You are on page 1of 21

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2006; 66:1717–1737


Published online 13 January 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nme.1537

Derivation of 3D masonry properties using numerical


homogenization technique

Chengqing Wu1, ∗, †, ‡ and Hong Hao2


1 School of Civil and Environmental Engineering, The University of Adelaide, SA 5005, Australia
2 School of Civil and Resource Engineering, The University of Western Australia, 35 Stirling Highway,
Crawley WA 6009, Australia

SUMMARY
Lots of research work has been conducted on homogenization technique, which derives global ho-
mogenized properties of masonry from the behaviour of the constitutive materials (brick and mortar).
Such a technique mainly focused on two-dimensional media in the previous studies with the out-of-
plane properties of masonry material neglected. In this paper, homogenization technique and damage
mechanics theory are used to model a three-dimensional masonry basic cell to numerically derive the
equivalent elastic properties, strength envelope, and failure characteristics of masonry material. The
basic cell is modelled with distinctive consideration of non-linear material properties of mortar and
brick. Various displacement boundaries are applied on the basic cell surfaces in the numerical simu-
lation. The detailed material properties of mortar and brick are modelled in a finite element program
in the numerical analysis. The stress–strain relations of masonry material under various conditions
are obtained from the simulation. The homogenized elastic properties and failure characteristics of
masonry material are derived from the simulation results. The homogenized 3D model is then utilized
to analyse the response of a masonry panel to airblast loads. The same panel is also analysed with
distinctive material modelling. The efficiency and accuracy of the homogenized model are demon-
strated. The homogenized material properties and failure model can be used to model large-scale
masonry structure response. Copyright 䉷 2006 John Wiley & Sons, Ltd.

KEY WORDS: homogenization technique; masonry; basic cell; numerical analysis

1. INTRODUCTION

Recently, researchers have paid much attention on investigating the complex mechanical
behaviour of masonry structures. Masonry is a composite structure, which is constituted by

∗ Correspondence to: C. Wu, School of Civil and Environmental Engineering, The University of Adelaide,
SA 5005, Australia.
† E-mail: cwu@civeng.adelaide.edu.au
‡ This work was completed in the University of Western Australia.

Contract/grant sponsor: Australian Research Council (ARC); contract/grant number: DP0451966

Received 16 March 2005


Revised 9 August 2005
Copyright 䉷 2006 John Wiley & Sons, Ltd. Accepted 13 September 2005
1718 C. WU AND H. HAO

blocks of bricks jointed by mortar joints. In order to understand the mechanical behaviour
of the masonry assembly, both continuum and discrete block methods have been developed
to analyse the linear and non-linear response of masonry structures. The discrete method is
usually carried out by schematizing the blocks as linear elastic and the interfaces governed
by Coulomb friction law [1–5]. This approach will lead to a very expensive computational
problem and is only applicable to modelling the behaviour of small specimens. The continuum
model can be used to simulate the large scale of masonry structures. However, this approach
needs obtaining the equivalent material properties of masonry. The homogenization technique,
which aims at deriving the behaviour of the composite from the geometry and behaviour of
the basic cell, can achieve this goal.
The techniques of homogenization, permitting to establish constitutive relations in terms of
averaged stresses and strains from the geometry and constitutive relations of the individual
components, are getting more and more popular among the masonry community in the last
decade [6–14]. Normally, three types of approaches have been developed. The first approach
is to use a simplified geometry to represent the complex geometry of the basic cell so that a
close-form solution of the homogenization problem is possible [7, 8]. This method is suitable
for linear elastic behaviour of masonry and also for a relatively simple extension to a non-
linear behaviour of masonry. The second approach is to apply the homogenization theory for
the periodic composite to the basic cell [9, 14]. Due to the complexity of masonry basic cell,
it is necessary to use the finite element method to obtain a numerical solution of a problem.
This approach is applicable for the analysis of non-linear behaviour of the complex masonry
basic cell by solving the problem for all possible macroscopic loading histories. The third
approach is to adopt micromechanical and microstructural model for a detailed simulation of
the interactions between the different internal components of the basic cell [11, 13]. Although
this approach is a very useful tool, its applications are limited since it is difficult to determine
several parameters in the micromechanical model for macroscopic analysis.
Most previous studies derived homogenization properties of masonry wall. Since masonry
wall is either not load carrying or used to carry only in-plane load, the out-of-plane properties
of the masonry wall was usually not considered. Recently, a numerical model with distinctive
consideration of the non-linear properties of mortar and brick has been used in the homogeniza-
tion process to derive the equivalent material properties of masonry wall, including equivalent
elastic moduli, Poisson’s ratio, strength envelope and failure patterns of masonry [12]. The
latter study, however, again only considered the in-plane characteristics of masonry.
Recently, a lot of research work has been conducted to simulate response and collapse of
masonry structure and masonry infilled RC structure to explosive loads [15, 16]. If continuum
model of masonry, i.e. mortar and brick are discritized and modelled individually, are employed
in the numerical model, it is very time consuming and computationally ineffective to simulate
its damage and collapse process of such structures, especially for large-scale structures. In
order to effectively analyse masonry wall or masonry infilled wall to explosive loads, 3D
homogenized masonry material properties are needed. The objective of this study is to derive
the homogenized masonry material properties. The previous work on deriving 2D masonry
material properties [12] is extended to 3D. A three-dimensional masonry basic cell is used
in the study. In the numerical model, the material properties of masonry and mortar such as
stress–strain relation, strength envelope and damage characteristics, are modelled in detail in the
basic cell. In the homogenization process, the brick failure is simulated by double scalar damage
model and the concept of fracture energy based damage is used for the failure of mortar joint.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1719

By applying various displacement boundaries, the stress–strain relations of the basic cell under
different conditions are derived and the equivalent material properties and failure characteristics
of the basic cell are obtained. The derived homogenized 3D masonry material model is used
to analyse a masonry panel subjected to airblast loads, and the same masonry panel is also
analysed with the continuum model in which mortar and brick are modelled distinctively. The
efficiency and accuracy of the homogenized 3D model is verified. The derived homogenized
material model can be employed to analyse large-scale masonry structures or masonry infilled
structures to explosive loads.

2. HOMOGENIZATION OF MASONRY BASIC CELL

Masonry is a composite medium characterized by a regular arrangement of bricks that are


periodically distributed and jointed with mortar. Therefore, a repetitive cell, which includes
the geometry of the structure and the material properties, can be determined. The smaller
repetitive cell (see Figure 1), which contains all the geometric and constitutive information
of the masonry, is employed as a basic cell for the investigation of the composite material.
To establish the macroconstitutive stress–strain relation of masonry unit, the components of
averaged stress and strain ¯ ij and ¯ij are defined by the integral over the basic cell as

1
¯ ij = ij dV (1)
V V

1
¯ ij = ij dV (2)
V V

where V is the volume of the basic cell, ij and ij are stress and strain in an element.
Using stress–strain relations of the brick and mortar material, it can numerically reproduce the
equivalent stress–strain relation of the basic cell with homogenization technique by applying
various displacement boundary conditions on the surfaces of the basic cell. The equivalent
material properties of the basic cell can be derived from the simulated stress–strain relations
in the numerical analysis.

3. MATERIAL MODEL FOR MORTAR AND BRICK

Due to the difference existing between brick units and mortar, a complex interaction between
the two masonry components occurs with masonry deformation. In general, non-linear behaviour
of masonry unit is dominated by mortar joint. In order to derive the homogenized inelastic
material properties of masonry basic cell, a reliable material model for masonry components
(brick and mortar) is important.

3.1. Mortar joint


For mortar joint in uniaxial stress state, analytically convenient functions of the stress–strain
diagrams in uniaxial compression and tensions can be approximately described with exponential

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1720 C. WU AND H. HAO

Basic cell

Homogenized continuum

Figure 1. Basic cell for masonry and homogenization process.

forms. They are [12], respectively,


 = Ec c , c  co (3)
 = co exp(−c (c − co )/co ), c > co (4)
and
 = Et t , t  to (5)
 = to exp(−t (t − to )/to ), t > to (6)

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1721

where co , to are initial compressive and tensile yield stress, respectively, and it has co = Ec co
and to = Et to , in which co and to are the threshold strains that initiate compressive and
tensile fracturing of the material; t and c are parameters that control post-failure stress–strain
relations of materials in tension and compression, respectively. The uniaxial versions of the
damage variables, D̄c and D̄t in compression and tension, respectively, are defined as [17]
Gc Gt
D̄c = , D̄t = (7)
Gcf Gtf
where Gcf and Gtf are the total fracture energy per volume in compression and tension and are
defined as
 ∞  ∞
Gcf =  dc , Gtf =  dt (8)
0 0

and the fracture energy density Gc and Gt are in the form of


 c  t
Gc =  dc , Gt =  dt (9)
0 0

Based on the continuum damage theory, damaged elastic constitutive matrix Ẽijkl is described as
Ẽijkl = (1 − D̄)E0ijkl (10)
in which E0ijkl is undamaged elastic constitutive matrix, D̄ is degradation damage variable. For
quasi-brittle materials such as mortar, the degradation of stiffness from microcracking occurs
in tension and compression and the response from tensile and compressive damage are quite
different. To account for the different damage responses of mortar in tension and compression,
the degradation damage variable is defined as [18]
D̄ = 1 − (1 − D̄t )(1 − D̄c ) (11)
where D̄t and D̄c are determined by Equation (7). This definition implies the total damage
depends on both tensile and compressive damage of the mortar material, but it is not a direct
summation of the two damage scalars.
The yield strength of quasi-brittle materials is of non-linearity and can be modelled by
Drucker–Prager strength criterion. In the present study, it is described as

F = I1 + J2 − c(D̄) = 0 (12)
where J2 is the second invariant of the stress deviator and I1 is the first invariant of the stress
tensor, given by J2 = 0.5sij sij , I1 = (1 + 2 + 3 ).  and c are material constants. Suppos-
ing that the post-failure surface satisfies the isotropic softening rule, the constant  remains
unchanged, while c will degrade with damage as
c(D̄) = c0 (1 − D̄) (13)
The flow rule gives the relation between the plastic flow direction and the plastic strain rate.
Usually, the non-associative flow rule is necessary to control the dilatancy in modelling frictional
materials [19]. The plastic potential function in the present study is proposed as

 = I1 +  J2 (14)

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1722 C. WU AND H. HAO

where  is a parameter that controls the dilatancy. An associated plastic flow theory will be
initiated if the parameter  is equal to 1.
In the incremental theory of plasticity, the total strain increment dij consisting of the elastic
p
part deij and plastic part dij is
p
dij = deij + dij (15)
and the plastic strain rate can be expressed by a non-associated flow rule in the form of

p *
dij = d (16)
*ij
where d is plastic multiplier that governs the plastic flow after yielding. The elastic–plastic
stress–strain relation is expressed as
ep
dij = Ẽijkl dkl (17)
where
ep Ẽijtu (*F /*rs )(*/*tu )Ẽrskl
Ẽijkl = Ẽijkl − (18)
h + (*F /*mn )Ẽmnpq (*/*pq )
in which h denotes the hardening parameter
*F * *F *c *
h= − p − (19)
*ij *ij *c *pij *ij
and the plastic multiplier can be expressed as
(*F /*ij )Ẽijkl dkl
d = (20)
h + (*F /*mn )Ẽmnpq (*/*pq )
With the aid of Equations (12)–(14) and (19), (20), the hardening parameter can be calculated.

3.2. Damage model for brick


Under high stress, the damage of brick might occur owing to compressive crushing and tensile
splitting. It can be characterized by two damage scalars, namely, Dt and Dc corresponding,
respectively, to the damage measured in tension and compression states. They are defined as
Dt = 1 − exp(−t (¯ + − + +
0 )/0 ) and Dc = 1 − exp(−c (¯ − − − −
0 )/0 ) (21)

in which t and c are damage parameters associated with the two damage scalars; + 0 and

0 are threshold strains in uniaxial tension and compression, respectively, while ¯ and ¯− are
+

equivalent tensile and compressive strains, defined as


   
¯ + = (+
i )
2 and ¯ − = (−
i )
2 (22)
i=1,3 i=1,3

where +i is the positive principal strain, the superscript (+) means it vanishes if it is negative.
−
i is the negative principal strain and it vanishes if it is positive. Obviously, it can be seen

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1723

from Equation (21), when equivalent tensile strain ¯+ is larger than threshold strain + 0 in
uniaxial tension, it will produce tension-induced damage Dt , and when equivalent compressive
strain ¯− is larger than threshold strain −
0 in uniaxial compression, it will induce compression
damage Dc . Combining the tension and compression damage, the cumulative damage scalar
for brick material D can be determined as follows:

D = t Dt + c Dc , Ḋt > 0, Ḋc > 0 and t + c = 1 (23)

in which the weights t and c are defined by the following expressions:

 Hi [+ + −
i (i + i )]  Hi [− + −
i (i + i )]
t = ; c = (24)
i=1,3 ˜2 i=1,3 ˜2

+ − 2
where ˜ = i=1,3 (i + i ) is the effective strain; Hi [x] = 0 when x < 0 and Hi [x] = x when
x  0. t and c are defining the contribution of each type of damage for general loading. It
can be verified that in uniaxial tension, t = 1, c = 0, D = Dt and vice versa in compression.
Similarly, the yield strength of brick is also modelled by Drucker–Prager strength criterion as
indicated in Equation (12) and elasto-plastic constitutive model is used for brick.

4. NUMERICAL RESULTS

The above three-dimensional models for masonry and mortar are implemented into a finite
element program to simulate the stress–strain relations of the basic cell in the numerical
analysis. In the present study, the exponential forms c and t of the stress–strain curves
for mortar to describe the post-failure state in Equations (4) and (6) are assumed to be one,
implying plastic stress–strain relation after yielding. The material parameters t and c of brick
that determine its damage evolution are also assumed to be one. More detailed descriptions of
how to choose the parameter values can be found in a previous study [12].
Figure 2 shows the mesh used in the numerical model of the basic cell. In the numerical
model, the brick and mortar are discretized individually. The brick dimension which is widely
used in Australia (http://www.australbrick.com.au) is 230 × 110 × 76 mm3 and 10 mm thick
mortar joint is adopted in the analysis. Both bed and head joints are considered having the
same material properties. Table I lists the material properties of mortar and brick used in
the numerical simulation. By applying various displacement boundary conditions, a complete
monotonic stress–strain relation can be simulated through the homogenization process. The
reason to choose displacement boundaries is to avoid the incompatible deformation between
the basic cells. It should be noted that current research is to derive equivalent material properties
of masonry, not homogenized masonry wall. The results can be applied to walls of different
thickness with or without opening. Before the simulation, the influence of element size on
computational accuracy is investigated. This is carried out by halving the size of the element
for both brick and mortar while keeping loads on the basic cell unchanged until the difference
between the results obtained with two consecutive element sizes is less than 5%. The final
numerical model used in the simulation is shown in Figure 2.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1724 C. WU AND H. HAO

Figure 2. Numerical model of the masonry basic cell: (a) unit; and (b) mortar.

Table I. Material properties for brick and mortar.


(a) Brick
Ec , Et (GPa) Gc (GPa) t (MPa) c (MPa) +
0 −
0
11.0 4.4 0.25 2.6 52 0.00024 0.0047
(b) Mortar
Ec , Et (GPa) Gc (GPa) t (MPa) c (MPa) to co
5.8 2.2 0.33 1.0 14.0 0.00045 0.0024

4.1. Stress–strain relations


To have a general picture of material properties under different loading cases, response of
the basic cell under various stresses states are analysed. For compressive stress state, uniform
compressive displacements are applied to the surfaces of the basic cell and the corresponding
average stress and strain of the basic cell are estimated. Figure 3 gives the representative
homogenized stress–strain relations of the basic cell under various compressive–compressive
conditions. As shown in Figure 3(a)–(c), without lateral restraints the uniaxial compressive
strength in Z (vertical) direction is 17.8 MPa. It is the smallest as compared with 20.9 and
26.8 MPa, respectively, in X and Y directions. The reason that the uniaxial compressive strength
in Y direction (out-of-plane) is the largest is because of the influence of the joint distribution
pattern and the scale of the unit. The dimension of the basic cell in Y direction is only 76 mm,
much shorter in comparison with the other directions (240 and 240 mm). It should be noted that
the uniaxial compressive stress (17.8 MPa) of masonry in Z direction is slightly higher than that
of mortar (14.0 MPa), but far less than that of brick (52.0 MPa), indicating that compressive

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1725

(a) (b) (c)

(d) (e) (f)

(g) (h)

Figure 3. Stress–strain relation of the basic cell in compression–compression stress state.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1726 C. WU AND H. HAO

strength of masonry unit is governed by mortar joint. Using the empirical formulae proposed
in other studies, the ultimate masonry strength for the present masonry structure is 14.3 MPa
[20], 21.6 MPa [21] and 26.9 MPa [22], respectively. The present prediction is close to the
experimental result by Hendry and Malek [20]. The physical conditions of masonry specimens
used by Hendry and Malek [20] are similar to the present numerical model.
It should be noted that the size and structure effect are not considered when deriving the
homogenized material properties from the basic cell. Owing to structure effect, the out-of-plane
direction of a masonry wall could be weaker because of bending moment but the material
property in this direction is stronger. This is because the model is 3D and bending should not
bother it. The above results indicate that the out-of-plane properties of masonry are very different
from the in-plane properties. Most of previous studies did not take into consideration this and
assumed the same material properties in both X and Y directions. This could underestimate
the masonry wall capacity to resist out-of-plane loads.
The compressive strength increases under biaxial loading as shown in Figure 3(d) and (e).
Their post-failure strains are smaller than those under uniaxial loading, indicating a more brittle
failure of the basic cell. It should be noted that due to different dimensions of the basic cell in
Y and Z directions, the ratio of displacement in Y and Z directions shown in Figure 3(e) is
set to 1:3 (v:w) so that the ratio of strain boundary conditions in Y and Z directions is about
1:1. With the lateral confinements in X and Y directions, the ultimate compressive strength
in Z direction is 38.2 MPa (Figure 3(f)), which is much higher than that obtained without
lateral restraint. Similarly, if Y direction is constrained, the compressive strength in X and
Z directions also increases significantly as shown in Figure 3(g) in comparison with those in
Figure 3(d). As expected, with the confinements in the three directions, the compressive strength
of the basic cell under triaxial loading increases substantially as shown in Figure 3(h). The
compressive strength of the basic cell under triaxial confinement can reach and even exceed
the brick compressive strength.
The stress–strain curves of masonry material in compressive–tensile and tensile–tensile stress
states are also derived. Figure 4 illustrates the stress–strain relations of the basic cell in various
conditions. As shown in Figure 4(a), without lateral constraint the ultimate uniaxial tensile
strength in Z direction is 0.87 MPa, much smaller than the compressive strength of masonry.
It is also slightly smaller than the tensile strength of mortar. The uniaxial tensile strength
in Y direction reaches 1.35 MPa (Figure 4(c)), substantially higher than those in X and Z
directions, indicating that the out-of-plane tensile strength of masonry is quite different from
that of the in-plane tensile strength of masonry. The effect of confinements on the ultimate
tensile strength of masonry is shown in Figure 4(d)–(f). As shown, in both tensile–compressive
(Figure 4(e)) and tensile–tensile (Figure 4(d)) state, the ultimate tensile strength decreases. It
should be noted that under tensile–compressive state, the masonry unit fails due to tensile strain
before the compressive strength reaches the maximum value. Figure 4(f) shows that the ultimate
tensile strength in the three directions decreases under triaxial tensile loading condition.
The compressive-shear and tensile-shear stress–strain curves of the basic cell are also inves-
tigated. Figure 5 illustrates the representative stress–strain relations in different cases. It shows
that under pure shear condition, the ultimate shear strength
zx is 1.04 MPa (Figure 5(a)) and

zy is 0.88 MPa (Figure 5(b)). This demonstrates that the out-of-plane shear strength
zy of the
basic cell is smaller than the in-plane shear strength, although the out-of-plane compressive and
tensile strength of the basic cell are larger than the in-plane strength. In the compressive-shear
state (Figure 5(d)), the shear strength increases while in the tensile-shear state (Figure 5(c)),

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1727

(a) (b) (c)

(d) (e) (f)

Figure 4. Stress–strain relation of the basic cell in compressive–tension stress state.

the shear strength decreases. It is noteworthy that under compressive-shear state, the masonry
unit fails owing to shear strain before the compressive strength reaches the maximum value.

4.2. Masonry properties


The properties of masonry usually show anisotropic. They can be obtained from the simulated
stress–strain relations shown in Figures 3–5. The equivalent elastic moduli and Poisson’s ratio
of the masonry material can be calculated using the stress–strain relations corresponding to the
uniaxial compressive conditions in the three directions as
¯ xx ¯yy ¯zz
Ēxx = , ¯ xy = , ¯ xz = (25)
¯xx ¯xx ¯xx

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1728 C. WU AND H. HAO

1.2E+06 1.0E+06

u
v

τ zy (Pa)
τ zx (Pa)

u
6.0E+05 5.0E+05
v

0.0E+00 0.0E+00
0 0.0005 0.001 0 0.0005 0.001
(a) ε zx (b) ε zy
z
y

x
2.0E+06
8.0E+05
w
u τ zx

σ zz
ε zz ε zx
0.0E+00
σ zz τ zx (Pa)

u - 0.001 - 5E- 04 0 0.0005 0.001


τ zx (Pa)

w
4.0E+05
u:w=1 w
u:w=1 u

τ zx - 2.0E+06

σ zz
u
w
0.0E+00
- 4.0E+06
0 0.0002 0.0004
(c) ε zz ε zx (d) σ zz (Pa)

Figure 5. Stress–strain relation of the basic cell in compressive–tension-shear stress state.

¯ yy ¯xx ¯zz
Ēyy = , ¯ yx = , ¯ yz = (26)
¯yy ¯yy ¯yy

¯ zz ¯xx ¯yy
Ēzz = , ¯ zx = , ¯ zy = (27)
¯zz ¯zz ¯zz

Using the stress–strain relations corresponding to the three pure shear conditions, the shear
moduli can be calculated as

¯xy
¯yz
¯zx
Ḡxy = , Ḡyz = , Ḡzx = (28)
¯xy ¯yz ¯zx

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1729

J2
2.4E+07 2.4E+07 J2

F4 F3
1.2E+07 1.2E+07

F2
I1
p3 F1
I1
0.0E+00 p2 p1 pt
-1.2E+08 -8.0E+07 -4.0E+07 0.0E+00 4.0E+07 -1.2E+08 -8.0E+07 -4.0E+07 0.0E+00 4.0E+07
(a) (b)

Figure 6. Strength envelope
√ for masonry in I1 ∼ J2 space: (a) simulated
√ two-invariant points in
I1 ∼ J2 space; and (b) yield surface of masonry in I1 ∼ J2 space.

The equivalent elastic moduli of the masonry material are then obtained as Exx = 7.748 GPa,
Eyy = 8.925 GPa, Ezz = 6.412 GPa, zx = 0.263, zy = 0.275, xy = 0.283, xz = 0.293, yx =
0.305, yz = 0.308, Ḡxy = 3.424 GPa, Ḡyz = 2.781 GPa, and Ḡzx = 3.101 GPa. As shown, the
out-of-plane elastic modulus (Y direction) is the largest while the elastic modulus in the vertical
direction is the smallest among the three directions. The Poisson’s ratios are different in the
three directions due to the different distribution patterns of mortar. It should be noted that the
present in-plane results are similar to those obtained in the previous study [12].

4.3. Yield surface


The strength properties of quasi-brittle materials such as masonry are important parameters
needed in masonry structure failure modelling. Using the ultimate strength from the stress–
strain relations given in Figures 3–5, the equivalent strength envelope can be derived. The
yield strength of masonry can be modelled √by the first invariant of the stress tensor √ I1 and
the second invariant of the stress deviator J2 , i.e. two-invariant
√ model in I 1 ∼ J2 space.
Figure 6(a) shows the simulated two-invariant points in I1 ∼ J2 space. It is suggested that
masonry failure is represented with four parts as illustrated in Figure 6(b). The first part of the
model F̄1 with I1 > 0 captures masonry failure in tension. The second part F̄2 with 0 > I1 > p1
corresponds to shear failure. The third part F̄3 with p1 > I1 > p2 and the fourth part F̄4 with
p2 > I1 > p3 are associated with compressive failure. The piecewise linear parts of the strength
model can be described by

F̄i = i I1 + J2 − ki = 0, i = 1, 2, 3 (29)

From the simulated results, the parameters i , ki are derived as −0.78 and 0.89 MPa for F̄2 ,
−0.54 and 0.89 MPa for F̄2 , −0.17 and 8.1 MPa for F̄3 , and parameters p1 , p2 and p3 are
calculated as −17.1, −66.3 and −109.2 MPa, respectively. The fourth part F̄4 is a cap model
that can be expressed as

F̄4 = (I1 − p2 )2 + R 2 J2 − (p3 − p2 )2 = 0 (30)

where R is ratio of major to minor axis of cap. Using the simulated data, it is found that R is
equal to 2.91. The cap model controls the masonry failure under high compressive stress. The

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1730 C. WU AND H. HAO

cap segment of the strength model was not included in many previous studies, for example, in
the two-dimensional model [12].
It should be noted that the orthotropic non-linear behaviour of masonry usually cannot be
adequately represented by assuming a yield surface based only on the stress invariants if the
maximum strengths in different directions are greatly different. In the present study, the yield
surface is approximately estimated with the stress invariants because there are clear trends (four
parts as shown in Figure 6(b)) for two-invariant points. If the points of the two invariants are
not in clear trends, such as the case for hollow brick masonry which is currently under study,
a yield function for anisotropic materials should be employed [23, 24].
Using the simulated stress–strain relations of the basic cell, the threshold tensile strains
in three directions corresponding to the maximum tensile stresses can be obtained. They are
2.35 × 10−4 , 3.53 × 10−4 , and 2.21 × 10−4 in X, Y , and Z directions, respectively. The thresh-
old
 tensile strain of the equivalent material is then determined according to the formulae
(¯+ + + −4 in the present study. The threshold strain in the vertical
2 2 2
xx + ¯yy + ¯zz )/3, as 2.76 × 10
direction is the smallest while the threshold strain in Y direction (out-of-plane) is the largest.
This is because of the different distributions of mortar joints as discussed above. The equivalent
threshold strain can be used in the material masonry damage model.

4.4. The 3D homogenized model


Using the homogenized masonry material properties, threshold tensile strains and the yield
surface of masonry presented in the above section, a numerical model for homogenized masonry
material can be developed. The constitutive relations of the homogenized masonry material at
macro level can be derived as
ep
¯ ¯ tu )Ērskl
Ēijtu (*F̄ /*¯ rs )(*/*
Ēijkl = Ēijkl − (31)
h̄ + (*F̄ /*¯ mn )Ēmnpq (*/*¯ ¯ pq )
ep
where Ēijkl is the elastoplastic incremental material matrix; F̄ is the yield function at the macro

level which has been homogenized as a function in the I1 ∼ J2 space; ¯ is a plastic potential
function which is identical to the yield function if an associated flow rule is applied; h̄ is a
stiffness hardening factor; and Ēijkl is the damaged elastic matrix, which can be expressed as

Ēijkl = (1 − Dm )Ē0ijkl (32)

where Ē0ijkl is initial homogenized anisotropic elastic modulus matrix, Dm is damage scalar,
defined as
¯  + − ¯ + )/¯ + )
Dm = 1 − exp(−(¯ (33)
0 0

where ¯ + is the equivalent tensile strain of masonry; ¯ is a damage parameter, it is set to 0.5
based on the previous study [12]; and ¯+ 0 is the threshold strain and it is equal to 2.76 × 10
−4

as derived above.

4.5. Numerical verification of the homogenized masonry model


In the present study, boundary conditions applied to the surface of the basic cell are assumed
uniform in the homogenization process to derive the material properties. Thus the basic cell

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1731

Figure 7. (a) Continuum model; and (b) homogenized model of the 2 × 2


matrix formed by the basic cells.

should be as small as possible in the macrolevel. On the other hand, to obtain reliable results,
the basic cell should be large enough in comparison with the heterogeneity size so that the
geometry influence of the different components can be taken into consideration. Therefore, the
size effect of the basic cell, as well as the reliability of the homogenized material properties
in representing the masonry wall, should be investigated.
The continuum model formed by 2 × 2 basic cells as shown in Figure 7(a) is used to
examine the reliability of the homogenized material properties derived from the basic cell. The
continuum model means that mortar and brick are discretized and modelled separately in detail.
Three representative displacement boundary conditions, i.e. uniform compressive displacement,
uniform tensile displacement and shear-compressive displacement are applied in the present
study. Figure 8 shows the comparison of the simulated stress–strain relations of the basic cell
and the 2 × 2 basic cells of the masonry unit under different boundary conditions obtained
by the continuum model. As shown, the sizes of masonry unit have significant influence on
the mechanical properties of masonry structure. The maximum failure stress of larger size of
the masonry unit is obviously lower than that of the basic cell. It is also found from the
post-failure stage that a larger masonry unit is more brittle as compared with the basic cell.
This observation indicates the size or structural effect of the masonry wall. To eliminate this, in
deriving the homogenized material properties, the representative basic unit should be as small
as possible.
To check the reliability and accuracy of the homogenized material properties, the masonry
wall consisting of 2 × 2 basic cells is also analysed using the homogenized model as shown
in Figure 7(b). Figure 9 illustrates the stress–strain curves of the masonry wall obtained by
continuum model and homogenized model. It shows that the homogenized model gives excellent
prediction of the masonry wall response as compared with those obtained by detailed modelling
of bricks and mortar. The number of meshes used in the 3D homogenized model is far less
than those used in the continuum model as shown in Figure 7. On average, it takes 892.5 s in
an Intel Pentium 4 1.8 GHz with 512MB SDRAM memory for the continuum model to solve
the problem while it takes only 3.5 s for the homogenized model.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1732 C. WU AND H. HAO

ε zz
1.0E+06
- 0.04 - 0.02 0 2 × 2 basic cells
0.0E+00 1.0E+06 2 × 2 basic cells
Basic cell
Basic cell

w ε zz ε zx
0.0E+00
- 0.001 - 5E- 04 0 0.0005 0.001
w

σ zz (Pa)

σ zz (Pa)

σ zx (Pa)
- 1.0E+07 5.0E+05 w
u
w
- 1.0E+06

w
u:w=1
u
w
2 × 2 basic cells
0.0E+00 - 2.0E+06
Basic cell - 2.0E+07
0 0.0006 0.0012
(a) (b) ε zz (c) σ zz (Pa)

Figure 8. Comparison of the simulated stress–strain relations of two different sizes of the masonry
unit: (a) uniaxial compression; (b) uniaxial tension; and (c) compression-shear.

Continuum model Continuum model


Homogenized model Homogenized model
ε zz
- 0.02 - 0.01 0 8.0E+05
0.0E+00
σ zz (Pa)
σ zz (Pa)

4.0E+05
- 1.0E+07

0.0E+00
- 2.0E+07 0 0.0005 0.001
(a) (b) ε zz

Figure 9. Continuum model and homogenized model of masonry unit: (a) uniaxial compression for
the 2 × 2 basic cell; and (b) uniaxial tension for the 2 × 2 basic cell.

4.6. Numerical application of the homogenized masonry model


To demonstrate the efficiency of the developed homogenized material model, an 1.92 m × 1.92 m
masonry panel consisting of the above 8 × 8 basic cells of masonry under airblast loads is
analysed using the homogenized model and continuum model as shown in Figure 10(a) and (b).

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1733

Figure 10. (a) Continuum model; and (b) homogenized model of the masonry panel.

1.0E+06
1.5E+05
7.5E+05
1.0E+05
Pressure (Pa)

Pressure (Pa)

5.0E+05
5.0E+04
2.5E+05

0.0E+00 0.0E+00
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
-2.5E+05 t (s) -5.0E+04 t (s)
(a) (b)

Figure 11. Typical pressure time history of airblast loads.

The four side boundaries of the masonry panels in Figure 10(a) and (b) are fixed. The airblast
loads estimated from a previous study are used in the present study [25]. Figure 11(a) and
(b) show two typical airblast pressure time histories obtained from a surface explosion. The
predicted pressure time histories are applied on the surface of the masonry panel as input for
the analysis. The material models of mortar and brick as well as homogenized material model
for masonry including the equivalent elastic properties, strength envelope and damage threshold
are programmed and linked to an available computer program LS-DYNA3D through its user
subroutine capability [26].
Figure 12 illustrates deformation of the masonry panels at 0.025 s under airblast pressure
time history shown in Figure 11(a) by continuum model and homogenized model. The whole
masonry panels are blown off due to relatively high peak airblast pressure and relatively
short duration. As shown, the homogenized model gives good prediction of the masonry panel
response as compared with that obtained by detailed modelling of bricks and mortar. About
15 332 s is taken to solve the problem for the continuum model in an Intel Pentium 4 1.8 GHz
with 512MB SDRAM memory and only 305 s is spent for the homogenized model.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1734 C. WU AND H. HAO

Figure 12. Deformation of the masonry panels to airblast loads Type 1 at 0.025 s with:
(a) continuum model; and (b) homogenized model.

Figure 13. Deformation of the masonry panels to airblast loads Type 2 at 0.105 s with:
(a) continuum model; and (b) homogenized model.

When the peak airblast pressure is relatively low and its duration relatively long, the per-
formances of the masonry panels are quite different. Figure 13 shows the deformation of the
masonry panels at 0.105 s under airblast pressure time history shown in Figure 11(b) with
the two different models. Owing to relatively low peak airblast pressure and relatively long
duration, the masonry panels are not blown off but bounce back because of the negative air-
blast pressure. As shown, similar phenomena are observed from both continuum model and
homogenized model. It should be noted that although general performances of masonry panels
to airblast loads with both continuum model and homogenized model are similar, blown off of
bricks can be observed clearly in the continuum model but it is not accurately predicted in the

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1735

homogenized model because of the uniform material properties used in the homogenized model.
Therefore, if fragmentation is of concern, more detailed masonry model needs be developed.

5. CONCLUSIONS

Using homogenization technique, the equivalent material properties of masonry such as elastic
moduli, strength envelope and failure characteristics have been derived by numerical simulation
of the basic cell under various boundary conditions in three-dimensional space. The numerical
results indicate that the behaviour of masonry is anisotropic. The out-of-plane elastic modulus
of the masonry unit is the largest while the elastic modulus in the vertical direction is the
smallest. The failure behaviour of masonry can be characterized by tensile failure (yield surface
F̄1 ), shear failure (yield surface F̄2 ), compressive failure (yield surface F̄3 ) and high pressure
failure (cap part of yield surface F̄4 ). Based on the homogenized material properties, a 3D
homogenized model for masonry material is developed. It was then used to analyse masonry
panel response under airblast load. It is demonstrated that the 3D homogenized model can
simulate the behaviour of masonry panel to airblast load very well. The accuracy and efficiency
of the homogenized model is verified. The homogenized masonry material properties can be
used to model masonry walls of structures to blast loads.

NOMENCLATURE

¯ ij , ¯ij the components of averaged stress and strain in a basic cell


V volume of a basic cell
ij , ij stress and strain components in an element
co , to compressive and tensile yield stress for mortar
Ec , Et elastic moduli for compression and tension of mortar
co , to threshold strains for compression and tension of mortar
c parameter that controls post-failure stress–strain relations of mortar in com-
pression
t parameter that controls post-failure stress–strain relations of mortar in ten-
sion
D̄c damage variable in compression for mortar
D̄t damage variable in tension for mortar
Gcf , Gtf the total fracture energy per volume in compression and tension
Gc , Gt the fracture energy density in compression and tension
˜ ij , ˜ij stress and strain components after consideration of damage variable
E0ijkl , Ẽijkl undamaged and damaged elastic constitutive matrix
D̄ degradation damage variable
kl strain tensor
p
kl plastic strain tensor
J2 the second invariant of the stress deviator
I1 the first invariant of the stress tensor
 material constant for Drucker–Prager strength model
c0 initial material constant for Drucker–Prager strength model

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
1736 C. WU AND H. HAO

c material constant for Drucker–Prager strength model after consideration of


damage
F the yield function
 the plastic potential function for mortar
 a parameter that controls the dilatancy
dij total strain increment
deij elastic strain increment
p
dij plastic strain increment
d plastic multiplier that governs the plastic flow after yielding
ep
Ẽijkl damaged elastoplastic constitutive matrix
h the hardening parameter
Dc damage measured in compression for brick
Dt damage measured in tension for brick
c parameter that controls post-failure stress–strain relations of materials in
compression for brick
t parameter that controls post-failure stress–strain relations of materials in
tension for brick
+0 threshold strains in uniaxial compression for brick
−0 threshold strains in uniaxial tension for brick
¯+ equivalent compressive strain
¯− equivalent tensile strains
c the weight for compression
t the weight for tension
Ēxx , Ēyy , Ēzz the equivalent elastic moduli for masonry material
Ḡxy , Ḡyz , Ḡzx the equivalent shear moduli for masonry material
xy , yz , zx the equivalent Poisson’s ratio for masonry material
i , ki , p1 , p2 , p3 the parameters for the equivalent strength envelope of masonry material
F̄ the yield function for masonry material
¯ a plastic potential function for masonry material
h̄ a stiffness hardening factor for masonry material
Ēijkl the damaged homogenized anisotropic elastic matrix
Ē0ijkl initial homogenized anisotropic elastic modulus matrix
Dm damage scalar for masonry material
¯ a damage parameter

ACKNOWLEDGEMENTS
The authors would like to acknowledge the financial support from the Australian Research Council
(ARC) under grant number DP0451966 for carrying out this research work.

REFERENCES
1. Grimaldi A, Luciano R, Sacco E. Nonlinear dynamic analysis of masonry structures via FEM. In Computing
Methods in Applied Science and Engineering, Johnson FB (ed.). Nova Science Publishers: New York, 1992;
373–382.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737
DERIVATION OF 3D MASONRY PROPERTIES 1737

2. Lofti HR, Shing PB. Interface model applied to fracture of masonry structures. Journal of Structural
Engineering (ASCE) 1994; 120:63 – 80.
3. Pagnoni T. Seismic analysis of masonry and block structures with discrete element method. In Proceedings
of the 10th European Conference On Earthquake Engineering, Guma G (ed.), vol. 3. Balkema: Rotterdam,
The Netherlands, 1994; 1669 –1674.
4. Giambanco G, Rizzo S, Spallino R. Numerical analysis of masonry structures via interface models. Computer
Methods in Applied Mechanics and Engineering 2001; 190:6493 – 6511.
5. Casolo S. Modelling in-plane micro-structure of masonry walls by rigid elements. International Journal of
Solids and Structures 2004; 41:3625 – 3641.
6. Bakhvalvo N, Panasenko G. Homogenization: Averaging Processes in Periodic Media. Kluwer Academic
Publishers: Dordrecht, The Netherlands, 1989.
7. Pande GN, Liang JX, Middleton J. Equivalent elastic moduli for brick masonry. Computers and Geotechnics
1989; 8:243 –265.
8. Pietruszczak S, Niu XA. A mathematical description of macroscoptic behavior of unit masonry. International
Journal of Solids and Structures 1992; 29(5):531–546.
9. Anthoine A. Derivation of the in-plane elastic characteristics of masonry through homogenization theory.
International Journal of Solids and Structures 1995; 32(2):137–163.
10. Pegon P, Anthoine A. Numerical strategies for solving continuum damage problems with softening: application
to the homogenization of masonry. Computers and Structures 1997; 64(1– 4):623 – 642.
11. Luciano R, Sacco E. Homogenization technique and damage model for masonry material. International
Journal of Solids and Structures 1997; 34(24):3191– 3208.
12. Ma GW, Hao H, Lu Y. Homogenization of masonry using numerical simulation. Journal of Engineering
Mechanics (ASCE) 2001; 127(5):421– 431.
13. Zucchini A, Lourenco PB. A micro-mechanical model for the homogenisation of masonry. International
Journal of Solids and Structures 2002; 39(12):3233 – 3255.
14. Anthoine A. Homogenisation of periodic masonry: plane stress, generalized plane strain or 3D modeling?
Communications in Numerical Methods in Engineering 1997; 13:319 –326.
15. Dennis ST, Baylot JT, Woodson SC. Response of 1/4-scale concrete masonry unit walls to blast. Journal
of Engineering Mechanics (ASCE) 2002; 128(2):134 –142.
16. Wu C, Hao H. Dynamic response and damage analysis of masonry structures and masonry infilled RC
frames to blast ground motion. Engineering Structures 2005; 27:323 –333.
17. Lubliner J, Oliver J, Oller S, Onate E. A plastic-damage model for concrete. International Journal of Solids
and Structures 1989; 25(3):299 –326.
18. Lee J, Fenves GL. Plastic-damage model for cyclic loading of concrete structures. Journal of Engineering
Mechanics (ASCE) 1998; 124(8):892–900.
19. Chen WF, Han DJ. Plasticity for Structural Engineering. Springer: New York, 1988.
20. Hendry AW, Malek MH. Characteristic compressive strength of brickwork from collected test results. Masonry
International 1986; 7:15 –24.
21. Kirtschigg K. On the failure mechanism of masonry subject of compression. Proceedings of the 7th
International Brick Masonry Conference, Melbourne, Australia, 1985; 625 – 629.
22. Mann W. Statistical evaluation of tests on masonry by potential functions. Proceedings of the 6th International
Brick Masonry Conference, Rome, Italy, 1982; 86 – 89.
23. Barlat F, Lege DJ, Berm JC. A six-component yield function for anisotropic materials. International Journal
of Plasticity 1991; 7:693 –712.
24. Yu MH. New System of Strength Theory. Xi’an Jiaotong University Press: China, 1992.
25. Wu C, Hao H. Modeling of simultaneous ground shock and airblast pressure on nearby structures from
surface explosions. International Journal of Impact Engineering 2005; 31(6):699 –717.
26. Whirley RG, Englemann BE. DYNA3D: a nonlinear explicit three-dimensional finite element code for solid
and structural mechanics—user’s manual. Report, UCRL-MA-107254 Revision 1, Lawrence Livermore National
Laboratory, Livermore, CA, November 1991.

Copyright 䉷 2006 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2006; 66:1717–1737

You might also like