Vous êtes sur la page 1sur 8

Modeling Fracture in Masonry

Mario M. Attard1; Alfonso Nappi2; and Francis Tin-Loi3

Abstract: A finite element procedure developed for the study of fracture in concrete is extended for the simulation of tensile and/or shear
fracture in masonry. Triangular units are grouped into rectangular zones mimicking brick units with surrounding mortar joints. Fracture is
captured through a constitutive softening-fracture law at the boundary interface nodes. The mortar joint, which is a plane of weakness, can
be modeled as an interface of zero thickness or of a given thickness. At each nodal location, there exist essentially two nodes, the relative
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

displacement 共i.e., crack opening or sliding兲 of which is related to the conjugate internodal force by the appropriate softening relationship.
The model is ideally suited to the modeling of fracture in masonry because fracture usually runs along a horizontal or vertical joint in the
mortar or is approximately vertical in the brick unit. The inelastic failure properties are divided into those for the mortar joints and those
for fracture within the brick units. The inelastic failure surface is modeled using a Mohr–Coulomb failure surface with a tension cut-off.
Examples which include: Direct tension, microshear, and three-point bending of masonry panels are used to verify the formulation.
DOI: 10.1061/共ASCE兲0733-9445共2007兲133:10共1385兲
CE Database subject headings: Cracking; Fracture; Masonry; Softening; Finite elements; Concrete.

Introduction Tin-Loi 共1999, 2000兲 also used a discrete formulation for the
modeling of masonry and were able to describe some essential
The need to preserve and protect historical structures built using features such as joint slip, and brick cracks.
masonry has seen the emergence of research directed toward the Attard and Tin-Loi 共1999, 2005兲 developed a discrete element
analytical modeling of fracture in masonry 共see, e.g., Page 1978, representation based on a particle/interface type model for the
1980; Lourenço 1996; Lourenço and Rots 1997; Lourenço et al. simulation of fracture/cracking in concrete. The approach used a
1997, 1999; Sutcliffe et al. 1999; Nappi and Tin-Loi 1999, 2000; linear complementarity format and an enumerative mathematical
Guinea et al. 2000; van Zijl 2001; Giambanco et al. 2001兲. The programming algorithm to obtain equilibrium solutions to a non-
fracturing process in masonry is complex because masonry is the holonomic rate formulation. When a bifurcation was encountered,
composite of two brittle materials, the brick unit and the mortar the equilibrium path with the minimum incremental external work
joint, which can have very different material properties. was chosen as the critical path. The constitutive law was a single
The weakness element in a masonry panel is typically the branch softening law written in terms of forces and displace-
mortar joints but fracture can also occur in the brick units. Be- ments. The advantages of this formulation were that, as with the
cause the joints follow a pattern of crisscrossing horizontal and discrete crack models, there was no length scale required. The
vertical lines, fracture in the joints which follow the crisscrossing formulation also allowed the tracking of interacting and/or
pattern, may involve tension induced cracks and sliding and dila- branching cracks without remeshing. The formulation of Attard
tancy induced by shear. Giambanco et al. 共2001兲 discusses the and Tin-Loi 共1999, 2005兲 is extended in this paper for the simu-
two main approaches used for the analysis of fracture in masonry; lation of fracture in masonry under tension and/or shear.
the macromodeling approach and the micromodeling approach. The basic unit in the formulation is a triangle formed by as-
The analytical work described in this paper falls under the cat- sembling nine constant strain triangles and condensing out the
egory of micromodeling, where masonry is modeled as a discon- freedoms at the vertices 关see Fig. 1共a兲兴. There are two nodes on
tinuous assembly of blocks connected to each other by interface each of the three sides/interfaces of the triangular unit. The posi-
joints. The formulation can be classified as discrete. Nappi and tion of the interface nodes is set at Li / 2n from the vertices, where
Li⫽interface length 共length of the side of the triangular unit兲 and
1 n an integer here set to 10 关refer to Fig. 1共a兲兴. The material within
Associate Professor, School of Civil and Environmental Engineering,
the triangular unit remains linear elastic if the inelastic constitu-
Univ. of New South Wales, Sydney 2052 Australia. E-mail: m.attard@
unsw.edu.au tive relationship for the interface forces is softening. Masonry is
2
Professor, Dept. of Civil Engineering, Univ. of Trieste. E-mail: modeled by combining the triangle units to form a brick with the
nappi@univ.trieste.it interfaces around the brick representing the mortar interface. The
3
Professor, School of Civil and Environmental Engineering, Univ. of mortar interface is then assumed to have zero thickness 共the mor-
New South Wales, Sydney 2052 Australia. E-mail: f.tinloi@unsw.edu.au tar interface can also be modeled with a definite thickness兲 and is
Note. Associate Editor: Khalid M. Mosalam. Discussion open until assigned the inelastic constitutive properties of the mortar. The
March 1, 2008. Separate discussions must be submitted for individual interior interfaces of the brick have the inelastic constitutive prop-
papers. To extend the closing date by one month, a written request must
erties of the brick. A mortar of a given thickness can also be
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on July 22, 2003; modeled. Fig. 1共b兲 shows the simplest model for the masonry
approved on April 9, 2007. This paper is part of the Journal of Struc- unit.
tural Engineering, Vol. 133, No. 10, October 1, 2007. ©ASCE, ISSN Generalized interface displacements are defined which corre-
0733-9445/2007/10-1385–1392/$25.00. spond to the outward normal and tangential 共counterclockwise

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007 / 1385

J. Struct. Eng. 2007.133:1385-1392.


plete structure, a piecewise linear assemblage of the inelastic fail-
ure surfaces is adopted following the procedures in Maier 共1970兲.
For the nodal inelastic failure surface shown in Fig. 2,
␭i⫽interface multiplier vector 共analogous to plastic multipliers
used in plasticity兲, defined at an interface node as

冤冥
␭n
␭i = ␭s1 共1兲
␭s2
where ␭n⫽multiplier associated with the normal interface inelas-
tic tensile displacement activated by tension failure; ␭s1 and
␭s2⫽multipliers associated with sliding and normal interface in-
elastic displacements when the Mohr–Coulomb portions of inter-
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

face inelastic failure surface is activated; Qi⫽resultant interface


force vector at the interface node; Qny⫽inelastic failure normal
force; and Qsy⫽inelastic failure shear force defined by the normal
to the Mohr–Coulomb surface. The interface normality matrix Ni
and the dilatancy matrix Vi are given by

Ni = 冋0
1
cos ␾
sin ␾
− cos ␾
sin ␾
, 册
Fig. 1. Modeling of masonry units
Vi = 冋 0
1
cos ␺
sin ␺
− cos ␺
sin ␺
册 共2兲

defined as positive兲 displacements at the interface nodes. The con- The normal inelastic failure force Qny is estimated from the prod-
jugate generalized forces are the outward normal Qn and shear uct of the material inelastic failure tensile stress f t obtained from
force Qs at the interface nodes 关see Fig. 1共a兲兴. a pure tension test, with half the interface length Li and the panel
thickness ti given by

Inelastic Constitutive Law f tL it i


Qny = 共3兲
2
At the level of the interface nodes, the inelastic failure surface is This is an average measure of the inelastic failure capacity since
a function of the normal and shear interface forces. The inelastic the traction along the interface may not be uniform.
failure surface for the interface forces is taken as a linear Mohr- The normal distance from the origin to the Mohr–Coulomb
Coulomb failure surface with a tension cut-off as depicted in Fig. inelastic failure surface is denoted by the force Qsy and defined by
2. The angle ␾⫽friction angle, whereas ␺⫽dilatancy angle. The
cohesion is represented by the symbol “c”. In the present formu- f ⬘c L it i
lation, the friction angle and dilatancy angle are taken as constant Qsy = 共1 − sin ␾兲Liti = c cos ␾ 共4兲
4 2
throughout the damage process. The inelastic failure surface is
defined in terms of interface forces as opposed to conventional where f ⬘c ⫽compressive strength of the interface material and
continuum models, which are based on stresses. No compression c⫽cohesion. Note that Qsy represents a combination of shear and
cap is considered in this paper. The flow rule is nonassociated some tension depending on the friction angle. It is used to define
when the friction and dilatancy angles are different. For a com- the Mohr–Coulomb failure surface. The inelastic failure forces
are grouped into an interface node failure force vector ri given by

冤 冥
Qny
ri = Qsy 共5兲
Qsy
When the interface force at a node Qi intersects the failure
surface either at the tension cutoff plane or the Mohr–Coulomb
plane, the activated inelastic constitutive law is taken as a single
branch softening curve. Fig. 3 shows a single branch softening
law for the normal interface force. The multiplier ␭n can only
have positive values along the descending branch. Elastic unload-
ing is allowed from the descending branch. Once the critical crack
opening displacement ␭nc is reached, the crack can open and close
freely unless the cracked surfaces come into contact. No previ-
sions for cracks coming back into contact have been made in this
work. The critical crack opening displacement used with the
single branch softening constitutive law is approximated from the
Fig. 2. Mohr Coulomb with tension cut-off inelastic failure surface Mode 1 fracture energy GIf by

1386 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007

J. Struct. Eng. 2007.133:1385-1392.


tancy matrices N and V, respectively, and an initial inelastic fail-
ure load vector r are all assembled for the complete structure.
Inelastic failure is activated if the structure interface generalized
force vector Q intersects the assembled structure inelastic failure
surface. This can be ascertained by examining the projections of
Q in all phases and must therefore satisfy the following condition:
NTQ 艋 r + H␭ = ␰

␸ = NTQ − r − H␭ 艋 0

=␣NTQe − r + 共NTZV − H兲␭ 艋 0

=␣NTQe − r + W␭ 艋 0 共10兲
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Single branch softening law


in which ␰⫽inelastic failure surface evolution vector; ␭⫽vector
of inelastic multipliers; ␣⫽loading parameter; Qe⫽vector of unit
1.5GIf elastic generalized forces; Z⫽structural matrix relating the inelas-
␭nc = 共6兲 tic generalized forces to the inelastic multipliers 共see Attard and
ft
Tin-Loi 2005兲; and ␸⫽potential function vector. If the ith inelas-
When the Mohr–Coulomb failure surface is activated the inelastic tic failure plane is active then the corresponding ith element of
constitutive law is defined with respect to the normal to the the potential vector ␸ is zero. The following constraints must also
Mohr–Coulomb plane. A single branch softening law is again be satisfied:
applied which represents a cohesion loss until the critical sliding
displacement ␭sc is reached after which the damaged zone main- ␭ 艌 0, ␸ T␭ = 0 共11兲
tains a residual friction level. The critical sliding displacement is Initially, only positive or zero values for the inelastic multipliers
approximated using the Mode 2 fracture energy GIIf by are acceptable 共this constraint will be relaxed for those multipliers
1.5GIIf which have exceeded the critical value兲. Eqs. 共10兲 and 共11兲 can be
␭ns = 共7兲 reset in a finite incremental nonholonomic form in a quasi-
c
prescribed displacement format which can be expressed as 共see
The evolution of the inelastic failure surface at an interface node Attard and Tin-Loi 2005 for details兲
is represented by an interface hardening 共softening兲 matrix Hi
defined by 0 艋 z = p + Mx ⬜ x 艌 0

冤 冥
Qny Qny Qny z = − ⌬␸, x = ⌬␭
− − ␤ − ␤
␭nc ␭ns ␭ns
Qsy Qsy p = − ⌬␩NTQe, M=W 共12兲
Hi = − ␤ − 0 共8兲
␭nc ␭ns where ⌬␩⫽increment in a displacement scaling parameter.
Qsy Qsy Complementarity is enforced between the function vector ⌬␸ and
− ␤ 0 − the increment of the multiplier vector ⌬␭. The quasiprescribed
␭nc ␭ns
displacement format is such that the load vector is maintained.
where ␤⫽interaction parameter. The off-diagonal terms in the The problem is solved in incremental steps as a series of linear
previous matrix represent interaction between softening/damage complementarity problems. At each event, a set of active multi-
of the normal interface force and the interface shear capacity. If ␤ pliers is maintained and updated, with unloading inelastic failure
is taken as unity the inelastic failure surface contracts in an iso- points removed from the active set. The linear complementarity
tropic manner. The evolution of the interface inelastic failure sur- problem, Eq. 共12兲 is solved for those active multipliers on the
face can be represented by the vector ␰i given by descending branch of the constitutive relationship. For the set of
multipliers which have reached either the critical crack opening
␰ i = r i + H i␭ i 共9兲
displacement or critical shear sliding displacement, a coupled set
of questions are solved 共see Attard and Tin-Loi 2005兲.
Unique solutions are guaranteed if the M matrix is positive
Solution Algorithm definite and standard algorithms such that of Lemke 共1965兲 can
be used to solve for such instances. If the M matrix is not positive
To track the inelastic displacements such as the opening and clo- definite then a multiplicity of solutions could be possible indicat-
sure of cracks and/or inelastic sliding due to shear, the path- ing a bifurcation has been reached. Either an enumerative proce-
dependent softening behavior in rate or finite incremental form dure to solve the LCP is used, as described in Bolzon et al. 共1997兲
共Bolzon et al. 1995, 1997兲 will need to be solved. A more detailed or for problems which involve a large number of variables either
description of the algorithm used here can be found in Attard and the strategy outlined in Tin-Loi and Tseng 共2003兲 could be em-
Tin-Loi 共2005兲. Only a brief description will be given here. For ployed or a single event by event strategy as described below
the complete structure such as a masonry panel or wall, the inter- could be employed. When multiple solutions are detected, the
face generalized forces are collected into a structure generalized equilibrium solution which provides the minimum increment in
force vector Q. Similarly, a structure hardening matrix H, the external work should be taken as the critical solution. This ap-
vector of inelastic multipliers ␭, the structure normality and dila- proach becomes important when investigating softening and es-

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007 / 1387

J. Struct. Eng. 2007.133:1385-1392.


Table 1. Material Properties for Brick Wall Tension Simulation 共Adapted
from Lourenço 1996; Lourenço et al. 1999兲
Property Mortar Brick
Tensile inelastic failure stress 共MPa兲 0.5 1.3
Critical tensile opening displacement 共mm兲 0.04 0.104
Shear inelastic failure stress 共MPa兲 0.60 1.56
Critical shear opening displacement 共mm兲 0.167 0.433
Friction angle ␾ 共degrees兲 36.9 36.9
Dilatancy angle ␺ 共degrees兲 0, 11.3 0, 11.3
Elastic modulus 共MPa兲 — 5,000
Fig. 4. Masonry detail for infinitely long wall under tensile loading
Poisson ratio — 0.2
parallel to the bed joint
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

pecially for interacting and/or branching cracks. If one restricts


the solution algorithm to only single events at each step, the num-
ber of possible multiple solutions to the equilibrium path is re- failure zone to that used by Lourenço 共1996兲, whereas Case B
duced because simultaneous inelastic failure has been excluded. provided no restriction on which interfaces could suffer inelastic
This is called a single event by event strategy. If one included behavior. Lourenço 共1996兲 and Lourenço et al. 共1999兲 provided
imperfections as in a real structure, it could be argued that simul- numerical results using continuum elements for the brick units
taneous inelastic failure produced by symmetry would not be and interface elements for the mortar joints.
possible as in the real world. When several locations within a Figs. 5 and 6 show the results for Case A. Tensile inelastic
structure simultaneously reach inelastic failure, the minimum failure occurred along the vertical mortar joints followed by shear
work would be done if only one softens and the rest elastically inelastic failure along the horizontal joint. Figs. 5共b and c兲 show
unload 共excluding the solution where all elastically unload兲. For the elastic displacements of the final failure mechanism. The
large problems the single event by event strategy is adopted here. stress versus displacement results in Fig. 6 show a qualitative
agreement with 共Lourenço 1996兲 although the present results
have one more peak. Fig. 5 shows the progression of inelastic
Numerical Examples failure points within the finite element mesh for several points
along the load–deflection curve. Initially, the inelastic failure zone
Three examples are presented which demonstrate the application is activated in a symmetric pattern with tensile inelastic failure
of the proposed model for the analysis of fracture in masonry. The along the two vertical mortar interfaces. One of the two interfaces
first example deals with tension parallel to a bed joint but in- then dominates with unloading from the other. This asymmetric
volves both tension cracking and cohesion/shear degradation. The behavior is similar to that identified in a two notch tensile test
second example explores the capability of the model to handle when the loading platens are allowed to rotate. The tensile crack
shear under different confining pressures. The last example looks in one of the vertical mortar joints then propagates through the
at results for the three point bending of a masonry panel with vertical interface until it reaches the horizontal interface and the
either weak or strong mortar and demonstrates the ability of the cracks open reaching the critical crack opening displacement.
algorithm to track complex crack paths. Tensile cracking then propagates through the other vertical inter-
face. Once both vertical interfaces have cracks opening freely,
shear inelastic failure is activated along the horizontal interface.
Tensile Behavior Parallel to Bed Joint There are three peaks in the stress versus displacement results.
The first two corresponding to inelastic failure and snap back
Lourenço 共1996兲 and Lourenço et al. 共1999兲 examined a masonry associated with the vertical interfaces. The third peak is associ-
panel detail used to model an infinitely large wall under tension as ated with inelastic failure along the entire horizontal interface.
shown in Fig. 4. Essentially a panel consisting of two and half Lourenço’s results had only two peaks as his results predicted a
bricks in length and two half bricks in height is loaded by axial symmetric propagation of the tensile cracks in the vertical inter-
tension at one end. The axial loading is parallel to the bed joint. faces. Lourenço 共1996兲 also incorporated dilatancy effects, which
Lourenço 共1996兲 allowed inelastic failure only along the mortar produce continuing hardening of the stress versus displacement
interfaces at the midsection region 共see Fig. 4兲. Fig. 4 also shows relationship. The results using the proposed algorithm for a dila-
the mesh used for the present formulation with each brick unit tancy angle of 0 and 11.3° compare reasonable well with
modeled using a ten by three grid. Rollers were used along the top Lourenço’s results.
and bottom boundaries and a rigid boundary was placed at the To demonstrate the versatility of the proposed method, Case B
opposite end to the loading. The masonry units were 900⫻ 600 had no restriction on the potential inelastic failure zone and as
⫻ 100 mm calcium silicate units with thin glued joints. The bed stated earlier, several of the material properties for the brick units
joints were approximately 2 mm thick, whereas the head joints were scaled from the mortar properties. The weakest element of
were approximately 3 mm thick. The mesh shown in Fig. 4 as- masonry unit was the mortar joint. Figs. 7 and 8 show the stress
sumes a zero thickness for the mortar joint. The material proper- versus displacement results and the ultimate failure mode, respec-
ties for the mortar and brick unit are listed in Table 1. Only the tively. The failure mode included tensile inelastic failure along
tensile strength of the brick unit was quoted in Lourenço 共1996兲 three vertical mortar interfaces and tensile inelastic failure
and hence several properties for the brick unit have been scaled through the brick unit. No shear plane inelastic failure was acti-
using the ratio of the brick to mortar tensile strength. Two cases vated. The four peaks in the stress versus displacement results
have been investigated. Case A restricted the potential inelastic correspond to failure along the vertical interfaces.

1388 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007

J. Struct. Eng. 2007.133:1385-1392.


Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Case A—comparison of stress versus displacement


predictions

9兲. A vertical distributed confining force was applied to the rigid


top boundary. The masonry specimen had initial confining
stresses of 0.1, 0.5, or 1.0 MPa. Forces were applied at one end of
the rigid element which produced a idealized state of pure shear at
the midsection of horizontal mortar interface. Inelastic behavior
was only permitted along the mortar interface. The brick units
measured 200 mm in length, 50 mm in height, and 100 mm in
thickness. The mortar bed was approximately 15 mm thick. A grid
of 32⫻ 6 was used for the brick units, whereas the mortar was

Fig. 7. Case B—comparison of stress versus displacement


Fig. 5. Case A—masonry panel under tension predictions

Mortar Interface under Shear and Confinement

van der Pluijm 共1992, 1993兲, designed a shear test rig to subject a
mortar joint between two brick units to pure shear under different
levels of confining stress. The main objective was to trace the
average shear mortar bond stress as a function of the shear dis-
placement across the mortar interface. The test rig consist of ac-
tuators to apply the normal confining stress and actuators which
loaded the brick unit with a combination of bending and shear
producing pure shear on the mortar joint. Fig. 9 shows the finite
element model adopted here to simulate van der Pluijm’s micros-
hear test. The top and bottom test loading frames were modeled
using rigid elements along the top horizontal boundary and by Fig. 8. Case B—failure mode and final deformation profile of
providing a fixed horizontal boundary along the bottom 共see Fig. masonry panel under tension

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007 / 1389

J. Struct. Eng. 2007.133:1385-1392.


Fig. 9. Simulation model for the microshear test of van der Pluijm Fig. 10. Scaled failure mechanism
共1993兲
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

modeled using a grid of 32⫻ 3. The material properties for the because of the use of a single branch softening constitutive law.
masonry components are listed in Table 2. Experimental results At the peak stress, all of the mortar interface nodes along the top
for the mortar and brick combination identified as JG-B were layer of the mortar joint have activated the Mohr–Coulomb fail-
given in van der Pluijm 共1993兲. An average shear stress at the ure surface and are softening at the same rate, hence on a struc-
horizontal mortar interface was calculated and plotted against an tural level the softening behavior of the model would appear
average shear displacement based on the displacements at the linear. The present model could be improved by using a bilinear
gauge points indicated in Fig. 9. softening constitutive law. As well, the present model uses a con-
For the numerical model, as the load was increased, a zone of stant friction angle and dilatancy angle. The model may give a
inelastic shear failure spread throughout the horizontal layers of better comparison with the experimental results if the normality
the mesh representing the mortar joint. As the inelastic shear zone and dilatancy matrices defined in Eq. 共2兲 were functions of the
moved across the specimen from left to right, the top layer re- evolution of the inelastic failure surface.
mained active, whereas the middle and bottom layers unloaded.
At the peak load, inelastic shear failure had activated the interface
of the top layer of the mortar joint. The critical shear displace- Three-Point Bending of a Masonry Panel
ment for the interface nodes was reached almost simultaneously
because all nodes essentially had the same shear displacement. At Guinea et al. 共2000兲, produced two small-scale masonry panels
a structural level, the branching from softening to the residual using small scale bricks approximately 48 mm long
friction level therefore appears almost instantaneously. Fig. 10 ⫻ 10 mm in height⫻ 25 mm thick. The panels were constructed
shows the scaled deformation mode at failure with the top layer of using nine courses of ten brick units with a total height of ap-
the mortar being sheared and expanding out as dilatancy was no proximately 112 mm. The dimensions of the panels are shown in
hindered. Fig. 11 compares the experimental results of van der Fig. 12. A midspan notch of approximately 39 mm was sawn. The
Pluijm 共1993兲 with the results of the proposed formulation for notch extended through a mortar joint then through a brick unit
varying initial confining stresses of 0.1, 0.5, and 1.0 MPa. Al- and finally through a second mortar joint. The top of the notch
though the proposed model compares reasonably with van der was below a brick unit. The panels were used for a three point
Pluijm’s experimental results, a major difference is in the shape of bending test 共refer to Fig. 12兲.
the postpeak curve is evident. van der Pluijm’s results show a Prisms and cubes made from the same batched mortar and
nonlinear softening behavior, whereas the proposed model shows from the parent brick material were used to obtain base material
an almost linear softening behavior. This can partly be explained data including the splitting tensile strength and compression

Table 2. Material Properties for Microshear Test Simulation 共Van Zijl


1996兲
Property Mortar JG-B Brick JG
Tensile inelastic failure 0.62 —
stress 共MPa兲
Critical tensile opening 0.0387 —
displacement 共mm兲
Shear inelastic failure stress 0.673 —
共MPa兲
Critical shear opening ␴ = 0.1 MPa 0.0525 —
displacement 共mm兲
␴ = 0.5 MPa 0.09
␴ = 1.0 MPa 0.137
Friction angle ␾ 共degrees兲 47.7 —
Dilatancy angle ␺ 共degrees兲 26.9 —
Elastic modulus 共MPa兲 6,035 16,700 Fig. 11. Comparison of microshear simulation with experimental
Poisson’s ratio 0.28 0.28 results of van der Pluijm 共1993兲

1390 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007

J. Struct. Eng. 2007.133:1385-1392.


Fig. 12. Dimensions of three-point bending test 共Panel P4 from
Guinea et al. 2000兲
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

strength, and the Mode 1 fracture energies for both the mortar and
brick unit. Table 3 lists the material properties used in the simu-
lation here. In Table 3, Mortar A is based on the mortar used by
Guinea et al. 共2000兲. The mortar and brick units used by Guinea Fig. 13. Comparisons with the three-point bending test results of
et al. 共2000兲 have similar strengths with the mortar having a ten- Guinea et al. 共2000兲
sile strength of approximately 73% of the brick units. The failure
mode therefore consisted of a single vertical crack penetrating the
panel through the midspan region above the notch 关see Fig. tensile strength than the mortar joint, the damage zone within the
14共a兲兴. The fracture was primarily Mode 1. panel for the weaker mortar types, B and C, is far broader than for
To simulate the three point bending test, the brick units were Mortar Type A where the damage zone is highly localized
modeled using a 4 ⫻ 4 grid. The mortar interface was modeled throughout the loading history. During the simulation for Mortar
with zero thickness. The elastic modulus for the composite was Types B and C a large number of mortar joints became inelastic at
selected from a back fit to the experimental results and was taken some stage during the loading but unloaded as the damage zone
as 18,000 MPa. A comparison of the simulation for Mortar Type progressed through the panel. Before the fracture penetrated the
A with the test results of Panel P4 of Guinea et al. 共2000兲 is brick units there is an initial shear failure in the mortar joint with
shown in Fig. 13. The comparison shows generally the same lin- sliding on the horizontal mortar joint above and below the brick
ear and postpeak response although the simulation did not have unit, as well as the mortar joint on either side of the brick unit.
the second peak evident in the postpeak response of the test Once the crack penetrates the brick unit and progresses into the
results. vertical mortar joint above the brick, there is unloading of the
To demonstrate the versatility of the proposed formulation to inelastic zones around the brick. The load deflection result for
track branching and interacting cracks as well as shear failure in Mortar Type B has several peaks along the postpeak path. The
the joints, the simulation was run with two further mortar defini- secondary hardening in the postpeak path are generally associated
tions. Mortar B was a weak mortar based on the mortar used in
the simulation of the panel under tension parallel to the bed joint
by Lourenço 共1996兲. Mortar C was the same as Mortar B but with
reduced shear strength. Table 3 gives the material properties for
the mortars designated as Mortar B and Mortar C. Fig. 13 com-
pares the load versus load point deflection of simulations results
for Mortars B and C with the test results of Guinea et al. 共2000兲.
Figs. 14共b and c兲 show the elastic displacements produced by the
numerical simulation. Because the brick unit has a much higher

Table 3. Material Properties for Three-Point Bending Test 共Guinea et al.


2000兲
Mortar Mortar Mortar
A B C Brick
Tensile inelastic failure 3.3 0.5 0.5 4.5
stress 共MPa兲
Critical tensile opening 0.02 0.04 0.04 0.0262
displacement 共mm兲
Shear inelastic failure stress — 0.60 0.20 —
共MPa兲
Critical shear opening — 0.167 0.167 —
displacement 共mm兲
Friction angle ␾ 共degrees兲 — 36.9 36.9 —
Dilatancy angle ␺ 共degrees兲 — 11.3 11.3 —
Elastic modulus 共MPa兲 — — — 18,000
Poisson’s ratio — — — 0.2 Fig. 14. Elastic displacements at failure

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007 / 1391

J. Struct. Eng. 2007.133:1385-1392.


with the crack reaching the base of a brick unit and the inelastic of Concrete Structures, FRAMCOS, Vol. 2, F. H. Wittmann, ed.,
failure zones penetrating around the brick unit dissipating energy Freiburg, 885–898.
through sliding in the horizontal mortar joint and tension in the Bolzon, G., Maier, G., and Tin-Loi, F. 共1997兲. “On multiplicity of solu-
vertical joint. Softening from the secondary peaks are generally tions in quasi-brittle fracture computations.” Comput. Mech., 19,
associated with the crack having penetrated the brick units. The 511–516.
Giambanco, G., Rizzo, S., and Spallino, R. 共2001兲. “Numerical analysis
final failure mode for Mortar Type B involved a straight crack
of masonry structures via interface models.” Comput. Methods Appl.
through the mid of the panel above the notch to the load point.
Mech. Eng., 6494–6511.
The simulation for Mortar Type C showed an asymmetric failure Guinea, G. V., Hussein, G., Elices, M., and Planas, J. 共2000兲. “Microme-
mode with, as shown in Fig. 14共c兲 much more shear failure in the chanical modeling of brick-masonry fracture.” Cem. Concr. Res., 30,
mortar joints. The load deflection response for Mortar Type C 731–737.
showed a residual strength in the postpeak behavior associated Lemke, C. E. 共1965兲. “Bimatrix equilibrium points and mathematical
with the residual friction strength being reached in the horizontal programming.” Manage. Sci., 11, 681–689.
mortar joint. It must be remembered however that the present Lourenço, P. B. 共1996兲. “Computational strategies for masonry struc-
tures.” Ph.D. thesis, Delft Univ. of Technology, Delft, The
Downloaded from ascelibrary.org by University of Waterloo on 12/16/14. Copyright ASCE. For personal use only; all rights reserved.

formulation is limited in that it does not have a compression cap


and therefore excludes the possibility of compression failure at Netherlands.
the midspan. Lourenço, P. B., and Rots, J. G. 共1997兲. “Multisurface interface model for
analysis of masonry structures.” J. Eng. Mech., 123共7兲, 660–668.
Lourenço, P. B., Rots, J. G., and Blaauwendraad, J. 共1997兲. “Current
possibilities of masonry modelling.” Finite elements in engineering
Conclusions
and science, Hendriks, Jongedijk, J. G. Rots, and Spanje, eds.,
Balkema, Rotterdam, The Netherlands, 285–295.
A finite element formulation which models the brick unit and the Lourenço, P. B., Rots, J. G., and Van Der Pluijm, R. 共1999兲. “Understand-
mortar interface using triangular elements with nodes along the ing the tensile behaviour of masonry parallel to the bed joints: A
side interface of the triangular units has been presented. The numerical approach.” Masonry Int., 12共3兲, 96–103.
inelastic process is captured through constitutive laws for the nor- Maier, G. 共1970兲. “A matrix structural theory of piecewise-linear elasto-
mal and shear forces at the interface nodes. The present formula- plasticity with interacting inelastic failure planes.” Meccanica, 5, 54–
tion is based on a single branch softening law. The cracking and 66.
inelastic shear path is limited to the interfaces of the units. Mate- Nappi, A., and Tin-Loi, F. 共1999兲. “A discrete formulation for the numeri-
rial properties are required for both the mortar joint and the brick cal analysis of masonry structures.” Proc., APCOM ’99, 4th Asia-
unit. A Mohr–Coulomb failure surface with a tension cutoff has Pacific Conf. on Computational Mechanics—Computational Mechan-
been adopted. The present formulation uses a constant friction ics for the Next Millennium, Vol. 1, C. M. Wang, K. H. Lee, and K. K.
and dilatancy angle but future work will introduce a decaying Ang, eds., Elsevier Science Ltd., Amsterdam, The Netherlands,
dilatancy angle linked to the evolution of the failure surface. The 81–86.
Nappi, A., and Tin-Loi, F. 共2000兲. “A numerical model for masonry
path dependent softening behavior was solved using an finite in-
implemented in the framework of a discrete formulation.” Struct. Eng.
cremental nonholonomic form set in a linear complementarity
Mech., 11共2兲, 171–184.
format. The advantages of the present procedure is that inelastic Page, A. W. 共1978兲. “Finite element model for masonry.” J. Struct. Div.,
behavior involving the interaction of tension, sliding, and dila- 108共8兲, 1267–1285.
tancy in the mortar joints and brick units can be tracked. Page, A. W. 共1980兲. “A biaxial failure criterion for brick masonry in the
tension range.” Int. J. Masonry Constr., 1, 26–29.
Sutcliffe, D. J., Yu, H. S., and Page, A. W. 共1999兲. “Computational
Acknowledgments mechanics-limit analysis of unreinforced masonry shear walls.”
ACMSM 16, Sydney, Australia, 23–28.
The writers gratefully acknowledge the financial support of the Tin-Loi, F., and Tseng, P. 共2003兲. “Efficient computation of multiple so-
Australian Research Council Large Grant. lutions in quasibrittle fracture analysis.” Comput. Methods Appl.
Mech. Eng., 192, 1377–1388.
van der Pluijm, R. 共1992兲. “Material properties of masonry and its com-
ponents under tension and shear.” 6th Canadian Masonry Symp., Uni-
References versity of Saskatchewan, Sask., Canada, 675–686.
van der Pluijm, R. 共1993兲, “Shear behaviour of bed joints.” 6th North
Attard, M. M., and Tin-Loi, F. 共1999兲. “Fracture simulation using a dis- American Masonry Conf., 125–136.
crete triangular element.” ACMSM 16, Sydney, Australia, 11–16. van Zijl, G. P. A. G. 共1996兲. “Shear transfer across bed joints in masonry:
Attard, M. M., and Tin-Loi, F. 共2005兲. “Numerical simulation of qua- A numerical study.” TU-DELFT Rep. No. 03.21.0.22.28, Delft Univ.
sibrittle fracture in concrete.” Eng. Fract. Mech., 72共3兲, 387–411. of Technology, Delft, The Netherlands.
Bolzon, G., Maier, G., and Tin-Loi, F. 共1995兲. “Holonomic and nonholo- van Zijl, G. P. A. G. 共2001兲. “A discrete crack modelling strategy for
nomic simulations of quasi-brittle fracture: A comparative study of masonry structures.” Structural engineering, mechanics, and compu-
mathematical programming approaches.” Proc., Fracture Mechanics tation, 745–752.

1392 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / OCTOBER 2007

J. Struct. Eng. 2007.133:1385-1392.

Vous aimerez peut-être aussi