angle-Edge Contacts in Same Block; (b) Two Half-An-
gle-Edge Contacts in Different Blocks; (c) One Half-Angle-Edge
Contact and One Half-Angle-Half-Angle Contact; (d) Two Half-
Angle-Half-Angle Contacts
1112 / JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999
J. Struct. Eng., 1999, 125(10): 1109-1117
2. Two half-angle-edge contacts in different blocks: Fig.
5(b) shows that Vertex 1 of the top block is in contact
with Edge 34 of the bottom block, and Vertex 4 of the
bottom block is in contact with Edge 12 of the top block.
The contact length L is equal to the length of Line 14.
3. One half-angle-edge and one half-angle-half-angle con-
tacts: Referring to Fig. 5(c), Vertex 1 of the top block is
in contact with Edge 34 of the bottom block, and Vertex
2 of the top block is in contact with Vertex 4 of
the bottom block. The contact length L is the length of
Line 12.
4. Two half-angle-half-angle contacts: In Fig. 5(d), Vertices
1 and 2 of the top block are in contact with Vertices 3
and 4 of the bottom block. The contact length L is the
length of Line 12 (or Line 34)
The failure criteria of the mortar are then given as follows:
1. Tensile failure: The tensile strength of the mortar is as-
sumed to be ␴t , whereas the tensile normal stress of the
mortar is ␴. The tensile failure criterion of the mortar is
written as
␴ ⱖ ␴t (5)
Or expressed by the normal spring force as
d ⭈ kn ⱖ ␴t ⭈ l (6)
where kn = stiffness of the normal spring; d = opening
distance; and l = effective length. If the normal spring
force satisfies (6), then the mortar has tensile failure. In
that case, the mortar no longer provides bonding force.
Both the normal spring and the shear spring are thus
taken off.
2. Shear failure: The shear failure of mortar is assumed to
follow the Mohr-Coulomb criterion. The shear strength
of the mortar ␶f is written as
␶f = ␶0 ⫹␴n tan ␾ (7)
where ␶0 = cohesion of the mortar; ␾ = internal friction
angle, ␴n = normal stress; and the positive sign in (7) is
taken for the compressive ␴n . When the mortar is shear
failure, then its shear stress ␶ satisfies
␶ ⱖ ␶0 ⫾ ␴n tan ␾ (8)
Or expressed by the shear spring force as
s ⭈ ks ⱖ ␶0 ⭈ l ⫾ d ⭈ kn ⭈ tan ␾ (9)
where ks = stiffness of the shear spring; and s = sliding
distance. The shear spring is therefore taken off provided
the mortar has shear failure. The normal spring is also
taken off if the normal stress is tensile. However, the
normal spring is used as the penetration-resistant com-
ponent for the compressive normal stress.
The analysis of RC structures is similar to that of masonry
structures. The concrete structures are cut into subblocks by
artificial joints. The artificial joints become incipient with the
finite strength of concrete. Both the tensile failure and shear
failure of concrete blocks are investigated. The shear failure
of concrete is assumed to follow the Mohr-Coulomb criterion.
The reinforcements are assumed to perfectly bond with con-
crete, and they are modeled by the bolt elements in DDA. The
stress-strain relation of the reinforcement is assumed to be bi-
linear, and the plastic modulus of the reinforcement is chosen
to be Ep = 0.02Es , where Es is the elastic modulus of the
reinforcement.
 RESULTS AND DISCUSSION
The modified DDA is verified first by investigating the de-
flection of a tip-loaded cantilever beam. Referring to Fig. 6,
the length of the beam is L = 8 m, height H = 1 m, width B
= 1 m, and load P = 1 ton (9.80665 kN). The Young’s modulus
of the beam is E = 105 ton/m2 (980.665 MPa), Poisson’s ratio
␯ = 0.2, and the stiffness of the contact springs is kn = ks =
1.65 ⫻ 105 ton/m (1.62 ⫻ 106 N/mm). This beam is cut into
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.
128 rectangular subblocks, and 128 triangular subblocks, re-
spectively. Fig. 7 shows the deflection of the beam cut by
various artificial joints and the exact solution. It is found that
the deflection predicted by the model with rectangular sub-
blocks deviates from the exact solution, whereas the deflection
by the model with triangular subblocks agrees well with the
exact one. The reason for this finding is that a linear DDA
block can only undertake uniform extension in two directions
or simple shearing according to its first-order displacement
function. The characteristic of the linear DDA block makes
the rectangular subblocks deform only into parallelograms.
This fact indicates that it is unwise to use rectangular sub-
blocks to refine a flexural element subjected to bending. On
the other hand, the deformed shape for the beam refined by
the triangular subblocks is not restrained by the first-order dis-
placement function used in DDA.
On the analysis of the framed masonry wall, the structure
is cut into subblocks by the artificial joints. Each brick is sim-
ulated by a block, while the reinforced concrete frame is cut
FIG. 6. Tip-Loaded Cantilever Beam: (a) Cut into 128 Rectan-
gular Subblocks; (b) Cut into 128 Triangular Subblocks
FIG. 7. Deflection of Beam Cut by Various Artificial Joints
JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999 / 1113
J. Struct. Eng., 1999, 125(10): 1109-1117
into triangular concrete subblocks and the reinforcements are
modeled by the bolt elements. The wooden window in the
partially filled masonry wall is neglected in the numerical
model. The element meshes for the specimens: an RC frame,
an RC frame partially filled with masonry wall, and an RC
frame completely filled with masonry wall, are shown in Fig.
8. The RC frame is cut into 498 triangular concrete subblocks,
the tension and compression reinforcements are modeled as
114 bolts, and the stirrups are modeled as 54 bolts. Because
the specimens are overreinforced, the effect of the space of
stirrups on the behavior of RC frame is not significant. For
simplicity, the stirrups modeled by the bolts are thus arranged
in equal space. The input material properties are the same as
those of experimental specimens, which are determined by
test. The elastic modulus of steel is Es = 1.96 ⫻ 107 N/cm2
(1.96 ⫻ 105 MPa), and yield stress fsy = 3.74 ⫻ 104 N/cm2
(374 MPa). The stress-strain relation of the steel is assumed
to be bilinear, and the plastic modulus of the steel is taken to
be Ep = 0.02Es . The compressive strength of the concrete is
FIG. 8. Element Meshes of Specimens: (a) RC Frame; (b) RC
Frame Partially Filled with Masonry Wall; (c) RC Frame Com-
pletely Filled with Masonry Wall
 f ⬘c = 26.66 MPa, and the elastic modulus Ec = 4,696 兹 f c⬘ =
2.4247 ⫻ 106 N/cm2 (24,247 MPa). The tensile strength of the
concrete is f t⬘ = 271 N/cm2 (2.71 MPa). The cohesion is cho-
sen to be 280 N/cm2 (2.8 MPa). The elastic modulus of the
brick is Eb = 2.087 ⫻ 106 N/cm2 (2.087 ⫻ 104 MPa). The
tensile strength of the interface mortar is 98 N/cm2 (0.98
MPa), and its shear strength is
␶f = 3.64 ⫹ 0.75␴n (kg/cm2) (10a)
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.
tion in the failure regions. However, due to the characteristic
of the artificial joints, the predicted stress distribution does not
agree well with the experimental cracking pattern, but it re-
flects the numerical failure behavior.
Similar studies were made for the framed masonry wall. The
load-deflection relation of the RC frame partially filled with
the masonry wall is presented in Fig. 12. It is found that the
DDA solutions agree well with the experimental results. How-
ever, because the wooden window is neglected in the numer-

or ical model, the defections predicted by DDA are larger than

the experimental results. Fig. 13 shows the failure of this
␶f = 0.35672 ⫹ 0.0735␴n (MPa) (10b) framed wall after yielding of the structure. During the exper-
iment, it is observed that the failures concentrate on the RC
The stiffness of contact springs is chosen to be kn = ks = 1.96 frame, whereas there is no obvious crack found in the masonry
⫻ 107 N/mm. wall. Many horizontal cracks are found to concentrate on the
The monotonic loading is adopted in this study and the load- center region of the left column around the left upper corner
deflection relation of the RC frame is presented in Fig. 9. It of the wall. Numerous cracks are also found in the top and
is found that the numerical solutions agree satisfactorily with bottom of the left column. The failure of the right column is
the experimental results. Fig. 10 illustrates the failure behavior
of the RC frame after yielding of the structure. Referring to
Fig. 10(a), it is observed that the failure regions of the ex-
perimental specimen concentrate on the top and bottom of the
columns, the left and right ends of the beam, and the joints of
the column and the beam. As expected, most of the failures
of columns and beam are flexural failure, while there will be
mixed failures at the joints of column and beam. It is found
that there are a lot of inclined cracks at the joints of the test
specimen. The failure configuration predicted by the refined
DDA is presented in Fig. 10(b). The solid lines indicate the
failure surfaces. One can see that the failure regions predicted
by DDA agree with those of the experimental results. How-
ever, because the specimen is cut by the artificial joints into
triangular subblocks and its failure is assumed to be along the
boundary of subblocks, the numerical solutions show both
flexural failure and shear failure in the columns and beam.
This finding disagrees with the experimental results, in which
the flexural failure is the dominant one except in the joints of
beam and column. Although the artificial joints need improv-
ing in that the failure direction is prescribed and they may
create an additional failure mode, the numerical model even-
tually can predict the acceptable failure regions. Fig. 11 shows
the deformed shape and the distribution of principal stress of
this failure frame. It is found that the stress distribution of this
frame is complicated and there is observable stress concentra-

FIG. 10. Failure Configuration of RC Frame: (a) Experimental

Result; (b) DDA Result

FIG. 9. Comparison of DDA Solutions with Experimental Re-

sults: Load-Deflection Relation of RC Frame FIG. 11. Principal Stress Distribution of RC Frame

1114 / JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999

J. Struct. Eng., 1999, 125(10): 1109-1117

 ever, numerous failures are found in the masonry wall. There
are some reasons for the conflicting finding. First, the strength
of mortar is much lower than that of concrete. Second, the
wooden window frame is neglected in the numerical model
and its induced confined loading to the wall is not included.
Third, the rectangular brick block is not easily deformed. In
addition, because of the shortage of the artificial joints, the
additional shear failure mode is also found in the columns and
beam. Fig. 14 shows the deformed shape and the distribution
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.

of the principal stress of this failure partially filled frame. It

is found, as in the experimental results, that most of the loads
are carried by the frame. Similar to the RC frame, the failure
of the structure induces complicated stress distribution and
stress concentration, but the stress distribution only reflects the
numerical failure behavior.
Fig. 15 shows the load-deflection relation of the RC frame
completely filled with masonry wall. It is found that the DDA
solutions agree with the experimental results, but the predicted
displacements are smaller. When the load is greater than 550
kN, the floor beam of the test specimen is found to be totally
broken. Thus, the load-deflection relation for a load greater

FIG. 12. Comparison of DDA Solutions with Experimental Re-

sults: Load-Deflection Relation of RC Frame Partially Filled with
Masonry Wall

FIG. 14. Principal Stress Distribution of RC Frame Partially

Filled with Masonry Wall

FIG. 13. Failure Configuration of RC Frame Partially Filled

with Masonry Wall: (a) Experimental Result; (b) DDA Result

found to be similar to that of the pure RC frame, while there

are some cracks in the upper side of the beam. The reason for
the different failure configuration of the left column is that the
partially filled masonry wall induces the short column effect
on the left column. Thus, the failure of the left column con-
centrates on the contact area of the column and the wall. The
failure predicted by the refined DDA is presented in Fig. 13(b)
and the solid lines indicate the failure surfaces. One can see
that the short column effect of the left column is obviously FIG. 15. Comparison of DDA Solutions with Experimental Re-
found, and the failure regions of beam and columns predicted sults: Load-Deflection Relation of RC Frame Completely Filled
by DDA agree with those of the experimental results. How- with Masonry Wall

JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999 / 1115

J. Struct. Eng., 1999, 125(10): 1109-1117


Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.
FIG. 16. Failure Configuration of RC Frame Completely Filled
with Masonry Wall: (a) Experimental Result; (b) DDA Result
FIG. 17. Principal Stress Distribution of RC Frame Completely
Filled with Masonry Wall
than 550 kN is not presented in Fig. 15. The failure of this
framed masonry wall after yielding of structure is shown in
Fig. 16. Referring to Fig. 16(a), it is found that the failures
occurred in the columns, the left joint, and the masonry wall.
There is no observable crack in the beam except at the left
joint of the beam and column. The left column carries most
of the load and there are many horizontal cracks in it. The
failure configuration of this structure predicted by DDA is il-
lustrated in Fig. 16(b) and the solid lines indicate the failure
surfaces. One can also find that the failure regions agree with
experimental results, but there are more failure surfaces in the
wall. The principal stress distribution of this failure framed
masonry wall is given in Fig. 17. It is found that the principal
direction of the masonry wall is no longer uniform in the di-
agonal direction if the crack of the mortar and separation of
the bricks are considered. According to the principal stress
distribution of the masonry wall, the approximate equivalent
1116 / JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999
J. Struct. Eng., 1999, 125(10): 1109-1117
strut (Smith 1966) is inapplicable for the discontinuous ma-
sonry wall.
CONCLUSIONS
The structural behaviors of the framed masonry wall sub-
jected to in-plane monotonic loading are investigated by a full-
scale test and the method of DDA. The concept of artificial
joints is adopted to refine DDA so that it can be used to an-
alyze the continuous and discontinuous behavior of the ma-
sonry structure. The framed masonry walls are cut into sub-
blocks by the artificial joints. Each brick is simulated by a
block, while the reinforced concrete frame is cut into triangular
concrete subblocks and the reinforcements are modeled by the
bolts. The reinforcements and concrete are assumed to be in
perfect bond. On the failure analysis, both the tensile failure
and the shear failure of mortar and concrete are considered,
but the mixed mode failure is neglected. The shear failure is
assumed to follow the Mohr-Coulomb failure criterion.
The numerical model is verified by comparing the numerical
solutions with the exact solution and experimental results. A
satisfactory agreement is obtained. The triangular subblocks
are proved to be capable of simulating the bending behavior
of flexural elements. The structural behaviors and stress dis-
tributions of the framed masonry wall show that the charac-
teristic of the masonry structure is highly influenced by the
failure of mortar. The failure of mortar induces the discontin-
uous and nonlinear behavior of the masonry structures. Ac-
cording to the distribution of the principal stress of the ma-
sonry wall in the RC frame, it is found that the equivalent
strut cannot replace the discontinuous masonry wall. In addi-
tion, the filled masonry wall affects the behavior of the framed
masonry structure dominantly. The partially filled masonry
wall induces a short column effect and leads to a severe failure
of the column. On the other hand, the completely filled ma-
sonry wall increases the stiffness of the structure and the ad-
jacent column fails in the configuration of nearly uniform
cracks.
The proposed numerical model is therefore proved to be
capable of simulating the discontinuous behavior of the ma-
sonry structure subjected to in-plane monotonic loading and
identifying the failure regions of the structure. However, some
important factors need further investigating. The nonlinear
constitutive relations of brick, mortar, and concrete are rec-
ommended to be considered. Furthermore, the out of plane
failure of the masonry wall, the bond slip of reinforcements,
and the effects of the energy release of fracture are also highly
recommended to be fully studied in the future.
ACKNOWLEDGMENTS
This study is supported by the National Science Council of Republic
of China under Grant NSC87-2211-E-006-041.
APPENDIX. REFERENCES
Abrams, D. P., and Paulson, T. J. (1991). ‘‘Modeling earthquake response
of masonry building structures.’’ ACI Struct. J., 88(4), 475–485.
Achyutha, H., Jagadish, R., Rao, P. S., and Rahman, S. S. (1986). ‘‘Finite
element simulation of the elastic behavior of infilled frames with open-
ings.’’ Comp. and Struct., 23(5), 685–696.
Ali, S. S., and Page, A. W. (1988). ‘‘Finite element model for masonry
subjected to concentrated loads.’’ J. Struct. Engrg., ASCE, 114(8),
1761–1784.
Chang, C. T. (1994). ‘‘Nonlinear dynamic discontinuous deformation
analysis with finite element meshed block system,’’ PhD dissertation,
Dept. of Civ. Engrg., University of California, Berkeley, Calif.
Chiou, Y. J., Tzeng, J. C., and Hwang, S. C. (1998). ‘‘Discontinuous
deformation analysis for reinforced concrete frames infilled with ma-
sonry walls.’’ Struct. Engrg. Mech., 6(2), 201–215.
Clough, R. W., Gulkan, P., Mayes, R. L., and Manos, G. C. (1990).
‘‘Seismic testing of single-story masonry house: Part 2.’’ J. Struct.
Engrg., ASCE, 116(1), 257–274.
 Cundall, P. A., and Strack, O. D. L. (1979). ‘‘A discrete numerical model
for granular assemblies.’’ Géotechnique, London, 29(1), 47–65.
Dawe, J. L., and Seah, C. K. (1989a). ‘‘Behaviour of masonry infilled
steel frames.’’ Can. J. Civ. Engrg., Ottawa, 16, 865–876.
Dawe, J. L., and Seah, C. K. (1989b). ‘‘Masonry infilled steel frames
subject to dynamic load.’’ Can. J. Civ. Engrg., Ottawa, 16, 877–885.
El Haddad, M. H. (1991). ‘‘Finite element analysis of infilled frames
considering cracking and separation phenomena.’’ Comp. and Struct.,
41(3), 439–447.
El Shabrawi, A., and Verdel, T. (1995). ‘‘Modelling of ancient masonry
Downloaded from ascelibrary.org by VNR Vignana Jyothi Institute of Engineering & Technology on 12/06/16. Copyright ASCE. For personal use only; all rights reserved.
structures by the distinct element method under dynamic loads.’’ Struc-
tural studies of historical buildings IV—Volume 2: Dynamic, repairs
& restoration, C. A. Brebbia and B. Leftheris, eds., Computational
Mechanics Publications, Southampton, U.K.
Gulkan, P., Clough, R. W., Mayes, R. L., and Manos, G. C. (1990).
‘‘Seismic testing of single-story masonry house: Part 1.’’ J. Struct.
Engrg., ASCE, 116(1), 235–256.
Haider, S. (1996). ‘‘In-plane cyclic response of reinforced concrete frames
with unreinforced masonry infills,’’ PhD dissertation, Dept. of Civ.
Engrg., William Marsh Rice University, Houston.
Ke, T. C. (1993). ‘‘Simulated testing of two dimensional heterogeneous
and discontinuous rock masses using discontinuous deformation anal-
ysis,’’ PhD dissertation, Dept. of Civ. Engrg., University of California,
Berkeley, Calif.
Liauw, T. C. (1972). ‘‘An approximate method of analysis for infilled
frames with or without opening.’’ Build. Sci., 7, 233–238.
Lin, C. T. (1995). ‘‘Extensions to the discontinuous deformation analysis
JOURNAL OF STRUCTURAL ENGINEERING / OCTOBER 1999 / 1117
J. Struct. Eng., 1999, 125(10): 1109-1117
for jointed rock masses and other blocky systems,’’ PhD dissertation,
Dept. of Civ., Envir. and Arch. Engrg., University of Colorado, Boul-
der, Colo.
Lotfi, H. R., and Shing, P. B. (1991). ‘‘An appraisal of smeared crack
models for masonry shear wall analysis.’’ Comp. and Struct., 41(3),
413–425.
May, I. M., and Naji, J. H. (1991). ‘‘Nonlinear analysis of infilled frames
under monotonic and cyclic loading.’’ Comp. and Struct., 38(2), 149–
160.
Mehrabi, A. B. (1994). ‘‘Behavior of masonry-infilled reinforced concrete
frames subjected to lateral loadings,’’ PhD dissertation, University of
Colorado, Boulder, Colo.
Papia, M. (1988). ‘‘Analysis of infilled frames using a coupled finite
element and boundary element solution scheme.’’ Int. J. Numer. Meth-
ods in Engrg., 26, 731–742.
Saneinejad, A., and Hobbs, B. (1995). ‘‘Inelastic design of infilled
frames.’’ J. Struct. Engrg., ASCE, 121(4), 634–650.
Shi, G. H. (1988). ‘‘Discontinuous deformation analysis: A new numerical
model for the statics and dynamics of block systems,’’ PhD disserta-
tion, Dept. of Civ. Engrg., University of California, Berkeley,
Calif.
Shyu, K. (1993). ‘‘Nodal-based discontinuous deformation analysis,’’
PhD dissertation, Dept. of Civ. Engrg., University of California, Berke-
ley, Calif.
Smith, B. S. (1966). ‘‘Behavior of square infilled frames.’’ J. Struct. Div.,
ASCE, 92(1), 381–403.
Thiruvengadam, V. (1985). ‘‘On the natural frequencies of infilled
frames.’’ Earthquake Engrg. Struct. Dyn., 13, 401–419.