Homogenizationof Masonry Using Numerical Simulations

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245286256

Homogenization of Masonry Using Numerical Simulations

Article  in  Journal of Engineering Mechanics · May 2001


DOI: 10.1061/(ASCE)0733-9399(2001)127:5(421)

CITATIONS READS
51 572

3 authors, including:

Hong Hao Yong Lu


Curtin University The University of Edinburgh
739 PUBLICATIONS   16,221 CITATIONS    196 PUBLICATIONS   4,289 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Integrated Health Monitoring Systems for Infrastructure Structures in Operational Environments View project

Structural performance of GPC vs OPC beams View project

All content following this page was uploaded by Hong Hao on 18 August 2016.

The user has requested enhancement of the downloaded file.


HOMOGENIZATION OF MASONRY USING NUMERICAL SIMULATIONS
By Guowei Ma,1 Hong Hao,2 Member, ASCE, and Yong Lu3

ABSTRACT: Homogenization is one of the most important steps in the numerical analysis of masonry structures
where the continuum method is used. In the present study, equivalent elastic properties, strength envelope, and
different failure patterns of masonry material are homogenized by numerically simulating responses of a rep-
resentative volume element (RVE) under different stress conditions. The RVE is modeled with distinctive con-
sideration of the material properties of mortar and brick. In the numerical simulation, various displacement
boundaries are applied on the RVE surfaces to derive the stress-strain relation under different conditions. The
Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

equivalent overall material properties of the RVE are averaged by integrating the stresses and strains over the
entire area. Failure of masonry is defined by three different modes, namely, tensile failure of mortar (Mode I),
shear failure of mortar or combined shear failure of brick and mortar (Mode II), and compressive failure of
brick (Mode III). The homogenized elastic properties and failure model can be used to analyze large-scale
masonry structures.

INTRODUCTION eters in microlevel which are difficult to determine on a ma-


crolevel analysis. Thus, homogenization to obtain the equiv-
Masonry is a two-phase composite material formed by reg- alent material properties of masonry is still a challenge in a
ularly distributed bricks and mortars. Generally, two different continuum model.
models [i.e., the continuum model (Page 1978; Lotfi and Shing To predict the in-plane load-carrying and deformation ca-
1991), and the discrete model (Pagnoni 1994)], have been de- pacity of a masonry wall, knowledge of the deformation and
veloped to perform linear and nonlinear analyses of masonry failure characteristics of the masonry under biaxial stress is
structures. The latter approach of using the discrete element required. The strength of masonry in various stress states has
method is computationally intensive and is only suitable for been summarized by Hendry (1990). Genna et al. (1998) in-
simulating the fracture behavior of small laboratory specimens. vestigated the influences of adopted constitutive model to cal-
The continuum model considering the masonry material as a culate stress and plastic strain distributions of old masonry
continuum medium is applicable to analyzing a large-scale ma- buildings. It has been found that the maximum and minimum
sonry wall. However, to obtain reliable equivalent material principal stresses and the plastic zone are quite different based
properties of masonry material homogenization is most critical on different constitutive models. The strength model of ma-
in the numerical analysis. sonry is generally obtained from empirical approximation
Homogenization techniques of masonry material can be based on laboratory data (Page 1980; Hamid and Drysdale
classified into three types. The traditional homogenization of 1981; Pande et al. 1994). However, it will be very expensive
masonry is mainly based on empirical approximation by con- and almost impossible to conduct laboratory tests of masonry
sidering the effects of the volume ratios, the physical, and specimen in various physical, material, and loading conditions.
material properties of brick and mortar joint (Pande et al. Laboratory tests with the aid of numerical simulation may be
1989; Hendry 1990). With the development of homogenization an effective way to homogenize the equivalent strength prop-
theory, a homogeneous numerical model of masonry wall has erties of masonry.
been obtained by modeling the masonry as a periodic com- In this study, a typical unit of masonry is selected to serve
posite continuum (Anthoine 1995). Overall elastic moduli for as a representative volume element (RVE). The unit is mod-
one type of masonry are derived by using the homogenization eled numerically to calculate the equivalent elastic constants
theory for periodic media in conjunction with the finite-ele- and failure modes of masonry material. Both the bricks and
ment method. Although this approach is suitable to analyze mortar joints are idealized as isotropic material having their
large masonry structures, it is not adequate for detailed stress own properties such as stiffness, strength, and damage char-
analysis and for capturing the various failure mechanisms of acteristics. In the homogenization process, failure of the in-
masonry assemblages. The other homogenization technique is dividual material in the unit is divided into three modes,
based on micromechanics and microstructural concepts. A rep- namely Mode I—tensile failure of mortar joint, Mode II—
resentative volume element containing all the geometric and shear failure of mortar joint and brick, and Mode III—com-
constitutive information of the masonry is employed (Luciano pressive failure of brick. A fracturing law is associated with
and Sacco 1997; Alpa and Monetto 1994). The approach is the tensile failure of the mortar joint, while the shear failure
certainly a valuable research tool and also a viable alternative accounts for the variations of shear strength as a function of
to the costly and often time-consuming laboratory experi- normal stress (Mohr-Coulomb criterion with degrading shear
ments. However, the applications of the approach are limited strength). The brick failure is simulated by the Mazars (1986)
because the micromechanics always requires several param- damage model, which assumes that the tensile damage and
compressive damage are determined by the equivalent tensile
1
Res. Assoc., School of Civ. and Struct. Engrg., Nanyang Technol. strain and compressive strain, respectively. Based on the nu-
Univ., Singapore 639798. merical results, the equivalent material properties such as the
2
Assoc. Prof., School of Civ. and Struct. Engrg., Nanyang Technol. elastic constants and failure modes of the unit are obtained.
Univ., Singapore 639798.
3
Asst. Prof., School of Civ. and Struct. Engrg., Nanyang Technol.
They can be used as the homogenized material properties in
Univ., Singapore 639798. numerical simulation of a full masonry structure based on the
Note. Associate Editor: Jacob Fish. Discussion open until October 1, continuum approach.
2001. To extend the closing date one month, a written request must be
filed with the ASCE Manager of Journals. The manuscript for this paper RVE
was submitted for review and possible publication on November 30,
1999; revised November 14, 2000. This paper is part of the Journal of An RVE of masonry should (1) include all the participant
Engineering Mechanics, Vol. 127, No. 5, May, 2001. 䉷ASCE, ISSN materials; (2) constitute the entire structure by periodic and
0733-9399/01/0005-0421–0431/$8.00 ⫹ $.50 per page. Paper No. 22161. continuous distribution; and (3) be the minimum unit satisfy-
JOURNAL OF ENGINEERING MECHANICS / MAY 2001 / 421

J. Eng. Mech., 2001, 127(5): 421-431


FIG. 1. Homogenization of Masonry Material

ing the first two conditions. The representative volume element


provides a valuable dividing boundary between the discrete
model and the continuum model (Fig. 1). For scales much
larger than the representative volume element the continuum
Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

mechanics can be used to reproduce the material as a whole. FIG. 2. Modified Mohr-Coulomb Criterion
Under conditions of an imposed macroscopically homogene-
ous stress or deformation field on the representative volume ␴ = En εn, εn ⱕ εn0 (4)
element without displacement jump, the average stress and
⫺␣n(εn⫺εn0)/εn0
strain fields are, respectively ␴ = ␴0 e , εn ⱖ εn0 (5)


¯ ij =
1
V
冕V
␴ij dV (1)
where ␣n = material parameter; and εn0 = threshold strain that
initiates tensile fracturing of the material. ␴0 is the elastic limit


stress of the mortar joint, and it has ␴0 = En εn0. The exponen-
1 tial decay as expressed above leads to the following total frac-
ε̄ij = εij dV (2)
V V ture release energy per unit thickness of masonry:


where V = volume of the representative volume element. To ⫹⬁

ensure the continuity of the displacement field between adja- G If = ␴dεn (6)
cent RVEs, the stress vector ␴ij nj should be continuous and 0

the strains should be compatible, where nj is the normal vector The differential of fracture energy is obtained as
of the RVE surface. Based on the constitutive relations of the
brick and the mortar materials, the equivalent stress-strain re- 1 1
lations of the RVE are homogenized by applying various com- dG I = ␴dεn ⫺ d␴εn (7)
2 2
patible displacement conditions on the RVE surfaces. Aniso-
tropic elastic parameters and the failure envelope of the RVE Considering the uniaxial tensile stress-strain relation as given
can be derived based on the simulated stress-strain relations. by (4) and (5), it has
dG I = 0 when εn ⱕ εn0 (8)
MATERIAL MODEL OF MORTAR JOINT
Most of the nonlinear deformation in brick masonry, until
failure, occurs only in the joints. Establishing a reliable ma-
dG I =
1
2
␴0 冉
1⫹
␣n
εn0 冊 e⫺␣n εn⫺εn0)/εn0 dεn when εn ⱖ εn0
(
(9)

terial model for the mortar joint is very important for analyz- Thus, the failure Mode I-induced tensile damage can be de-
ing the masonry load-bearing and deformation capacity. The fined as the ratio of released fracture energy to the total frac-
mortar joint has smaller thickness than its length, which is ture release energy as
always modeled by an interface element. The constitutive re-


εn
lation can be expressed as

再 冎 冋 册再 冎
dG I
␶ Et 0 εt 0
= (3) I
␴ 0 En εn D =
G If
(10)

where ␶ and ␴ = shear stress and normal stress, respectively;


Et and En = shear and elastic moduli of mortar, respectively; Mode II Failure
and εt and εn = relative tangential and normal strains, respec- The Mohr-Coulomb yield criterion is perhaps the most com-
tively. monly used criterion for modeling rock joints, mortar joints,
In the present study, the interface model is assumed satis- and other similar interfaces. In the compression-shear region,
fying Coulomb frictional law in compressive-shear state. The it is reasonable to use the Mohr-Coulomb criterion, which can
plastic work is produced by the shear stress over the perma- be expressed as
nent shear strain of mortar. However, failure in tensile is due
to the tensile fracturing of the mortar joint. Fig. 2 shows a F(␴, ␶) = 兩 ␶ 兩 ⫹ ␮␴ ⫺ c(G II) = 0 (11)
modified Mohr-Coulomb yield criterion that distinguishes the
failure condition into the two regions, namely ␴ ⱕ 0 and ␴ where ␮ and c = frictional coefficient and cohesion, respec-
ⱖ 0. The tensile failure (Mode I) which corresponds to the tively; and ␴ = compressive stress and is negative. The co-
normal stress in the tensile region with ␴ ⱖ 0 is controlled by hesion c associated with the plastic work is degraded as shown
released fracturing energy of the joint. The shear failure (Mode in Fig. 2; G II is the dissipated plastic work that results from
II) which associates with the compressive normal stress ␴ ⱕ shear failure of mortar joint.
0 is determined by the plastic work dissipated in the joint. The The incremental strain vector of mortar can be divided into
original yield line will degrade to the final yield line with the an elastic part and a plastic part
increase of joint damage (Fig. 2). d{ε} = d{ε}e ⫹ d{ε} p (12)
T
Mode I Failure where d{ε} = {dεt, dεn} . To avoid excessive plastic dilatancy,
a nonassociated flow rule is proposed. It is expressed as
The degradation of tensile strength can be expressed by an
exponential approximation as Q(␴, ␶) = ␩ 兩 ␶ 兩 ⫹ ␮␴ (13)

422 / JOURNAL OF ENGINEERING MECHANICS / MAY 2001

J. Eng. Mech., 2001, 127(5): 421-431


where ␩ = parameter that scales the dilatancy. The direction ␶ = Et εt0 e⫺␣t(εt⫺εt0)/εt0 when εt ⱖ εt0 (25)
of plastic relative displacements is governed by the flow rule
where ␣t = material parameter; and εt0 = threshold shear strain
as
of mortar joint without lateral compression. They can be de-
⭸Q termined by laboratory test of direct shear of mortar.
d{ε} p = d␭ (14) The cohesion c is assumed to decrease depending on the
⭸{␴}
Mode I and II damage and is expressed as
where d␭ = plastic multiplier. Eq. (14) governs the plastic flow
after yielding. If the parameter ␩ is given as 1 an associated
plastic flow theory arises.
c= 冉1⫺
GI
G If
冊冉1⫺
G II
G IIf
冊 c0 = (1 ⫺ D I)(1 ⫺ D II)c0 (26)

According to the traditional plastic flow rule, the complete Thus, the hardening parameter A can be calculated with the
elastoplastic incremental stress-strain relationship is presented aid of (11), (13), (20), and (26). The incremental stiffness de-
Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

as follows: fined by (16) is obtained at each loading step.


d{␴} = [E ]epd{ε} (15) DAMAGE OF BRICK
where Damage of brick is composed of compressive crushing and
tensile splitting due to high compression. According to the
⭸Q ⭸F
[E ]e
[E ]e isotropic damage theory, the secant constitutive tensor can be
⭸{␴} ⭸{␴}T expressed as
[E ] = [E ] ⫺
ep e
(16)
⭸F ⭸Q
⫺A ⫹ [E ]e ⌳ijkl (D) = ⌳0ijkl (1 ⫺ D) (27)
⭸{␴}T ⭸{␴}
where ⌳0ijkl = initial stiffness of material; and D = damage.
The hardening parameter A is calculated according to the plas- The damage scalar consists of two parts, namely Dt due to
tic flow rule as tensile damage and Dc due to compressive damage based on
the Mazars (1986) theory. It is evaluated by the combination
1 ⭸F 1 ⭸F ⭸c
A=⫺ d{␴} = dG II (17) of
d␭ ⭸{␴}T d␭ ⭸c ⭸G II
D = At Dt ⫹ Ac Dc, D
˙ >0 (28)
The plastic work done by the shear stress ␶ depends on the
lateral compression ␴. When the combination of {␶, ␴} on the where At and Ac = balancing coefficients characterizing tension
strength surface is expressed as (11), the incremental plastic and compression, respectively. Dt and Dc corresponding re-
work is spectively to damage measured in tension and compression
can be expressed again in exponential approximation as
dG II = ( 兩 ␶ 兩 ⫹ ␮␴)dε tp = { 兩 ␶ 兩 ⫹ ␮␴ 0}{dε tp 0}T (18) ⫹
⫺ε0⫹)/ε0⫹ ⫺
⫺ε0⫺)/ε0⫺
Dt = 1 ⫺ e⫺␤t (˜ε ; Dc = 1 ⫺ e⫺␤c(˜ε (29a,b)
or where ε˜ and ε˜ = equivalent tensile and compressive strains,
⫹ ⫺

II
dG = {c p
0}d{ε} = {c} d{ε} T p
(19) respectively, and they are

Thus, with the aid of (14), the parameter A is deduced as


⭸F ⭸c ⭸Q
ε˜ ⫹ = 冑冘 3

1
⫹ 2
(ε ) ;
i ε˜ ⫺ = 冑冘 3

1
(εi⫺)2 (30a,b)

A= {c}T (20)
⭸c ⭸G II ⭸{␴} where ε⫹ i and εi = positive and negative principal strains, re-

spectively. The superscript ⫹ denotes that it vanishes if it is


The damage value at compressive-shear region can be ex- negative and vice versa for the superscript ⫺. ␤t and ␤c in (29)
pressed as are material parameters. They, as well as the threshold tensile
strain ε⫹
0 and the threshold compressive strain ε0 , can be de-

G II
D II = (21) termined from laboratory tests of uniaxial tension and com-
G IIf pression.
where The weights At and Ac are defined by the following expres-
sions:

冕 冘 冘
p
εt
i (εi ⫹ εi )]
Hi [ε⫹ i (εi ⫹ εi )]
⫹ ⫺
G = II
( 兩 ␶ 兩 ⫹ ␮␴)dε tp (22) Hi [ε⫺ ⫹ ⫺

At = ; Ac =
0 i=1–3 ε˜ 2 i=1–3 ε˜ 2
(31a,b)
and
where


⫹⬁

G IIf =
0
( 兩 ␶ 兩 ⫹ ␮␴)dε pt (23)
ε̃ = 冑冘 i=1–3
i ⫹ εi )
(ε⫹ ⫺ 2

The damage scalar D II gives the sliding status of the mortar


is the effective strain; Hi [x] = 0, when x < 0 and Hi [x] = 1,
joint. Value 0 of D II means that the joint is still in elastic phase
when x ⱖ 0. Since two different kinematics of damage are
and the shear stress does not reach the static strength of mortar,
used depending on the sign of the principal stresses, At and Ac
where value 1 of D II indicates that the mortar joint has been
are the coefficients defining the contribution of each type of
already in sliding phase.
damage for general loading. It can be verified that in uniaxial
The total dissipated energy due to friction can be calculated
tension At = 1, Ac = 0, D = Dt, and vice versa in compression.
by the relationship of the shear stress and strain with no lateral
compression on the surfaces. The relationship can also be ex- NUMERICAL SIMULATION OF RVE
pressed in an exponential form as
The material models proposed above are coded into a finite-
␶ = Et εt when εt ⱕ εt0 (24) difference program. The stress-strain relations of joint with an
JOURNAL OF ENGINEERING MECHANICS / MAY 2001 / 423

J. Eng. Mech., 2001, 127(5): 421-431


exponential form as expressed in (4) and (5) or (24) and (25) uniaxial and dimensionless stress state with ␤t = 1 and ␤c =
are illustrated in Fig. 3(a) in a dimensionless expression. It can 1. It shows that the damage extends with the increase of di-
be seen that the exponential model allows some flexibility in mensionless strain from the threshold strain. The material pa-
defining the pre- and postfailure stress-strain relation of ma- rameters of mortar joint and brick are listed in Table 1.
terial in uniaxial stress state by specifying different parameters. The physical models of RVEs in the present numerical sim-
In the present analysis, the material is considered in elastic ulations are shown in Fig. 4. The brick and mortar are discre-
state before the directional strain of the mortar joint reaches tized individually. The brick size is 225 ⫻ 75 ⫻ 125 mm and
the threshold strain. In postfailure state, the stress will decrease the mortar thickness is 10 mm. The same material properties
with increase of uniaxial strain. It is reasonable to take the for both bed and head joints are assumed in the present study.
parameters ␣n and ␣t as 1.0 and thus the threshold strains sat- In numerical simulation, vertical and horizontal displacements
isfy εn0 = ␴0 /En and εt0 = c0 /Et, where ␴0 is the uniaxial tensile are applied to the RVE surfaces. The advantage of using dis-
strength and c0 is the inherent cohesion of mortar joint. Both placement boundaries cannot only avoid incompatible defor-
Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

␴0 and c0 as well as the normal and tangential moduli En and mation between RVEs, but also makes it possible to obtain a
Et can be determined by common laboratory tests. The uniform complete monotonic stress-strain curve through the homoge-
exponential form with ␣n = ␣t = 1.0 results in the residual load nization process of (1) and (2).
resistance decrease to one-fifth of the ultimate strength when
the strain component is four times the threshold strain, and to Stress-Strain Curves in Various Conditions
zero when it is eight times the threshold strain.
The material parameters of brick can also be determined by Fig. 5 illustrates the homogenized stress-strain curves of
uniaxial tensile and compressive test data. Fig. 3(b) gives the RVEs in various compressive-compressive cases. The uniaxial
stress-strain relation and damage evolution law of material in compressive strength of the RVE in the vertical direction with-
out lateral restraint is 16.4 MPa, which is much lower than
27.3 MPa obtained when the RVE is restrained horizontally.
The horizontal strain εxx in Fig. 5(a) is not exactly zero due to
the Poisson’s ratio effect. But it vanishes when the RVE is
laterally restrained as shown in Fig. 5(b). With the increase of
the lateral confinement ratio, the ultimate strength in the ver-
tical direction can reach around 50 MPa. At the same time,
the ultimate horizontal strength also increases substantially
[Figs. 5(c and d)]. The uniaxial compressive strength in the
horizontal direction is slightly higher than that in the vertical
direction as shown in Figs. 5(a–e). This as well as the strain-
hardening phenomena in the horizontal direction is probably
due to the influence of the joint distribution pattern and the
different horizontal and vertical scales of the unit. It should be
noted that the uniaxial compressive strength (16.4 MPa) is
much lower than the brick compressive strength (52 MPa), but
is larger than the compressive strength of the mortar (14 MPa).
However, in confined conditions, the compressive strength of
the RVE increases substantially and reaches the compressive
strength of brick. These observations indicate that masonry
material exhibits plastic and pressure-dependent behavior in
the compressive-compressive stress state.
Fig. 6 compares the stress-strain curves obtained by analyzing
RVE (I) and RVE (II) in the vertical compression condition
without applying horizontal restraints. It shows that the com-
pressive strengths based on the two RVEs are almost the same.
Numerical results of the two RVEs under other boundary con-
ditions which are not shown here, are also very similar. This
indicates that the numerical results obtained from either RVE
can represent the material properties in a unit volume level.
Thus, all the subsequent calculations are performed with the
RVE (I).
Based on empirical formulas obtained from available test
data on masonry wall carried out in various countries around
FIG. 3. Dimensionless Stress-Strain Relation of: (a) Mortar Joint; (b) the world, the ultimate strength of the masonry considered in
Damage of Brick the present study is, respectively, 26.9 MPa (Mann 1982), 21.6
MPa (Kirtschigg 1985), and 14.3 MPa (Hendry and Malek
TABLE 1. Material Parameters for Brick and Mortar

E G ␴t ␴c
(MPa) (MPa) ␯ (MPa) (MPa) ε⫹
0 ε0⫺
(a) Brick
11,000 4,580 0.2 2.6 52 0.00024 0.0047
(b) Mortar
En Et ␴0 c0
(MPa) (MPa) (MPa) (MPa) ␮ εn0 εt0
2,200 1,100 1.0 1.2 0.8 0.00045 0.0011
FIG. 4. RVEs of Masonry: (a) RVE (I); (b) RVE (II)

424 / JOURNAL OF ENGINEERING MECHANICS / MAY 2001

J. Eng. Mech., 2001, 127(5): 421-431


Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

FIG. 5. Stress-Strain Curves of RVEs in Compression-Compression Stress State

the current numerical result. This is because the empirical for-


mulas obtained depend on the volume ratios, mortar types, and
brick strengths of the masonry tested.
In the tensile stress state, however, the masonry material
shows stiffness degradation characteristics rather than the plas-
ticity as in the compressive stress state. Uniform tensile dis-
placements are applied to the RVE surfaces in order to obtain
the tensile behavior of masonry material. Fig. 7 gives the
stress-strain curves in compressive- and tensile-tensile stress
states. The tensile strength in Fig. 7(a) is much smaller than
the compressive strength of masonry. The calculated uniaxial
tensile strength (about 0.85 MPa) is smaller than the uniaxial
tensile strength of mortar. This is probably due to the average
effect of the stress and strain components over the entire area
of the RVE. The lateral confinement conditions affect the ul-
timate tensile strength of the RVE as shown in Figs. 7(b–f ).
It is reasonable because of the Poisson’s ratio effect on the
tensile deformation. The horizontal tensile strength illustrated
in Fig. 7(e) is slightly smaller than the vertical one in Fig.
FIG. 6. Comparison of Stress-Strain Curves of Different Units 7(a). This is different from the compression-compression load-
ing case shown in Fig. 5, and is due again to the different joint
1986). The present prediction (16.4 MPa) is very close to the patterns and scales in the two directions of the RVE. However,
experimental result obtained by Hendry and Malek (1986) the difference of tensile strengths in the two directions is in-
which is based on test results of masonry specimens having significant, thus it is reasonable to use a unique tensile strength
similar physical conditions to the present numerical model. for masonry material. The tensile failure may dominate the
This indicates that the present numerical analysis with refined masonry failure when the structure is subjected to a complex
material models is effective and accurate in simulating the stress condition because of the poor tensile resistance.
compressive strength of masonry. It should be noted that the The compressive- and tensile-shear stress-strain relations of
compressive strengths predicted by the empirical formulas by masonry material need also be investigated since structural
Mann (1982) and Kirtschigg (1985) are somewhat higher than masonry is basically designed to resist compression and shear.
JOURNAL OF ENGINEERING MECHANICS / MAY 2001 / 425

J. Eng. Mech., 2001, 127(5): 421-431


Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

FIG. 7. Stress-Strain Curves of RVEs in Compression-Tension Stress State

Fig. 8 illustrates the stress-strain curves of RVEs in various The elastic modulus can be obtained from the three group
shear conditions. The higher the ratio of shear displacement to stress-strain curves with the following particular displacement
compressive displacement, the smaller the ultimate compres- boundary conditions: (1) ε¯ xx ≠ 0, ε¯ yy = 0, and ε¯ xy = 0; (2) ε¯ xx =
sive strength [Figs. 8(a–c)]. The ultimate compressive strength 0, ε¯ yy ≠ 0, and ε¯ xy = 0; and (3) ε¯ xx = 0, ε¯ yy = 0, and ε¯ xy ≠ 0.
decreases from the uniaxial compressive strength level as the They respectively correspond to the three cases of Figs. 5(b
shear-to-compression ratio increases. On the other hand, the and f ) and 8(d). Based on the three group data, the elastic
ultimate shear strength increases from the pure shear strength moduli can be calculated as
level to the maximum shear strength level when the shear and
␯¯ 21 = ␴¯ (2)
xx /␴ yy , ␯
¯ (2) ¯ 12 = ␴¯ (1)
yy /␴
¯ (1)
xx (33)

冉 冊冒
compressive displacements on the surfaces are the same as
shown in Fig. 8(c). The pure shear stress-strain curve is illus- ␴
¯ (2)

(1)
¯
¯ 11 = ␴¯ (1)
xx (1 ⫺ ␯
¯ 12 ␯¯ 21)/ε¯ (1)
xx = ␴ 1⫺ ε¯ (1)
xx yy
trated in Fig. 8(d). It shows that the shear strength of the RVE E ¯ (1) (34)
␴ ␴
xx (2) (1) xx
¯ yy ¯
xx

冉 冊冒
is about 1.1 MPa. In the tensile-shear state, the ultimate shear
strength decreases with the increase of the ratio of tensile-to- ␴ xx ␴
¯ (2) ¯ (1)
¯ 22 = ␴¯ (2)
yy (1 ⫺ ␯
¯ 12 ␯¯ 21)/ε¯ (2)
yy = ␴ 1⫺ ε¯ (2)
yy
E ¯ (2) (35)

¯ yy ␴
yy (2) yy
shear displacements, indicating the shear strength is reduced ¯ (1)
xx
with the presence of tensile stress in the material. ¯ = ␴¯ (3)
G ¯ (3)
xy /ε xy (36)
Table 2 lists three loading points respectively satisfying the
Elastic Properties
conditions above. The elastic moduli of the equivalent material
The masonry material is usually considered as an aniso- are then calculated as Ē11 = 7,899 MPa, Ē22 = 6,276 MPa,
tropic composite media. The elastic stress-strain relation in the ␯¯ 21 = 0.27, ␯¯ 12 = 0.31, and Ḡ = 2,884 MPa. It can be noticed
plane stress condition can be expressed as that the horizontal Young’s modulus is larger than the vertical
one. Both of them are smaller than the Young’s modulus of
¯ 11
E ¯ 11 ␯¯ 21
E brick, but they as well as the shear modulus are much larger
0 than those of mortar. The two Poisson’s ratios are different

再冎 再冎
1 ⫺ ␯¯ 12 ␯¯ 21 1 ⫺ ␯¯ 12 ␯¯ 21

¯ xx ε̄xx because of the different distribution patterns of mortar.
¯ 22 ␯¯ 12
E ¯ 22
E

¯ yy = 0 ε¯ yy (32) Strength Envelope
1 ⫺ ␯¯ 12 ␯¯ 21 1 ⫺ ␯¯ 12 ␯¯ 21

¯ xy ε̄xy
The strength properties of the equivalent material are very
0 0 ¯
G important when a continuum model is applied. Based on those
426 / JOURNAL OF ENGINEERING MECHANICS / MAY 2001

J. Eng. Mech., 2001, 127(5): 421-431


Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

FIG. 8. Stress-Strain Curves of RVEs in Compression-Tension-Shear Stress State

TABLE 2. Two Groups of Stresses and Strains in Elastic State 1


q= 兹(␴¯ 1 ⫺ ␴¯ 2)2 ⫹ (␴¯ 2 ⫺ ␴¯ 3)2 ⫹ (␴¯ 3 ⫺ ␴¯ 1)2 (38)

¯(1)
xx ␴
¯ (1)
yy ␴
¯ ␴
(2)
¯
xx
(2)
yy ␴
¯(3)
xy 兹2
ε̄
(1)
xx (MPa) (MPa) ε̄
(2)
yy (MPa) (MPa) ε̄
(3)
xy (MPa)
where ␴¯ 1, ␴¯ 2, and ␴¯ 3 are the principal stresses. Fig. 10(a) plots
⫺0.00116 ⫺10 ⫺3.13 ⫺0.00146 ⫺2.66 ⫺10 0.000312 0.9
all the strength points shown in Fig. 9 on a q-p plane. It is
illustrated clearly that the masonry failure can be divided into
three different modes as schematically illustrated in Fig. 10(b).
stress-strain curves shown in Figs. 5–8, the equivalent strength The first part with positive hydrostatic pressure p > 0 leads to
envelope can be derived by using the ultimate strength in var- Mode I failure due to mortar tension. The third part with p <
ious cases. The strength points in compressive- and tensile- p1, where the compression is very high, corresponds to brick
tensile stress states are illustrated in Fig. 9(a). The biaxial com- compressive failure. The other part with intermediate com-
pressive strength of masonry is much larger than the uniaxial pression is associated with Mode II failure induced by mortar
compressive strength, indicating that the strength of masonry and brick shear. This piecewise linear strength model is fea-
material is highly pressure-dependent. The tensile strength is, sible to the equivalent material in a continuum model. For the
as expected, much smaller than the compressive strength. The present masonry configuration and material properties, it has
strength points in compressive- and tensile-shear stresses are pt = 0.34 MPa, qc = 1.5 MPa, p1 = ⫺5.5 MPa, p2 = ⫺30 MPa,
plotted in Fig. 9(b). The uniaxial compressive strength, uni- q1 = 15 MPa, and q2 = 45 MPa.
axial tensile strength, and pure shear strength of the homog-
enized material are readily identified. Damage and Failure Modes
The q-p plot is commonly used for a geomaterial such as
rock, concrete, and soil, etc., in which p is hydrostatic stress Apparent stiffness degradation of the RVE in tensile stress
defined as state is illustrated in those stress-strain curves shown in Fig.
7. Based on those calculated stress-strain curves, the threshold
1 tensile strains of the equivalent material that correspond to the
p= (␴¯ 1 ⫹ ␴
¯2 ⫹ ␴
¯ 3) (37)
3 ultimate tensile stresses can be obtained. The average horizon-
tal, vertical, and equivalent threshold strains are 2.01 ⫻ 10⫺4,
and q is equivalent shear stress as 1.98 ⫻ 10⫺4, and 2.23 ⫻ 10⫺4. The threshold strain in the
JOURNAL OF ENGINEERING MECHANICS / MAY 2001 / 427

J. Eng. Mech., 2001, 127(5): 421-431


displacement boundary is applied to the two horizontal sur-
faces while the vertical surfaces are free. The tensile damage
is bigger than the compressive damage although the RVE is
subjected to a vertical compression. This is because of the
Poisson’s ratio effect and the much smaller tensile strength of
brick. However, the compressive damage becomes the domi-
nant factor of the brick failure when the horizontal displace-
ment of RVE is restrained [Fig. 11(c)]. The tensile damage
occurs only in the areas with large compressive damage [Fig.
11(d)]. This is reasonable because the equivalent tensile strain
is quite small in most parts of the brick due to confinement at
the horizontal direction.
Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

The tensile damage of the bed joint is illustrated in Fig.


12(a) when the RVE is loaded by a vertical outward displace-
ment at the bottom and top surfaces. It shows clearly that the
tensile failure is due to the mortar tensile damage. The brick,
however, still remains undamaged. Figs. 12(b and c) give the
plastic strain of RVE respectively loaded by a horizontal or a
vertical shear displacement at two opposite surfaces. It shows
that only the bed joint has plastic strain when RVE is subjected
to a horizontal shear; however, plastic strain develops in both
the head and bed joints when RVE is loaded by shear dis-
placement at the vertical surfaces.

SCALE EFFECT AND CONTINUUM MODEL FOR


MASONRY
Boundary conditions applied to the RVE may have signifi-
cant effects on the homogenization of materials (Suquet 1987).
FIG. 9. Strength Envelope of RVE: (a) Tensile-Tensile Stress States; Generally, two kinds of boundary conditions are used in a
(b) Tensile-Shear Stresses homogenization process: (1) uniform stress (␴ij nj = ␴¯ ij nj) or
(2) uniform strain (ui = ε̄ij xj) on the RVE surfaces. These
boundary conditions must reproduce, as closely as possible,
the in situ state of the RVE inside the material. It is immedi-
ately seen that the displacement field used in the present ho-
mogenization satisfies the uniform strain boundary condition.
Moreover, to obtain reliable results the RVE should be large
enough as compared to the heterogeneity size (microlevel), so
that the geometry effect of the different constitutes can be
included. On the other hand, the RVE must be as small as
possible in the macrolevel because the stress distribution and
displacement on the surfaces are assumed uniform in a ho-
mogenization process. Although the RVE used in the present
study satisfies the three basic requirements stated in the second
section, and is consistent with the unit employed by other re-
searchers (Anthoine 1995; Luciano and Sacco 1997), to ensure
the reliability and applicability of the homogenized properties,
size effect of the representative volume element, however,
should be examined.
A 2 ⫻ 2 matrix formed by the RVE (I) (which is denoted
as Unit I hereafter) in Fig. 4 is employed to check the size
effect of representative volume element. The bigger RVE de-
noted as Unit II can build the whole wall by periodical distri-
bution along both directions. Three special cases of displace-
ment conditions, namely uniform compressive displacement,
uniform tensile displacement, and simultaneous shear and
compressive displacements are applied on the horizontal sur-
faces of the two units. They approximately represent the stress-
FIG. 10. Strength Envelope in q-p Plane: (a) Strength Points; (b) Ma- strain states of the units in uniaxial compression, uniaxial ten-
sonry Failure Modes sion, and combined shear and compression.
The calculated uniaxial compressive stress-strain curves
vertical direction is slightly smaller than the horizontal one with respect to the two units are compared in Fig. 13(a). It is
because of the weak bonding at the bed joints. The average found that the elastic part of the two curves coincide very well,
equivalent tensile strain is slightly larger than the average hor- indicating that the elastic properties are almost independent of
izontal threshold strain due to the contribution of tensile strains the size of RVE used in homogenization analysis in the uni-
in both directions. Those average threshold strains can be used axial compression state. The initial failure stress of the Unit II
in the damage model of masonry. which is 15.1 MPa is a little bit smaller than that of the Unit
Figs. 11(a and b) illustrate respectively the contours of the I (16.4 MPa). However, strain hardening is found in the stress-
compressive and tensile damage of the RVE when compressive strain curve of the Unit II and the postfailure stress becomes
428 / JOURNAL OF ENGINEERING MECHANICS / MAY 2001

J. Eng. Mech., 2001, 127(5): 421-431


Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

FIG. 11. Compressive Failure of Brick: (a) Compressive Damage (Free Vertical Surfaces); (b) Tensile Damage (Free Vertical Surfaces); (c) Compressive
Damage (Restricted Vertical Surfaces); (d) Tensile Damage (Restricted Vertical Surfaces)

FIG. 12. Failure of Mortar Joint: (a) Tensile Failure of Mortar; (b) Plastic Strain of Bed Joint; (c) Plastic Strain of Head Joint

closer to the failure stress of the Unit I which almost has no two units are compared in Fig. 13(c). The displacement ratio
strain hardening in the postfailure stage. The stress-strain (shear displacement-to-compressive displacement) applied on
curves in uniaxial tensile state are shown in Fig. 13(b). The the two horizontal surfaces in the current case is 0.5. It is
tensile strength of Unit II is 0.73 MPa which is also smaller found that the failure stresses of compression and shear of Unit
than that of the Unit I (0.81 MPa). The postfailure part of the II is again slightly smaller than that of Unit I. This is consistent
stress-strain curve of the Unit II has more significant softening with the results of the other two stress states. The results are
than that of Unit I. However, the failure strains corresponding reasonable because Unit II has an area four times larger than
to the ultimate stresses of the two units are very close (about that of Unit I. This will introduce some geometry effect into
0.0002). This value is very important because it is used as a the average stress and strain values on RVEs. Nevertheless,
threshold to initial damage in a continuum damage model. In the above observations indicate that the influence of the RVE
compressive-shear state which is the common loading state of scale is negligible in the elastic part. The difference between
a masonry wall, the stress-strain curves homogenized for the the failure stresses of the two units is not significantly pro-
JOURNAL OF ENGINEERING MECHANICS / MAY 2001 / 429

J. Eng. Mech., 2001, 127(5): 421-431


Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

FIG. 13. Size Effect Investigation: (a) Uniaxial Compression; (b) Uniaxial Tension; (c) Compressive-Shear State

nounced either. The homogenized material properties with re- ulus is much smaller than the Young’s modulus. Failure of
spect to Unit I are acceptable to represent the macromaterial masonry can be cataloged into three types, namely tensile fail-
properties of masonry. ure due to mortar tensile damage (Mode I), shear failure of
Based on the homogenized material properties, a plastic brick and mortar (Mode II), and compressive failure of brick
damage constitutive relation of masonry in the macrolevel can (Mode III).
be derived as It should be noticed that the RVE employed in a general
homogenization process should be much smaller than the
¯ ]De¯ ⭸Q ⭸F T [E
¯ ¯
[E ¯ ]De¯ whole masonry structure. Otherwise the edge effect will affect
¯ ]D̄ ⫺
¯ ] = [E
ep e ⭸{␴}
¯ ⭸{␴}
¯ the macromaterial properties significantly, and in that case the
[E (39)
¯ ⫹ ⭸F T [E ¯ ]D̄e ⭸Q
¯ ¯ discrete element method is preferred. If the stress and strain
⫺A field in the wall varies intensively, the homogenized material
⭸{␴}¯ ⭸{␴}¯
properties may not be applicable either. Furthermore, the cur-
where [Ē ]ep = elastoplastic incremental material matrix; F̄ = rent study only considered the in-plane elastic and failure be-
yield function in macrolevel that has been homogenized in the havior of masonry; the out-of-plane effect is neglected. It is
fifth; section Q̄ = plastic potential function that is identical to not possible to predict microcracks produced in the individual
the yield function if an associated flow rule is applied; Ā = constitute by a continuum model; however, the failure pattern
stiffness hardening factor; and [E¯ ]eD̄ = damaged elastic matrix, (i.e., tensile failure), compressive failure, or combined failure
which can be expressed as of shear and compression is readily predicted.
¯ ]D̄e = (1 ⫺ D)[E
[E ¯ ¯ ]0 (40)
0
REFERENCES
where [Ē ] = initial anisotropic elastic modulus matrix. The
initial material constants in [Ē ]0 have been homogenized in Alpa, G., and Monetto, I. (1994). ‘‘Microstructural model for dry block
the fifth section. The damage scalar D̄ in (40) is presented as masonry walls with in-plane loading.’’ J. Mech. Phys. Solids, 42, 1159–
1175.

¯ ⫺¯ε⫹ ⫹
D̄ = 1 ⫺ e⫺␤(¯ε ε0
0 )/¯
(41) Anthoine, A. (1995). ‘‘Derivation of the in-plane elastic characteristics of
masonry through homogenization theory.’’ Int. J. Solids and Struct.,
where ε̄⫹ = equivalent tensile strain of masonry; ␤¯ = damage 32, 137–163.
parameter; and ε̄⫹0 = threshold tensile strain that has been cal-
Anthoine, A., Magonette, G., and Magenes, G. (1994). ‘‘Shear-compres-
culated in the fifth section. This model can be used in contin- sion testing and analysis of brick masonry walls.’’ Proc., 10th Eur.
Conf. on Earthquake Engrg., G. Duma, ed., Vol. 3, Balkema, Rotter-
uum damage analysis of masonry under arbitrary loading con- dam, The Netherlands, 1657–1662.
ditions. Genna, F., Pasqua, M. D., and Veroli, M. (1998). ‘‘Numerical analysis of
old masonry buildings: A comparison among constitutive models.’’
CONCLUDING REMARKS Engrg. Struct., 20(1–2), 37–53.
Hamid, A. A., and Drysdale, R. G. (1981). ‘‘Proposed failure criteria for
This paper has presented a numerical homogenization tech- concrete block masonry under biaxial stresses.’’ J. Struct. Div., ASCE,
nique of masonry. The elastic moduli, strength behavior, and 107(8), 1675–1687.
damage behavior have been obtained by numerical simulation Hendry, A. W. (1990). Structural masonry, Macmillan, New York.
Hendry, A. W., and Malek, M. H. (1986). ‘‘Characteristic compressive
of the representative volume element (RVE) responding to var- strength of brickwork from collected test results.’’ Masonry Int., 7, 15–
ious boundary conditions. They have been implemented into 24.
a continuum plastic damage model which can be used to an- Kirtschigg, K. (1985). ‘‘On the failure mechanism of masonry subject of
alyze a masonry structure. compression.’’ Proc., 7th Int. Brick Masonry Conf., Melbourne, Aus-
The numerical results show anisotropy of masonry. The cal- tralia, 625–629.
culated Young’s modulus in the horizontal direction is larger Lotfi, H. R., and Shing, P. B. (1991). ‘‘An appraisal of smeared crack
models for masonry shear wall analysis.’’ Comp. and Struct., 41, 413–
than the one in the vertical direction as observed in laboratory 425.
tests by other researchers. Deformation in the vertical direction Luciano, R., and Sacco, E. (1997). ‘‘Homogenization technique and dam-
is more critical than the deformation in the horizontal direction age mode for masonry material.’’ Int. J. Solids and Struct., 34(24),
because Poisson’s ratio ␯¯ 12 is larger than ␯¯ 21. The shear mod- 3191–3208.

430 / JOURNAL OF ENGINEERING MECHANICS / MAY 2001

J. Eng. Mech., 2001, 127(5): 421-431


Mann, W. (1982). ‘‘Statistical evaluation of tests on masonry by potential Engrg., G. Duma, ed., Vol. 3, Balkema, Rotterdam, The Netherlands,
functions.’’ Proc., 6th Int. Brick Masonry Conf., Rome, Italy, 86–89. 1669–1674.
Mazars, J. (1986). ‘‘A description of micro- and macroscale damage of Pande, G. N., Kralj, B., and Middleton, J. (1994). ‘‘Analysis of the com-
␣ ␤
concrete structures.’’ Engrg. Fract. Mech., 25(5/6), 729–737. pressive strength of masonry given by the equation fk = K( f ⬘)
b ( fm) .’’
Page, A. W. (1978). ‘‘Finite-element model for masonry.’’ J. Struct. Div., The Struct. Engr., London, 72(1), 7–12.
ASCE, 104(8), 1267–1285. Pande, G. N., Liang, J. X., and Middleton, J. (1989). ‘‘Equivalent elastic
Page, A. W. (1980). ‘‘A biaxial failure criterion for brick masonry in the moduli for brick masonry.’’ Comput. Geotech., 8, 243–265.
tension-tension range.’’ Int. J. Masonry Constr., 1(1), 26–29. Suquet, P. M. (1987). ‘‘Elements of homogenization for inelastic solid
Pagnoni, T. (1994). ‘‘Seismic analysis of masonry and block structures mechanics.’’ Homogenization techniques for composite media, E. San-
with discrete element method.’’ Proc., 10th Eur. Conf. on Earthquake chez-Palencia et al., eds., Springer, New York.
Downloaded from ascelibrary.org by CURTIN UNIV OF TECHNOLOGY on 08/18/16. Copyright ASCE. For personal use only; all rights reserved.

JOURNAL OF ENGINEERING MECHANICS / MAY 2001 / 431

View publication stats J. Eng. Mech., 2001, 127(5): 421-431

You might also like