You are on page 1of 11

Caqwrers & Slrwhwes Vol. 61, No. 4, pp. 735-74.

5, 1996
Pergamon copyrisht0 1996Elsevicr sdcaa Ltd
Fhted in Great Britain. AU rights rcserwd
00457949(9spn61-4 0045-7949/96 SI5.00 + 0.00

NUMERICAL MODELLING OF BRICK MASONRY PANELS


SUBJECT TO LATERAL LOADINGS
J. S. Lee, G. N. Pande,t J. MIddIeton and B. KraIj
Department of Civil Engineering, University of Wale-sSwansea, Singleton Park, Swansea SAZ 8PP, U.K.

(Received 26 December 1994)

Abstract-A homogenization technique has been introduced to investigate the elastic-brittle behaviour
of masonry panels subject to incremental lateral loading. For modelling the elastic behaviour of masonry,
two successive steps of homogenization are used to obtain equivalent elastic properties. In the first step,
brick units are homogenized with perpend joints to give equivalent elastic properties of a stacked system.
This stacked system, in the second step, is then homogenized with the bed joints to obtain equivalent
material properties for masonry. Tensile cracking is the only nonlinearity considered in this paper.
Cracking is judged on the basis of stresses and strengths of each of the constituent materials. The cracks
developed, if any, are also homogenized with the homogenized masonry and equivalent nonlinear
stress-strain relationships for cracked masonry are derived. The constitutive model is incorporated in a
three-dimensional finite element code. It has been verified and validated with experimental data on the
response of a set of laterally loaded rectangular masonry panels with and without openings. It is considered
that the model can be used for predicting the physical behaviour of laterally loaded panels of arbitrary
geometry and boundary conditions. Copyright 0 1996 Elsevier Science Ltd

INTRODUCTION performed a parametric analysis of masonry panels


by employing a plate element and changing material
Masonry panels with in-plane loadings have been
properties randomly.
widely investigated in the past, both experimentally
In this study, a three-dimensional numerical
and analytically. However, relatively few analyses
analysis of masonry panels subject to distributed
and methods of design of masonry panels subject to
lateral loading is considered and the results are
lateral loadings (out-of-plane) have been proposed
compared with experimental data. For this, a
and only approximate analytical solutions based on
20-noded solid element together with orthotropic
the yield line theory [ 1,2] have been considered. In
material properties derived from the strain energy of
this theory, the panel is assumed to be divided into
the composite material [A will be used throughout.
several rigid bodies according to the direction of
The constituent materials, brick, bed and perpend
maximum tensile stresses and the failure loads are
mortar joints, are assumed to have elastic-brittle
predicted on the basis of limit theorems. A similar
characteristics. Once cracking based on a maximum
approach based on fracture line theory was proposed
principal tensile stress criterion is detected, the cracks
by Sinha [3], while a simplified solution for partially
are homogenized with the adjacent orthotropic
grouted concrete masonry was introduced by Hamid
material to trace the crack propagation. Therefore,
et al. [4].
two stages of homogenization are used, i.e. one for
Accurate numerical analysis of masonry panels
the orthotropic material and the other for smeared
including lateral loadings, by its nature, has to be
cracking of the material. In other words, the
based on a three-dimensional model. This requires
constituent materials are not discretized individually
considerable computational effort which is further
and the homogenized masonry material representing
compounded by the need to discretize the brick and
the equivalent orthotropic properties is used in the
bed as well as perpend mortar joints. To simplify this
finite element discretization. Tensile cracks are again
procedure, Chong et al. [5] adopted a four-noded flat
introduced through homogenization with the ortho-
shell element and used smeared material properties,
troic material characteristics of the masonry. This
and calculated failure pressures closer to the
allows simulation of cracking, crack propagation and
experimental results when compared with that of the
crack patterns to be predicted.
yield line theory. Meanwhile, Lawrence and Lu [6]
Using the above approach it is possible to carry out
three-dimensional finite element analyses of masonry
panels with a reasonable computational effort. It is
t To whom all correspondence should be addressed. also shown that the pattern of cracks in masonry

735
136 J. S. Lee et al.

panels are in good agreement with the experimental where


data.

ORTHOTROPIC MATERIAL PROPERTIES OF MASONRY


(3)
Masonry panels subjected to lateral loading can be
numerically analysed if an accurate stress-strain 0
relationship is employed for each constituent material
and each constituent material is then discretized
individually. However, a three-dimensional analysis 0
of a masonry panel involving only a very simple
geometry would require a large number of elements 0
and the nonlinear analysis of the structure would be
almost intractable. To overcome this computational 19 =
difficulty, the orthotropic material properties pro- 0 0 o&o XY
0

posed by Pande et al. [7,8] are used to model the


masonry panel in the sense of an equivalent material. 0 0 0 ok 0
Here, only brief details of the procedure to obtain Y2

equivalent elastic parameters are given for clarity and


completeness, and the reader should refer to the
original papers for full details. I_
0 0 000
k
The basic assumptions made to derive the (4)
and the details of the orthotropic material properties
equivalent material properties through the strain
of masonry in terms of the properties of the
energy consideration are:
constituents are given in Appendix A where the
(1) brick and mortar are perfectly bonded;
homogenization is first applied to the brick and bed
(2) perpend mortar joints are assumed to be
mortar joint. It is noted that, in Pande et al. [7], the
continuous.
equivalent material properties were derived with the
The second assumption is necessary in the
brick and the perpend mortar joint being homogen-
homogenization procedure and it has been shown [9]
ized first.
that the assumption of continuous perpend joints
The equivalent orthotropic material properties
instead of staggered joints does not have any
defined above are used to construct the system
significant effect on the stress states of the constituent
stiffness matrix in the finite element analysis
materials although a local stress concentration within
procedure and, from this, equivalent stresses/strains
the perpend joint can not be obtained.
are then calculated. The stresses/strains in the
Let the orthotropic material properties of the constituent materials can be evaluated through
masonry panel be denoted by EX, ti;,, B, VXY, &, vyi,
structural relationships, i.e.
GX,, GYZ,CXZ,Fig. 1. If the stress/strain relationship
of the equivalent (masonry) material is represented by
bb = [&,]a

b = [b]C fibj = [sbj]a

or Uh, = [ShJl~, (5)

z = [C]S where, subscripts b, bj and hj represent brick, bed


(2) joint and perpend joint, respectively. The structural
relationships for strains can similarly be established.
The structural matrices S are listed in Appendix B.

MODELLING OF CRACK PROPAGATION

It is assumed that tensile cracking is the most


important nonlinearity that governs the behaviour of
masonry panels. Here, stresses in each constituent
material, i.e. brick, bed and perpend mortar joint,
are calculated through the structural relationship
defined in the previous section and a check for
cracking is made based on the maximum principal
stress criterion for each constituent material.
Fig. 1. Coordinate system used in masonry panel. Once cracking occurs in any constituent material,
Numerical modelling of brick masonry panels subject to lateral loadings 737

The kinematic or compatibility condition along the


interface is

Since the thickness of the crack, t, is negligible


Fig. 2. Representative elementary volume including a crack. compared with that of the elementary volume, i.e.,
t<<d, further refinement can be considered through
introduction of velocity discontinuities, g, which are
the effect is smeared onto the neighbouring equivalent a function of the deformation field and width of the
orthotropic material through homogenization tech- crack:
nique which has been previously applied to masonry
in a two-dimensional case [lo]. Here, a homogeniz-
ation (averaging) procedure based on the work of
Pietruszczak and Niu [l l] is adopted for the Based on the assumption of negligible crack width,
three-dimensional case and is described below: eqn (7) together with eqn (10) can then be rewritten
Let the stress/strain rate of the equivalent material as
after cracking (cracked masonry) be represented by
[6]i x [S]i’ + fig, (12)

where,

The stress/strain rate in eqn (6) is, in fact, the


volume average of the stress/strain in the two phase
material, i.e. intact masonry and cracks (Fig. 2)
The constitutive relationship for each constituent is
then defined as

L = ,uii’ + pjc, (7)

where, pj and pj represent the volume fraction of the U = [K*]g, (14)


constituent materials and, for simple geometry such
as in Fig. 2, can be defined as where [D] can be derived from eqn (l), and [K*] is a
6 x 3 matrix which can be represented by
bwt

1
P/ = 7
UK,, LI aCI &I aK3I 0
[K*JT = aKlz K12 CIKU KX aK32 0 (15)
where V = bdw. Here subscript i will be used to a& G3 a& K23 aK33 0
denote quantities relating to intact masonry and j to
denote quantities relating to cracks. In fact, eqn (7)
can be called an averaging rule or a mixture rule in or the stress-strain relationship of the crack can be
the sense of Hill [12]. rewritten as
Assuming perfect bonding at the interface of the
two phases, the equilibrium condition in the local wj = KlL (16)
X-Y-Z coordinate system, Fig. 2, can be established
as
where,

ayy= (+I
YY
= (.y
YY ‘KN 0 0

7, =
.i _
7,

‘i
7yz =I 7yz -
-

_
'j
7,

.i
7yz. 64
WI = [II;1 =

for elastic response and


0 KT
0 0 4
0
1 (17)
138 J. S. Lee et al.

Table 1. Material properties of masonry panel


Parameter Notation uuit Value
(1) (2) (3) (4)
Masonry panel
dimension LxHxT 5.5 x 2.6 x 0.102
flexural strength I( to bed joint NlZll-2 0.32
I to bed joint N mm-* 1.15
Brick
dimension lxhxtb mm 212 x 65 x 102
Young’s modulus Eb N mm-* 22,528
Poisson’s ratio Vb 0.25
tensile strength Ulb N mm-* 1.85
Mortar
thickness tbj, ih, mm 10
Young’s modulus Ebj, Ebj N mm-* 7400
Poisson ratio rblr Vhi 0.3
tensile strength Urbj N mn-2 0.32
Uthj N mm-* 1.15

1
Ku Ku KS respectively. Alternatively, KN and KT can be
WI = &I K22 K23 (18) arbitrarily assumed where the thickness t is very

[ KN K,,
K32 small. Finally, once cracking occurs, the components
of [a in eqn (18) are set to a very small number, i.e
for post-cracking response with K,xO.
The following relations can be obtained from
eqns (7), (9) and (10);
,N==+2b
-=
(1 - v)E
t t(1 + v)(l - 2v) [S]ii = [A]( + [B]&, (20)

where
K*,$,L
2t(l + v)

a=,_%_
KN -l-v
v (19) -042
+______ -Da K22 -0.45
PI = c2 c2 c2 PC2 c2

where, A,,and 16 are the Lam& constants, and KN and -& - 052 -Dss - 0~ KM
Kr are the normal and shear stiffnesses of the crack, 1c,c,c,c,- PC3

Experiment -
Model + _

0 2 4 6 8 10
Displacement (mm)
Fig. 3. Applied lateral pressure-displacement curve of a four-sided supported solid panel.
Numerical modelling of brick masonry panels subject to lateral loadings 139

C,=D,+$

Equation (20) can then be further refined as

[b]i’ = [A]i + [B]g = [A]i + [B] ; ([6]C - [d]c’?. (22)

Rearranging,

[S]i’ = [s’]c’ (23)

or

i’ = [Sill, (24)

where

and

-1 0 0 0 0 o-
&Y
I, s;, s;, $4 s;, X6

pi]= O O l O O O
s;, s;, s;, s;, s;, s;,
s;, s;, s;, si, x5 s;,
Fig. 4. Crack propagation in a four-sided supported 0 0 0 0 0 1
solid panel (a) y = 1.0 kN m-l; (b) q = 1.8kN m-‘;
(26)
(c) 4 = 2.2 kN m-‘; (d) q = 2.5 kN m-*.
Using eqns (12) and (23),

g = t ([S]i - [6]i’)

PI =
= ; ([S] - [S])i

= [S{]i, (27)
740 J. S. Lee ef al.

where a size of 6 x 6, elements of which are fnnctions of the


direction cosines of the major tensile principal stress.
It is noted that the principal stresses and their
(28) directions can be calculated by solving an eigenvalue
problem and, therefore, the transformation matrices
Here, [$I and [$I can again be called structural in eqn (32) can easily be obtained.
matrices through which the relationship between
strains of the constituent material and the homogen-
NUMERICAL EXAMPLE
ized equivalent material are established. Finally, the
constitutive relationship for the equivalent material is The two stage homogenization techniques de-
derived from the mixture rule, eqn (7) scribed in the previous sections are implemented in a
standard three-dimensional tinite element program.
d = pit+ + jlj# x 1’ = [O][S$ (29) The computer program is used to predict the results
of full scale experiments conducted by the British
or Ceramic Research Limited. Table 1 shows the
material data used in the current study. It is noted
ci= [LqC (30) that the tensile strength of the mortar required for the
model is the lesser of (a) actual tensile strength (b)
where bond strength between mortar and the brick. This is
because the failure of the bond between the brick and
[P] = [D][$]. (31) the mortar joint is not explicitly considered in the
model and these two constituents are assumed to be
So far, all the derivations are based on the local bonded together at all times during the history of
X-Y-Z system of axes in Fig. 2. The stress-strain deformation. In the present case, the tensile strength
relationship in the global x-p-z axis system will be of the mortar is assumed to be equal to the flexural
evaluated from the following relations: strength of the masonry panel as obtained from tests
on wallettes. Furthermore, the values of KN and KT
b = [T]B defined in the previous section for the initial elastic
response are arbitrarily chosen as 10 N mm-),
i = [T,]T (32) respectively. The brick used in the experiment is a
perforated unit with three holes and, in the numerical
and analysis, Young’s modulus from actual experimental
work is adopted. The Young’s modulus of mortar is
b = [Z-J-‘[DW][z-,]Z= [b;W]E (33) chosen as 74OONmn-* which corresponds to the
value also chosen by Ali and Page [ 131and Poisson’s
where ratios, vb, vj and are the typical values chosen
vbj,

arbitrarily.
[iW = m-‘[DqT,] (34) In experiments, incremental pressures are applied
to the whole area of the wall panel with an air bag
and [q and [T,] are the transformation matrices with and thin plywood is attached to the window hole, if

I I I I I I -A

Experiment - -
Model -t

I I I I I
0 2 4 6 8 10 12 14 16
Displacement (mm)
Fig. 5. Applied lateral pressure-displacement curve of a three-sided supported panel with opening.
Numerical modelling of brick masonry panels subject to lateral loadings 741

(4
Fig. -ack propagation in a three-side supported panel with opening (a) q = 0.30 kN
(b) q = 0.50 kN m-*; (c) q = 0.70 kN rn-$ (d) q = 0.96 kN rne2.
742 J. S. Lee et al.

any. The effect of window loading is modelled by From the application of the homogenization
analysing the window area separately and by technique and numerical analysis of the model panel,
applying reactions as incremental nodal forces at the the following conclusions can be drawn:
edge of the window. (a) the proposed two stage homogenization tech-
nique is convenient, as well as practical, and can be
Four-side supported solid panel used to minimize the numerical effort of the analysis
Figure 2 shows the relationship between applied of masonry panels involving lateral loading;
lateral pressure q and the displacement at the center (b) the numerical prediction of the ultimate load is
point 0 of a panel simply supported on all the sides. in good agreement with the test data and the crack
The initial cracking pressure is 0.9 kN m-* and the pattern also matches that of the failure mode of the
predicted ultimate pressure is virtually the same as test panel;
that from the test work. (c) more accurate material parameters, such as
The development of the pattern of cracking which tensile strength and tensile bond strength between
corresponds to the evolution of the value of g, in mortar and brick, would provide a more accurate
eqn (11) is illustrated in Fig. 3. It is seen to match model.
closely with the experimental failure mode, although Numerical analyses of masonry panels having
an absolutely symmetric pattern is not expected in different boundary conditions and windows have also
practical situations. The crack widths in Fig. 3 are been performed and they generally show good
relatively high and this is due to the following agreement with the experimental results.
reasons:
(1) the homogenized equivalent material proper- Acknowledgements-The research reported here was spon-
sored bv the British Research Establishment (BRE). U.K.
ties are used in the finite element analysis; The auihors would like to acknowledge the many ‘helpful
(2) the mesh adopted is rather coarse and though discussions they had with Mr S. K. Arora, Dr R. De Vekey
this mesh is capable of capturing global behaviour, it and Mr A. Ferguson of BRE regarding the experimental
is not fine enough to obtain crack widths accurately; programme of the testing of masonry panels.
(3) The contours are plotted with nodal point
values rather than Gauss point values. REFERENCES
In Fig. 3, one quarter of the brick panel is analysed
using 420 20-noded isoparametric solid elements 1. B. A. Haseltine, H. W. H. West and J. N. Tutt, Design
of walls to resist lateral loads. Srrucr . Engnr S&422-430
including three layers of elements in the Z-direction. (1977).
The total degrees of freedom in this case are 13,899. 2. B. A. Haseltine and J. N. Tutt, Implications of research
It is noted that the dead weight of masonry is on design recommendations. Struct. Engnr 64A, 341-
currently not included in the analysis since the 350 (1986).
3. B. P. Sinha, A simplified ultimate load analysis of
symmetric boundary conditions will not be applicable laterally loaded model orthotropic brickwork panels of
if dead weight is considered. low tensile strength. Struct. Engnr 56B, 81-84 (1978).
4. A. A. Hamid, S. R. Chandrakeerthy and 0. A.
Three-side supported panel with opening Elnawawy, Flexural tensile strength of partially grouted
concrete masonry. J. Strnct Engng ASCE 118, 3377-
Figure 4 shows the relationship between applied 3393 (1992).
lateral pressure q and the displacement at one corner 5. V. L. Chong, I. M. May, C. Southcombc and S. Y. A.
(marked “0” on the inset in the figure) of a panel Ma, An investigation of laterally loaded masonry
panels using non-linear finite element analysis. In: Proc.
simply supported on three-sides with an opening. It Int. Symp. Comp. Meth. Struct. Masonry (Edited by
shows that the predicted ultimate load is lower than J. Middleton and G. N. Pande), Swansea, pp. 49-63.
that by test, although only one set of the test results Books and Journals International, Swansea (1991).
is currently available to the authors. The initial 6. S. J. Lawrence and J. P. Lu, An elastic analysis of
cracking pressure, in this case, is 0.28 kN m-* and the laterally loaded mesonry walls with openings. In: Proc.
Int. Symp. Comp. Meth. Struct. Masonry, Swansea
opening has a dimension of 2.8 x 1.4 m. The crack (Edited by J. Middleton and G. N. Pande), pp. 3948.
propagation pattern is illustrated in Fig. 6 and it Books and Journals International, Swansea (1991).
corresponds well to that observed in the tests. I. G. N. Pande, B. Kralj and J. Middleton, Analysis of the
compressive strength of masonry given by the equation
6 =K(/I).&y).. S%uc?. Engnr ?1;7-12 (1994). _
8. G. N. Pande. J. X. Liana and J. Middleton. Equivalent
CONCLUSIONS
elastic mod& for brick-masonry. Comput.‘Ge&ech. 8,
Two stage homogenization techniques have been 243-265 (1989).
9. J. S. Lee, G. N. Pande, J. Middleton and B. Kralj,
applied to a masonry panel subject to incremental Analysis of tensile strength of masonry. In: Proc. 10th
lateral loading. In the iirst stage, brick and mortar Inr. Brick/Block Masonry Conf., Calgary, Canada,
joints are homogenized to artifically generate pp. 21-29 (1994).
equivalent material properties which are directly used 10. J. S. Lee, G. N. Pande and J. Middleton, Numerical
modelling of cracking in masonry structures. In Proc.
in the finite element modelling of the panel. Later, 2nd Inst. Symp. Camp. Methd. Struct. Masonry (Edited
tensile cracks are homogenized with the equivalent by G. N. Pande and J. Middleton), Swansea, pp. l-3.
material to model crack propagation. Books and Journals International, Swansea (1993).
Numerical modelling of brick masonry panels subject to lateral loadings 743

11. S. Pietruszczak and X. Niu, On the description of and the relationship below also is established
localized deformation. Int. J. numer. Anal. Meth.
Geomech. 17, 791-805 (1993). ,14
’ -- vxy
12. R. Hill, Elastic properties of reinforced solids: some vy* E’ . (A4)
x
theoretical principles. J. Mech. Phys. Solids 11,357-372
(1963). For the system of masonry panels, the homogenization is
13. S. Ali and A. W. Page, Finite element model for applied to the layered material and perpend joint based on
masonry subjected to concentrated loads. J. Struct. the assumption of continuous perpend joint. Now, volume
Engnr ASCE 114, 1761-1784 (1988). fractions of the constituent materials are
14. C. M. Gerrard, Elastic models of rock masses having
one, two and three sets of joints. ht. J. Rock Mech. I /&.._x (A%
Min. Sci. Geomech. Abstr. 19, 1523 (1982). kbg=l+lhji I+ thj ’

where, subscript eq and hj represent layered material and


APPENDIX A perpend joint, respectively. Defining the following co-
ORTHOTROPIC MATERIAL PROPERTIES OF MASONRY efficients:

Derivations of the equivalent orthotropic material


properties of the masonry, which includes brick and bed as
well as perpend mortar joint, are based on the strain energy
criterion. It has been initially applied to jointed rock
masses [14]. Referring to Fig. 1, volume fraction of brick
and bed joint can be described as

where subscripts b and bj represent the brick and bed joint,


respectively. Assuming homogeneous, isotropic properties
of constituent materials and defining the following
coefficients:

oL = /&%(l - Vf,) + /.kbjEbj(l - Vz)


1 2 x = xeg•l- xhl
(1 - Vb)(l - Vbj)

x, = rop(v;r + v&)
i = /LbVbEb(l - Vi) + /Ib,Vb,&.j(l - Vi)
2 2 1 - v&v&
(l - Vb)(l - Vbj)

1, = E

1 = X, + Xhj (‘46)

the orthotropic material properties of the masonry panel are


x= Xb + xbl. 642)
derived

the following orthotropic material properties for the layer


homogenized with brick and bed joint are obtained

E: = a - (viz

I,&?+& + 2
xb(~-$+2~bj(~-$
E; Eb Ebj

VA= ia

v; = x(1 - VLZ)
I
VA= vrz

v&= x(1 - vh) (A3)


144 J. S. Lee et al.

647)

APPENDIX B
iL,E: i$v&E;
STBUCl’UBAL lUCLATiONSHIP OF CON-
MATIZBULS
s;, = c’
{
(VL + v&v&) - -
Ez
-
E, I
A general form of structural matrix S can be written as

1
SII S12 s,3 0 0 0
St1 $2 su 0 0 0
&I S32 s33 0 0 0 s;,=c(py)
PI = 0 0 0 su 0 0 (Bl)
0 0 0 0 s55 0
0 0 0 0 0 s66 s;l=$ uw
yl

and the non-zero components of structural matrix between where


perpend joint and equivalent material, [S,] are
c’_-!--- WI
S,+/,ll= 1.0 S&44= 1.O Shj,&= 1.0 1 - v;zv:Y.

Finally, structural matrices of brick and bed joint [Ss] and


[Su] are derived from the following relationship:

Ub= [P&r

ubj = [Pt.&r’, (B6)

where the general form of [P] has the same form as in


eqn (Bl) and, for the brick,

Pb.22= 1.0 A,44= 1.0 Pb,s = 1.0

where

(B3)

To derive structural matrices of brick and bed joint, the


structural relationship for layered system [S’] is established
tirst. That is Gb
Pw=a, (B7)

where

1
Cb’_’ (Bg)
Numerical modelling of brick masonry panels subject to lateral loadings 745

Also, for the bed joint

&,@ = 1 .o Pbj,,, = 1 .o Pbj+55 = 1.0

(B9)
Pbj,,, =c@_y)
where

1
Pbj.12
= Cb, Vbj(1+ VbJ)- y+v c, = 1 - “6. (B1’3
{ ( >I

From eqns (B5) and (B6)

= =
Ubjl [Pbj][S']B [Sbj]dp Wl)

where

[sbl = [pbl[sl
v:yEbl v$vbjEbl
Pbj.32 = cbj vbj(l + vbj) -
E’I + Ex )I (B12)

You might also like