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

net/publication/273402080

Numerical prediction of the behavior, strength and elasticity of masonry in

compression

Article  in  Engineering Structures · May 2015

DOI: 10.1016/j.engstruct.2015.02.011

CITATIONS READS

57 1,036

3 authors:

Anastasios Drougkas Pere Roca

University of Leeds Universitat Politècnica de Catalunya
43 PUBLICATIONS   298 CITATIONS    167 PUBLICATIONS   3,134 CITATIONS   

SEE PROFILE SEE PROFILE

Climent Molins
Universitat Politècnica de Catalunya
133 PUBLICATIONS   2,081 CITATIONS   

SEE PROFILE

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

Non destructive testing View project

Offshore Code Comparison Collaboration Continuation, with Correlation (OC5) View project

All content following this page was uploaded by Anastasios Drougkas on 15 February 2018.

The user has requested enhancement of the downloaded file.

 Engineering Structures 90 (2015) 15–28

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Numerical prediction of the behavior, strength and elasticity of masonry

in compression
Anastasios Drougkas ⇑, Pere Roca, Climent Molins
Departament d’Enginyeria de la Construcció, Universitat Politècnica de Catalunya, Campus Diagonal Nord, Building C1, Jordi Girona 1-3 UPC, 08034 Barcelona, Spain

a r t i c l e i n f o a b s t r a c t

Article history: The paper focuses on the numerical prediction of the compressive response of masonry by means of
Received 18 December 2013 detailed micro-modeling techniques. In such a model, the material constituents (mortar and units) and
Revised 7 February 2015 the unit–mortar interfaces are separately described by means of specific constitutive equations.
Accepted 11 February 2015
The available modeling choices are evaluated through a literature review of related case studies. A sys-
tematic approach is proposed and is corroborated by the numerical simulation of a large number of
experimental cases from various sources. A total of fifty experimental results are simulated resulting
Keywords:
in an overall good prediction of the compressive strength and elastic modulus of the masonry composite,
Micro-modeling
Numerical modeling
as well as a realistic depiction of the failure mode.
Masonry This study focuses on the numerical prediction of the compressive strength of masonry, extending to
Compressive strength issues pertaining to global stiffness, failure mode, hardening and softening behavior of masonry and their
Young’s modulus numerical simulation. The masonry typology considered consists of solid units, primarily brick, and mor-
Plasticity tar of mostly lower strength and higher deformability.
Smeared cracking Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction subject to both vertical and horizontal actions, both of which acti-
vate compressive struts in masonry members. In meso-models and
Due to the large number of existing buildings composed of simplified micro-models the compressive strength of masonry is
masonry structural members such as load bearing walls, the given a value by the analyst a-priori.
numerical modeling and analysis of masonry structures are receiv- Several attempts of simulation of the compression behavior of
ing at present a growing amount of effort. Masonry buildings are masonry have been undertaken using three-dimensional models
found worldwide and encompass not only a large building stock, with varying purposes and results. This behavior, governed by
still in use, but also valuable architectural heritage. the interaction of the units and the mortar, may be strongly affect-
Given the importance of the seismic action and the potential ed by out-of-plane effects, as analytical models have indicated,
seismic vulnerability of masonry structures, a significant part of Hilsdorf [5]. One of the first numerical attempts involved a set of
this effort is being devoted to the numerical simulation of masonry elastic analyses on hollow concrete block masonry under concen-
failure governed by tension and shear. In particular, the shear fail- tric compression, Hamid and Chukwunenye [6]. A similar paramet-
ure of masonry walls has been modeled by means of simplified ric analysis was conducted to study the influence of elastic
micro-modeling using interface elements to model existing and properties and joint dimensions on stress distribution and masonry
arising planes of weakness, Lourenço and Rots [1]. Macro-models elasticity, Reddy et al. [7]. The differences between plane stress,
have also been employed for this purpose, taking into account generalized plane strain and three-dimensional modeling of
the orthotropic properties of masonry, Lourenço et al. [2], masonry in terms of strength, elasticity and stress distribution
Syrmakezis and Asteris [3] and Pelà et al. [4]. Instead, more limited have been discussed, Anthoine [8].
attention has been allocated to the detailed simulation of the Similar observations have been made using nonlinear models in
masonry response and failure in compression. However, an order to comment on the effects of model geometry on the predict-
accurate characterization of the compressive strength of masonry ed compressive strength of masonry, Berto et al. [9], Barbosa et al.
is needed in order to verify the capacity of masonry structures [10]. Finally, the general deficiencies of plane models in accurately
reproducing the behavior of single- and multi-leaf walls in
⇑ Corresponding author. numerical analyses have also been noted by Milani et al. [11,12].
E-mail addresses: anastasios.drougkas@upc.edu (A. Drougkas), pere.roca. According to the latter authors, 3D effects need to be taken into
fabregat@upc.edu (P. Roca), climent.molins@upc.edu (C. Molins). account for the correct derivation of the failure envelope of

http://dx.doi.org/10.1016/j.engstruct.2015.02.011
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
16 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

Notation

Eu Young’s modulus of units ec/3 strain corresponding to one third of the uniaxial
Em Young’s modulus of mortar compressive strength
Ec Young’s modulus of masonry ec strain corresponding to the uniaxial compressive
fc,u uniaxial compressive strength of units strength
fc,m uniaxial compressive strength of mortar eu ultimate strain for uniaxial compression
f c;exp experimentally derived compressive strength of mason- fc1 maximum principal stress
ry fc3 compressive failure stress
fc,num numerically derived compressive strength of masonry fcf failure compressive strength
ft,u tensile strength of units enst principal strain vector
ft,m tensile strength of mortar rc linear elastic stress vector
ft,i tensile strength of unit–mortar interface s linear elastic stress vector scaling factor
mu Poisson’s ratio of units Kr compressive strength scaling factor
mm Poisson’s ratio of mortar e0 initial strain
hu height of units brcr compressive strength reduction factor due to lateral
lu length of units cracking
tu width of units al,1 lateral damage variable 1
hm thickness of mortar bed joint al,2 lateral damage variable 2
lm thickness of mortar head joint alat average lateral damage variable
tm thickness of transversal mortar joint r residual strength factor
a internal damage variable vector fP peak confined compressive stress
GIf tensile fracture energy eP peak confined compressive strain
Gcf compressive fracture energy eU ultimate confined compressive strain
ecr crack strain fcr residual confined compressive strength
ecr,u ultimate crack strain tn interface normal traction
V volume of finite element kn interface normal stiffness
h characteristic element length Dn interface normal relative displacement
b shear retention factor I1 first stress invariant
J2 second deviatoric stress invariant

masonry under in-plane loads. Additionally, other case studies The modeling approach has been tested against existing
include the simulation of concentric, Furtmüller and Adam [13], experimental data, focusing exclusively on case studies where suf-
Schlegel [14], Vyas and Reddy [15], and eccentric compression of ficient material characterization has been carried out. In this
masonry, Adam et al. [16], Brencich et al. [17]. Finally, three- research, a total of fifty different experiments have been simulated.
dimensional periodic unit cells have been used for the verification Micro modeling requires knowledge of several elastic and strength
of the results of homogenization methods for masonry under tri- parameters, many of which are usually not measured in
axial normal and shear stress, Cecchi et al. [18], Zoucchini and experimental campaigns, the Young’s moduli being a case in point.
Lourenço [19]. The analyses carried out in this report and their comparison with
Application of modeling approaches proposed for the study of the experimental data may contribute to the enrichment of the
masonry in compression has been rather narrow in scope, normally inventory of numerical results on the analysis of masonry struc-
extending to only a very small number of experimental cases each. tures and provide a starting point for further investigation through
This fact also narrows the capacity for comparison between the parametric and sensitivity analyses.
results of experiments, empirical expressions and numerical Most of the case studies encountered in the literature and actu-
results concerning the compressive strength of differing types of ally adopted for the present research involve units stronger than
masonry. the mortars. This leads to compressive failure modes governed
The purpose of the modeling strategy presented herein is found by crushing of the mortar under multi-axial compression and
in the detailed simulation of the failure of masonry under compres- cracking of the units under combined compression and tension.
sive loading by means of a general approach combining versatility The proposed numerical approach can be applied to masonry
(the ability to analyze a variety of cases regarding geometry and with typical combinations of materials and structural arrange-
material constituents) and a moderate computational cost. The ments, but is especially suited for the analysis of traditional brick
strategy is based on detailed micro-modeling where specific masonries composed of lime mortars of low compressive strength.
constitutive equations are used separately for the material con- Modern brick masonries, built with cement mortar of a compres-
stituents (mortar and units) and the unit–mortar interfaces. sive strength close to or higher than that of the bricks may require
The applicability of the method focuses on masonry prisms and a different approach. The applicability of the method is further dis-
walls consisting of solid bricks and mortar arranged in stack bond, cussed in Section 5. The numerical analyses have been performed
running bond and Flemish bond walls and in English bond pillars. using the finite element program DIANA [20].
In principle, this approach can be applied to any type of masonry
bond, since geometrical peculiarities are taken into account explic-
itly in the model geometry. Adopting a detailed micro-modeling 2. Material models
approach for the simulation of failure in masonry means that, in
principle, geometry and morphology of the structure should not 2.1. General
be a limiting factor as long as the phenomena affecting strength,
elasticity and failure initiation and development are modeled The compressive strength of masonry is determined by, among
accurately. other factors, the properties of its constituent materials. According
 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28 17

to empirical expressions, such as the one found in [21], it is esti- A crack bandwidth equal to a characteristic dimension given by
pffiffiffiffi
mated by the compressive strength of both the units and the mor- h ¼ 3 V is adopted, where V is the volume of the finite element.
tar, normally assuming the former is higher than the latter. The Snap-back is avoided if the absolute initial slope of the softening
resulting strength of the composite lies between the two values, diagram for the given crack bandwidth is lower than the initial tan-
implying that the mortar fails under a stress level higher than its gent Young’s modulus. This holds for:
uniaxial compressive strength.
Furthermore, the failure mode commonly encountered in GIf E
masonry in compression, aside from the crushing of the joints, is h6 2
ð5Þ
vertical cracking of the units, caused by the lateral expansion of ft
the mortar in the joints. Failure at the unit–mortar interface in where E is the initial tangent Young’s modulus.
horizontal, vertical and transversal joints occurs by way of separa- Finally, cracked directions subjected to tension are not affected
tion under tension, especially in the vertical and transversal joints, by the Poisson effect, meaning that such loading does not lead to
and shear slipping. contraction in the perpendicular directions. Therefore, an
Therefore, detailed micro-modeling approaches for the simula- orthotropic formulation has been adopted for the Poisson’s ratios,
tion of masonry need to be able to model the nonlinear behavior of which are reduced at the same rate as the secant modulus after
the units and the mortar in tension leading to cracking and pres- cracking [22].
sure dependent behavior under multi-axial compression. Tensile Due to the rotating crack assumption, the crack orientation fol-
and shear failure at the unit mortar interface need to be modeled lows the direction of the principal stress. Therefore, retention of
as well. Thus the failure of the structure in both arising and macro- shear stiffness in the crack need not be modeled.
scopically existing planes of weakness needs to be accounted for.
For the present research, the numerical analyses have been per- 2.3. Compressive behavior
formed using a mixed pressure dependent plasticity model in com-
pression and a smeared cracking model in tension, organically Uniaxial unconfined compressive behavior is modeled using a
combined in a total strain non-linear model with secant unloading parabolic compression curve based on fracture energy [23]. The
[20,22]. Therefore, it is possible to simulate all failure modes nor- curve is defined by three characteristic strain values: the strain
mally expected to arise in masonry under compression. for which hardening is initiated at one third of the compressive
Damage due to tensile cracking is modeled using a rotating strength, the strain for which maximum stress is reached and the
crack model, in which stress–strain relationships are evaluated in strain for which softening is terminated. For a uniaxial compressive
the principal directions of the strain vector. The direction of the strength fc and an initial tangent modulus of E, the strain ec/3, at
cracks may therefore change according to the direction of the prin- which one third of the compressive strength has been reached, is
cipal strain. expressed as:
Shear behavior is explicitly governed by a relationship between
shear stress and shear strain, while six internal damage variables 1 fc
ak (assembled in the a vector), indicating the ratio of damaged to ec=3 ¼  ð6Þ
3 E
effective stress, monitor the deterioration of the material, which
The strain ec, at which the maximum compressive strength is
is non-recoverable.
This constitutive law has been used in numerical simulations reached, and is expressed as:
for concrete and masonry macro-models. It may be considered 5 fc
an attractive choice for the simulation of masonry since it is cap-
ec ¼  ¼ 5ec=3 ð7Þ
3 E
able of modeling its behavior in compression, tension and the
The ultimate strain eu, at which the material has terminated its
interaction of the failure modes.
softening in compression, which is expressed as:
c
2.2. Tensile behavior 3 Gf
eu ¼ ec  ð8Þ
2 hfc
The stress–strain relationship is elastic until the tensile strength
is reached. The expression of the tensile behavior in the post-peak where, in turn, Gcf is the compressive fracture energy and h is the
using values for the tensile strength ft and the tensile fracture ener- characteristic element length. In addition to the mesh insensitivity
gy GIf is accomplished through the use of an exponential softening of the model, strain localization and dissipation of the compressive
fracture energy is ensured by the geometry of the models them-
curve. For the post-peak of the tensile stress–strain relation, the
selves given that compressive damage is normally expected to
damaged stress is expressed as:
occur in the mortar joints. The parabolic compression curve is
ecr
rðecr Þ ¼ f t eecr;u ð1Þ defined for strain e by the piecewise equation:
8 e
where ecr is the crack strain and ecr,u the ultimate crack strain. To >
> f c 13 ec=3 if ec=3 < e 6 0
>
> 
calculate ecr,u the softening equation is rewritten as: >
>
>     
> 1 eec=3 eec=3 2
  >
>
> f c3 1 þ 4 ec ec=3  2 ec ec=3 if ec < e 6 ec=3
ecr <
rðecr Þ ¼ f t y ¼ f t yðxÞ ð2Þ f ¼
  2  ð9Þ
ecr;u > eec
> f c 1  eu ec
> if eu < e 6 ec
>
>
According to the definition of the tensile fracture energy: >
>
>0
> if e 6 eu
Z ecr ¼1 Z x¼1  >
>
:
GIf ¼h rðecr Þdecr ¼ f th yðxÞdx ecr;u ð3Þ
ecr ¼0 x¼0
Lateral pressure dependence, which accounts for increased
Therefore, the ultimate crack strain is: strength under confining stress, has been modeled using the four
GIf parameter Hsieh–Ting–Chen failure surface [22,24], which is
ecr;u ¼ ð4Þ defined as:
ft  h
18 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

pffiffiffiffi
J2 J f I1 εU εu εP εc εc/3
f ¼ C1 2 þ C2 2 þ C3 c1 þ C4  1 ¼ 0 ð10Þ
fc fc fc fc
fcr
where I1 and J2 are the stress invariants and fc1 is the maximum
principal stress. The numerical values of the material parameters
C1, C2, C3 and C4 are determined by uniaxial compression, uniaxial fc/3
tension, equal biaxial compression and triaxial compression tests on
the material in question. All the stresses in the criterion are
normalized by the uniaxial compressive strength, implying that
the behavior described by the model may be applied to materials
of different strength but similar behavior. The stress fc3, a negative
stress value which results in compressive failure, is determined by
scaling the linear elastic stress vector rc = sEenst such that Eq. (10) fc
holds, where s is the scaling factor sought, E the tangent Young’s
Unconfined
modulus and enst the principal strain vector. Thus, fc3 is defined as Confined
the minimum normal stress component of the stress vector. The fP
confined compressive strength is expressed as fcf = fc3 and its ratio
to the uniaxial compressive strength is expressed as fcf = Krfc. The Fig. 1. (a) Uniaxial compression curve and (b) compression curve under lateral
peak strain factor, relating peak to initial strain (eP = Kre0) is equal compression load.

to the ratio between confined and uniaxial compressive strength.

The peak stress is obtained by taking into account the influence
2.4. Interface behavior
of lateral cracking. This influence on the compressive behavior is
modeled using a reduction factor [25]. The reduction factor, denot-
Unit–mortar interface behavior is described using a discrete
ed as brcr ; is a function of the average lateral damage variable given
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cracking model based on a total deformation theory, in which
by alat ¼ a2l;1 þ a2l;2 ; where al,1 and al,2 are the two damage vari- interface tractions are expressed as a function of the total relative
ables, obtained from the a vector in the two lateral to the compres- displacements [20].
sive load directions, and is calculated by: Elastic behavior is assumed until the traction reaches the
assigned tensile strength and a brittle behavior is assumed
1 post-peak: normal stiffness is reduced to zero. Therefore, normal
br cr ¼ 61 ð11Þ
1 þ Kc tractions tn are expressed as:
where: 8
  < k n Dn
> if Dn 6 f t;i =kn
alat tn ¼ 0 if Dn > f t;i =kn ð20Þ
K c ¼ 0:27   0:37 ð12Þ >
:
e0
Finally, the peak stress is given by:
where kn is the normal interface stiffness, ft,i is the interface tensile
f P ¼ brcr  f cf ð13Þ strength and Dn is the interface relative displacement.
It is assumed that shear stiffness of the interface reduces to zero
The initial strain e0 is given by the equation:
after the onset of cracking, by adopting a shear retention factor of
n fc zero.
e0 ¼  ð14Þ
n1 E
with: 3. Modeling

fP
n ¼ 0:80 þ ð15Þ 3.1. Geometry
17
Increase in ductility due to lateral confinement is modeled A body of work involving detailed modeling of masonry under
according to the following stress–strain expression: in-plane loads, in which units and mortar are modeled separately,
  has been produced [8–10]. In these cases, the unit–mortar interac-
e  eP
f ¼ f P 1  ð1  rÞ 6 f cr ð16Þ tion results in out-off-plane stresses which may significantly alter
eU  eP the compressive behavior of the masonry composite. This is espe-
assuming a value for the ultimate strain: cially true in cases where there is significant discrepancy between
 3 the elastic characteristics of the two constituent materials, namely
fP the Young’s modulus and the Poisson’s ratio.
eU ¼ eP ð17Þ
fc It is appealing from the point of view of computational cost and
modeling effort to model wall structures, such as running or Flem-
The residual compressive strength fcr is expressed as:
ish bond masonry using simple plane geometrical models, such as
f cr ¼ f c r ð18Þ plane stress or plane strain. However, the results obtained demon-
strate some patterns of divergence from experimentally observed
Where: behavior and obtained compressive strength. In plane stress ana-
 3 lyses the units afford very low confinement to the mortar, while
f
r ¼ 0:1 P ð19Þ in plane strain analyses the confinement is excessive. The results
fc
of the under- and over-estimation of the effects of unit–mortar
A comparison of the basic parabolic curve for uniaxial compres- interaction in these cases are, expectedly, too low compressive
sive loading and its comparison to a curve under lateral compres- capacity in the former and too high compressive capacity in the
sion is presented in Fig. 1. latter. The failure modes are also characteristic of the above
 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28 19

(a) (b) (c) (d)

Fig. 2. Model geometries: (a) stack bond prism, (b) running bond wall, (c) Flemish bond wall and (d) English bond column.

modeling approaches, in plane stress more so than in plane strain,

with the former leading to excessive yielding in compression of the
weakest material and with the latter usually leading to excessive
cracking of the strongest and stiffest. The issue of directly modeling
the non-constant geometry of, for example, Flemish bond walls
and English bond columns raises further doubts concerning the
adequacy of plane methods. Fig. 2 illustrates various masonry
typologies of different degrees of geometrical complexity.
The regular assumptions accompanying plane analysis methods
are not sufficient due to the geometrical layout of the masonry
composite. The ratio of the thickness of the mortar joints, which
is of the order of 10 mm, over the thickness of the masonry, which
is of the order of 100 mm, invalidates the plane stress assumption
of zero thickness. This dimension ratio results in out-off-plane
effects, and stresses, to become locally significant near failure even
though the assumption is conceptually sound on a global, struc-
tural level. Similarly, the assumption of infinite masonry thickness
certainly does not hold globally, though it may be a legitimate sim-
plification for local effects given the usual ratio of dimensions
mentioned above. Generalized plane strain, although maintaining
some elements of phenomenological modeling, is usually a closer
approximation of the actual behavior and leads to better results
for the simulation of local failure [26,27].
Overall, the more homogenous the composite is, implying units
and mortar of similar properties, plane stress becomes more accu-
rate, while plane strain becomes more accurate for increasing
property disparities between the two material phases. The two
plane methods should generally provide accurate results for
extreme or particular cases, but their capacity to provide accurate
results for intermediate or general cases has not been
demonstrated.
Three-dimensional micro-modeling, while computationally the
most demanding approach, has been shown to produce the consis-
tently most accurate results in terms of capacity when applied to a
Fig. 3. Three-dimensional masonry model examples under vertical load with
large number of different experimental cases. No geometrical
planes of symmetry indicated. (a) Finite element mesh, (b) horizontal normal stress,
assumptions are made and, therefore, out-off-plane effects and (c) vertical normal stress, (d) in-plane shear stress.
stresses are taken into account organically, both on the structural
and on the local level, thus allowing the modeling of masonries along the thickness of each joint. Symmetric loading of symmetric
with a variety of different geometrical textures. Vertical and structures allows for three planes of symmetry to be applied, thus
horizontal compressive strength values obtained using this greatly reducing the size of the models.
method fall between those obtained from plane stress and plane The unit–mortar bond was modeled using a discrete cracking
strain. Overall, the comparison with available experimental data law applied on zero thickness 8 + 8 node plane quadrilateral
is favorable compared to that of plane methods. Interestingly, all interface elements, based on quadratic interpolation and a 4  4
the above mentioned modeling approaches produce equal initial Newton-Cotes integration scheme [20].
stiffness in compression for the masonry composite. Fig. 3 illus-
trates examples of three-dimensional models, including the overall 3.2. Solution method and loading
layout and the distribution of stress under vertical loading.
In the present work the models are composed of 20-node The analysis procedure for the solution of the physically non-
isoparametric solid elements based on quadratic interpolation linear problem was the Modified Newton-Raphson method. A Line
and 3  3  3 Gauss integration scheme [20]. The fineness of the Search algorithm was used to predict the iterative displacement
mesh was adjusted accordingly as to include at least two elements increment and to speed-up the convergence rate.
20 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

The loading was prescribed in displacement control. An energy mortar was calculated in this study using the values given for
based convergence criterion was used for the convergence check, the units and the masonry composite.
since displacement loading reduces the usefulness of the displace- In order to perform the analyses of the available case studies, it
ment norm criterion and the, mostly, unrestrained lateral expan- was necessary to assume values for the missing parameters. The
sion of the models reduces the buildup of internal forces, thus available literature overviewed in the present work offers an ade-
making the force norm less useful. For the convergence criterion quate amount of information on which to base these assumptions.
a 0.5% energy norm was adopted. The average value of the measured Poisson’s ratio in the set of
The finite element code DIANA, in which the above constitutive case studies is 0.13 for the units with values ranging from 0.07
laws and element types are implemented, was used for the ana- to 0.24 while for the mortars it is equal to 0.15 with values ranging
lyses, employing a parallel direct sparse solver [20]. from 0.07 to 0.2. In the cases where it was not reported, the Pois-
son’s ratio for the units was chosen as being equal to 0.15 while a
value between 0.15 and 0.25 was chosen depending on the type of
4. Inventory of experimental result data
mortar. In [32,38] a weak lime mortar was used, so a value of 0.25
was adopted. For [31,33,37,40,43], for which a cement/lime mortar
4.1. Overview
was used, a value of 0.20 was adopted. In [34,36] Portland cement
mortars were used and a value of 0.15 was used.
There is a relatively large inventory of existing experimental
The average ratio of tensile to compressive strength for the
data on masonry compression. However, the number of those
units was 9% with values ranging from 1.8% to 23.9%. For the mor-
qualifying for a numerical reproduction using micro-models is
tar the average was 8.2% with values ranging from 5% to 26%. For
fairly limited. The cause of this is the lack of sufficient material
all the mortars a ratio of 10% was assumed for the missing values.
characterization. Parametric analysis performed in this research
In the case of the units the ratio differed from case to case depend-
has indicated a clear, if rather strong, influence of the Young’s mod-
ing on the material used and the workmanship employed. For
ulus of both units and mortar on the compressive capacity and the
[33,37] a ratio of 10% was used since they involved wire cut solid
failure mode observed. Difficulties in measuring the Young’s mod-
clay bricks. The average value of all solid clay bricks in the inven-
uli of small size specimens, coupled with the reliance on empirical
tory had an average ratio of 9.3%. This value is similar to the 10%
expressions for the determination of the strength of the composite
usually assumed for masonry units. In [34] a ratio of 5% was
based solely on the compressive strength of the two constituents
assumed since the campaign was performed using hand molded
[21], have brought about a lack of motivation for taking adequate
bricks. This lower percentage was assumed in order to reflect the
measurements of it, especially in the case of mortars, despite
poorer quality and consistency of hand-made bricks compared to
observations on the effect of the deformation properties of the
machine molded bricks. In [7] soil–cement blocks were construct-
masonry constituent materials on the behavior of masonry com-
ed, therefore a 5% ratio was adopted. For [35], which involved gran-
posites [28–30]. Tensile strength measurements are also often
ite units, a ratio of 5% was chosen. According to [46] this is a ratio
neglected in unit characterization, even though unit cracking is
that fits the available experimental data on granites well.
commonly observed in compression of masonry.
The plasticity and cracking models require values for the com-
Certainly, there exists a strong correlation between the com-
pressive and mode-I fracture energies in order to describe the soft-
pressive strength and the Young’s modulus or the tensile strength,
ening behavior of the materials. The current status of the research
especially in modern types of units. However, the validity of a
on these values for clay bricks and, especially, lime mortars is quite
direct correlation between the two values is less clear for mason-
limited. Only a small number of the cases in the experimental
ries built with low strength mortars (such as traditional lime mor-
inventory include values for the compressive and tensile fracture
tars) where other factors, such as composition and ageing, may
energy of the units and/or the mortar. The values for the fracture
affect significantly on the resulting masonry stiffness.
energy of the units and the mortar were determined using Eqs.
In summary, the available candidate case studies for numerical
(21) and (22). The compressive fracture energy of both the units
simulation are few in number and mostly concerning stack bond
and the mortar was calculated assuming a ductility index para-
prisms [7,13–17,31–41]. Case studies of walls in running bond
meter of 1 mm, defined as
are not uncommon [33,40,42–44] but tests on walls in Flemish
bond are rare [45] while a few suitable examples of English bond d ¼ Gcf =f c ð21Þ
pillars were also identified [13,33]. The vast majority of these cases
involve the characterization of clay brick masonry with a few while the tensile fracture energy was calculated according to the
exceptions involving stone block masonry [14,35] or compressed following equation:
cement block masonry [15]. All mortars were either cement or 0:7
lime/cement mortars, with one exception where pure lime mortars GIf ¼ 0:025ð2f t Þ ð22Þ
were used [32].
which is based on the equation provided for the calculation of the
The average ratio of Young’s modulus to compressive strength
tensile fracture energy by CEB-FIP Model Code 1990 [47] adjusted
is 328 for the units, with values ranging from 14 to 1265. The aver-
to assume a ratio between tensile and compressive strength of 5%
age for the mortars was 699, with values ranging from 22 to 2094.
and a maximum aggregate size of 8 mm. The wide range of mortar
The average for the masonry was 356, with values ranging from 29
and unit materials used in masonry and the lack of information con-
to 1903. The highest figure corresponds to the only case with a
cerning the values for their fracture energy necessitates the adop-
ratio above 1000 while the remaining values, and their average,
tion of the above rather conservative values.
are much lower than 1000, which is the characteristic value recom-
Very few of the experimental cases provide any measurement
mended by design codes (e.g. [21]).
of the unit–mortar interface properties. Furthermore, only a small
number of works focused on these properties exists [48,49] which
4.2. Assumed values can hardly be used as a general guideline in a study of this extent.
Therefore, representative values had to be assumed. Throughout
The dimensions, compressive strength and Young’s modulus of the cases a tensile strength of 0.2 MPa was considered, while zero
the units and mortar were reported in all the examined cases, with Mode I fracture energy (brittle cracking) and zero shear retention
the exception of one case in which the Young’s modulus of the after the formation of the crack were assumed. Prior to failure
Table 1
Stack bond prism cases. Comparison of experimental and analysis results.

Case Ref. Eu [MPa] mu [–] fcu [MPa] ftu [MPa] hu [mm] lu [mm] tu [mm] Em [MPa] mm [–] fcm [MPa] ftm [MPa] hm [mm] lm [mm] tm [mm] fc,exp [MPa] fc,num [MPa] Ec,exp [MPa] Ec,num [MPa]
S1 [41] 9900 0.17 44.0 1.79 55 194 89 1750 0.16 6.20 0.62 7.5 – – 19.70 19.44 – 6456
S2 [32] 4200 0.16 23.0 3.10 45 270 135 125 {0.25} 1.25 0.19 10 – – 12.03 10.50 729 814
S3 [33] 976 {0.15} 5.7 {0.57} 75 230 105 1500 {0.20} 1.16 {0.12} 12 – – 1.67 2.21 467 1016
S4 [15] 14,500 0.18 11.5 0.71 80 260 120 6450 0.16 3.50 {0.35} 10 – – 6.15 6.83 – 13,017
S5 [33] 3370 {0.15} 23.00 {2.30} 75 230 105 8570 {0.20} 5.14 {0.51} 12 – – 6.70 7.40 2393 3525

A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

S6 [31] 135 {0.15} 9.3 0.46 52 212 104 795 {0.20} 9.37 {0.94} 10 – – 7.22 7.94 207 154
S7 [32] 4200 0.16 23.00 3.10 45 270 135 250 {0.25} 1.90 0.40 10 – – 13.73 14.80 1181 1287
S8 [14] 5500 0.11 12.00 0.90 113 240 175 2770 0.07 4.60 {0.46} 30 – – 4.68 5.44 4200 4702
S9 [33] 3370 {0.15} 23.00 {2.30} 75 230 105 5450 {0.20} 4.36 {0.44} 12 – – 7.40 8.17 3135 3628
S10 [33] 976 {0.15} 5.70 {0.57} 75 230 105 238 {0.20} 0.60 {0.06} 12 – – 1.25 2.17 379 735
S11 [15] 14,500 0.18 11.5 0.71 80 260 120 6450 0.16 3.50 {0.35} 7 – – 6.32 7.35 – 13,464
S12 [15] 14,500 0.18 11.5 0.71 80 260 120 6450 0.16 3.50 {0.35} 20 – – 5.01 6.21 – 10,400
S13 [16] 2000 0.10 13.8 3.30 55 250 110 1700 0.20 9.20 2.40 10 – – 14.55 13.26 1936 1950
S14 [13] 7500 0.07 30.0 1.30 65 40 40 220 0.20 1.70 0.10 20 – – 4.50 3.14 878 1475
S15 [37] 12,000 0.20 62.6 {6.26} 45 285 130 4200 {0.20} 6.20 {0.62} 10 – – 28.60 27.01 10000 9761
S16 [14] 5500 0.11 12.0 0.90 238 252 241 2770 0.07 4.60 {0.46} 12 – – 8.84 6.91 5517 5330
S17 [40] 16,700 0.15 66.0 1.20 52 210 100 2100 0.22 3.00 0.22 13 – – 11.73 13.77 6800 8294
S18 [41] 9900 0.17 44.0 1.79 55 194 89 8600 0.13 21.00 2.10 7.5 – – 34.70 36.75 – 9724
S19 [38] 5760 {0.15} 19.8 2.46 55 125 120 5490 {0.20} 2.62 0.35 10 – – 8.24 10.31 2132 5906
S20 [41] 15,000 0.13 58.9 2.74 57 200 98 11600 0.10 31.10 3.11 7.5 – – 48.20 50.59 – 14,593
S21 [33] 976 {0.15} 5.7 {0.57} 75 230 105 8570 {0.20} 5.14 {0.51} 12 – – 1.83 4.63 365 1098
S22 [41] 15,000 0.13 58.9 2.74 57 200 98 8600 0.13 21.00 2.10 7.5 – – 40.90 44.14 – 13,788
S23 [7] 8000 0.08 8.3 {0.42} 100 305 143 6600 0.19 3.45 {0.35} 30 – – 3.10 6.57 5900 7013
S24 [13] 7500 0.07 30.0 1.30 65 40 40 220 0.20 1.70 0.10 10 – – 9.33 5.74 1938 2380
S25 [41] 9900 0.17 44.0 1.79 55 194 89 6600 0.14 15.20 1.52 7.5 – – 27.00 31.32 – 9342
S26 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 7080 {0.20} 8.50 {0.85} 12 – – 10.00 14.38 3700 3585
S27 [41] 15,000 0.13 58.9 2.74 57 200 98 6600 0.14 15.20 1.52 7.5 – – 32.50 37.12 – 13,059
S28 [41] 9900 0.17 44.0 1.79 55 194 89 11600 0.10 31.10 3.11 7.5 – – 37.70 43.16 – 10,633
S29 [34] 5300 {0.15} 17.7 {0.89} 75 230 110 545 {0.15} 3.10 {0.31} 10 – – 4.00 9.74 2239 3034
S30 [36] 12,900 {0.15} 60.0 3.20 36 123 60 9590 {0.15} 12.00 0.97 7 – – 18.16 24.90 8000 12,601
S31 [41] 15,000 0.13 58.9 2.74 57 200 98 1750 0.16 6.20 0.62 7.5 – – 29.90 21.12 – 8063

21
22 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

the bond is considered perfect, therefore a large initial elastic stiff- been sorted and numerically named in ascending order according
ness was considered in the normal and perpendicular directions of to their statistical fit with the experimental compressive strength.
the interface, meaning that virtually all deformation in the inter- The material properties and dimensions for the models were
face is nonlinear. taken as reported in the experimental results and are shown in
The numerical parameters C1, C2, C3 and C4 were taken as being Tables 1–4. The values assumed as mentioned above for unknown
equal to 2.0108, 0.9714, 9.1412 and 0.2312 respectively. These val- quantities are displayed in brackets. Examples include several Pois-
ues correspond to a tensile strength equal to 10% of the compres- son’s ratios and values for the tensile strength. The experimentally
sive strength, a biaxial compression strength equal to 1.15 times achieved values, along with the numerical value for the compres-
the uniaxial and a compressive strength under biaxial pressure sive strength, are also shown in the same table.
equal to 80% of the compressive strength equal to 4.2 times the Computational effort remains substantial but not excessive: a
uniaxial strength. All four tests necessary for the complete deter- single small or medium sized wall may be analyzed in order to
mination of the four parameters are practically never available obtain the maximum load and part of the post-peak curve in two
for mortars used in masonry. This problem is compounded in the to three hours using a conventional PC. The attainment of the full
case of existing masonry structures, where material sampling for post-peak curve may cause computational time to double.
all four tests is very difficult. The problem is less crucial in the case
of masonry units, either clay or stone, since the behavior of mason- 5.2. Analysis of results
ry is not influenced by the pressure dependent behavior of the
units in compression. Since the failure mode in tension is governed With a few exceptions, the modeling strategy produced
by the smeared cracking model, the determination of the ratio of adequately accurate predictions of the compressive strength
tensile to compressive strength is not crucial. throughout the group of cases, which includes a wide range of
material combinations. Certain experimental results may be
regarded as dubious, such as S29 and S24, where the compressive
5. Results strength was too high and too low respectively, considering the
strength of the materials and the dimensions reported. Others,
5.1. Presentation of results such as F3, were executed using extremely strong and stiff mortar
and should be seen as outside the intended scope of this modeling
The cases available for numerical analysis include thirty-one approach.
stack bond cases (S), nine running bond masonry cases (R), three The accuracy of the method regarding the determination of the
Flemish bond masonry cases (F) and seven English bond pillar cas- Young’s modulus of masonry is comparably high. However, certain
es (P) for a total of fifty cases. The cases for each bond type have cases such as F1, F2 and F3 exhibited an experimental elastic

Table 2
Running bond wall cases. Comparison of experimental and analysis results.

Case Ref. Eu mu [–] fcu ftu hu lu tu Em mm [–] fcm ftm hm lm tm fc,exp fc,num Ec,exp Ec,num
[MPa] [MPa] [MPa] [mm] [mm] [mm] [MPa] [MPa] [MPa] [mm] [mm] [mm] [MPa] [MPa] [MPa] [MPa]
R1 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 5450 {0.20} 4.36 {0.44} 12 12 – 8.20 8.47 5232 3590
R2 [40] 4000 0.13 17.0 1.00 50 206 96 1650 {0.20} 6.90 {0.69} 12.5 10 – 13.60 14.21 3200 3095
R3 [33] 976 {0.15} 5.7 {0.57} 75 230 105 238 {0.20} 0.60 {0.06} 12 12 – 1.23 1.98 580 717
R4 [33] 976 {0.15} 5.7 {0.57} 75 230 105 1500 {0.20} 1.16 {0.12} 12 12 – 1.55 2.35 735 1033
R5 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 7080 {0.20} 8.50 {0.85} 12 12 – 12.60 14.40 4824 3702
R6 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 8570 {0.20} 5.14 {0.51} 12 12 – 9.60 7.24 5024 3782
R7 [43] 6740 0.17 36.5 {1.50} 35 110 50 970 {0.20} 3.20 {0.32} 5 5 – 8.60 11.30 3700 3949
R8 [33] 976 {0.15} 5.7 {0.57} 75 230 105 8570 {0.20} 5.14 {0.51} 12 12 – 1.18 4.46 400 1254
R9 [44] 22,000 0.15 61.0 10.70 30 105 49 8880 0.2 12.30 1.58 5 5 – 30.14 37.81 – 18,540

Table 3
Flemish bond wall cases. Comparison of experimental and analysis results.

Case Ref. Eu mu fcu ftu hu lu tu Em mm fcm ftm hm lm tm fc,exp fc,num Ec,exp Ec,num
[MPa] [–] [MPa] [MPa] [mm] [mm] [mm] [MPa] [–] [MPa] [MPa] [mm] [mm] [mm] [MPa] [MPa] [MPa] [MPa]
F1 [45] 4870 0.09 26.9 4.90 55 250 120 1180 0.06 3.20 0.90 10 10 10 11.00 4.29 1651 3107
F2 [45] 4870 0.09 26.9 4.90 55 250 120 5650 0.09 12.70 3.90 10 10 10 14.50 16.70 3833 5002
F3 [45] 4870 0.09 26.9 4.90 55 250 120 17800 0.12 95.00 15.70 10 10 10 17.80 29.58 4567 6390

Table 4
English bond pillar cases. Comparison of experimental and analysis results.

Case Ref. Eu mu [–] fcu ftu hu lu tu Em mm [–] fcm ftm hm lm tm fc,exp fc,num Ec,exp Ec,num
[MPa] [MPa] [MPa] [mm] [mm] [mm] [MPa] [MPa] [MPa] [mm] [mm] [mm] [MPa] [MPa] [MPa] [MPa]
P1 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 8570 {0.20} 5.14 {0.51} 12 20 20 6.70 6.65 3317 4005
P2 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 5450 {0.20} 4.36 {0.44} 12 20 20 8.70 8.13 3789 3684
P3 [33] 976 {0.15} 5.7 {0.57} 75 230 105 238 {0.20} 0.60 {0.06} 12 20 20 1.46 2.07 377 690
P4 [33] 3370 {0.15} 23.0 {2.30} 75 230 105 7080 {0.20} 8.50 {0.85} 12 20 20 13.60 14.24 3677 3865
P5 [33] 976 {0.15} 5.7 {0.57} 75 230 105 1500 {0.20} 1.16 {0.12} 12 20 20 1.44 2.28 381 1056
P6 [33] 976 {0.15} 5.7 {0.57} 75 230 105 8570 {0.20} 5.14 {0.51} 12 20 20 1.38 4.13 376 1510
P7 [13] 7500 0.07 30.0 1.3 65 290 150 220 0.2 1.70 0.10 10 10 10 5.55 9.12 661 2007
 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28 23

modulus much lower than what would be expected considering that of both constituent materials (mortar and units). This low
the elastic moduli provided for the constituents. compressive strength cannot be correctly predicted using this ana-
Fig. 4 illustrates the relation between numerical and lysis method as can also not be accounted for by usual models for
experimental values of the compressive strength, showing a good predicting the compressive strength of masonry. Finally, the Pois-
agreement between the experimental and the numerical results. son’s ratio reported for the mortar in the F1 case, a pozzolana-lime
Throughout all cases the coefficient of determination was 0.969 mortar with low strength, is very low compared to the average
for the numerical prediction. Similarly, Fig. 5 illustrates the relation derived from the experimental inventory. A higher value for this
between the numerical and experimental values of the Young’s parameter, which would increase the amount of confinement on
modulus of the masonry composite in vertical compression. the mortar layers, would result in an increase in the numerically
Throughout the cases the coefficient of determination was 0.892. predicted masonry compressive strength. Disregarding the above
The ratios of masonry Young’s modulus over the compressive mentioned cases increases the coefficient of determination for
strength are roughly equal to the ones produced by the experimen- the prediction of compressive strength to 0.976.
tal results. The dominant failure mode obtained was a combination of mor-
Concerning the comparison of the numerical and experimental tar yielding in compression and unit cracking for the majority of
results, certain comments should be made regarding the applica- the cases. Two major types of secondary damage were registered.
bility of the proposed method. As has been previously stated, the Firstly, some initial cracking of the mortar at the unit–mortar inter-
method is intended to be applied in cases of units with higher com- face near the outer faces of the masonry. Secondly, compressive
pressive strength and lower deformability than the mortar. All of yielding of the units in the case of masonries with mortar stronger
the examined cases comply with this intention, with the exception in compression than the units. The observed failure modes will be
of the F3 case, involving a mortar that is three times as strong in discussed in the following paragraphs, with some emphasis on
compression as the bricks, and the S6 case, in which the two com- their dependence on material properties.
ponents have equal strength. Great discrepancy between the elastic characteristics of the
Some remarks must be also made with regard to a few units and the mortar enhances the confinement afforded on the
experimental measurements that do not comply with the general mortar, resulting in a higher ratio between the masonry compres-
trend of the full inventory of experimental cases considered. These sive strength and the mortar compressive strength. Therefore, the
cases, labelled S6, S21, S23, R8, F3 and P6, are the only ones for relation between the compressive strength of the constituents
which the compressive strength of the composite is lower than and the compressive strength of the composite is partly dependent
on the Young’s moduli as well, especially in cases where the global
60 failure mode is governed by mortar yielding in compression.
Compressive yielding mainly takes place in the mortar joints.
50 Crushing failure of the mortar, represented numerically by plastic
strains, normally initiates near the face of the masonry, where
40 horizontal confinement is lowest, and develops toward the interior
fc,num [N/mm2]

of the joint. In the running bond and Flemish bond wall cases, this
compressive yielding failure exhibited a repeating pattern across
30
the face of the structure, both in the horizontal and the vertical
joints. Lateral expansion of the crushed mortar causes cracking
20 damage in the units at the interface near the free surface of the
masonry. The non-uniform development of compressive damage
10 along the depth of the horizontal joints indicates that plane meth-
ods, as already discussed, may be inadequate for an accurate
description of the phenomenon. The development of mortar yield-
0
0 10 20 30 40 50 60 ing is illustrated in Fig. 6.
fc,exp [N/mm2] Compressive yielding of the units to an extent that affects the
global failure mode only occurs in cases of very strong mortar.
Fig. 4. Experimental vs. numerical compressive strength. Coefficient of determina- For mortars with compressive strength comparable to or higher
tion R2 = 0.969. than that of the units, compressive yielding of the units may occur
in the pre-peak range and initiate overall failure.
Notable compressive yielding of the units may also take place
15000
after extensive cracking, covering nearly the entirety of the unit.
This only happens far in the post-peak and does not affect the
stress–strain curve near the peak.
Cracking damage mainly takes place in the units and the head
Ec,num [N/mm2]

10000
and transversal joints. The appearance of extensive cracking in
the units, represented numerically by crack strains, and its influ-
ence on the global failure, are not directly dependent upon the
ratio of the Young’s moduli. For high ratios, lateral tension on the
5000
units increases but the onset of extensive unit cracking may not
necessarily occur before the yielding of the mortar.
The amount of vertical confinement afforded on the mortar and
the amount of vertical splitting on the units are directly dependent
0
0 5000 10000 15000 on the mismatch of elastic properties of the constituent materials.
Ec,exp [N/mm2] The prevalence of one of the two failure modes over the other
depends on both the elastic and inelastic properties.
Fig. 5. Experimental vs. numerical Young’s modulus. Coefficient of determination Overall, the numerically obtained failure mode is chiefly gov-
R2 = 0.884. erned by mortar compressive yielding which develops during the
24 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

(a) (b) (c)

Fig. 6. Distribution of principal plastic strains in bed joint mortar for case S18 for increasing vertical stress levels: (a) yielding near the edge at the pre-peak range, (b) yielding
extended toward the interior of the joint at peak load, (c) yielding of entire joint in the post-peak range.

(a) (b) (c)

Fig. 7. Distribution of principal crack strains for case R9 for varying levels of vertical stress: (a) initial cracking in the head joints in the pre-peak range, (b) vertical cracks in
the units around the bed joints at the peak load, (c) vertical cracks in the units at the post-peak range.

hardening of the structure. The compressive strength of the units masonry. However, the tensile strength of the units plays a more
does not appear to play a direct role in the compressive strength substantial role, as demonstrated by the cracking development
of the composite for the majority of the cases. Unit compressive near and after the load peak. The Poisson’s ratio of the mortars
yielding is not involved in the initiation of failure; conversely, with low Young’s modulus is also a strong influencing factor as it
tensile unit strength is more directly involved, especially in the affects the lateral expansion of the mortar under vertical compres-
post-peak. Cracking of the units near the unit–mortar interface ini- sion and, therefore, the amount of horizontal confinement afforded
tiates in the outer surface of the masonry and remains superficial to it by the units. It can be thus concluded that the tensile strength
without extending toward the interior of the masonry. The more of the units and the Poisson’s ratio of the mortar are of importance
critical vertical cracks in the units may extend into the interior of in masonry in compression and their function is linked.
the masonry in the post-peak, a further indication that plane ana- In the majority of cases where a relatively high Poisson’s of 0.2
lysis methods are inadequate. The development of cracking in the was assumed the model exhibited a tendency to slightly overesti-
units is illustrated in Fig. 7. mate the compressive strength of masonry. This is especially
A comparison of experimentally and numerically derived apparent in the cases from [33,38,43]. A value of 0.15, which is clo-
stress–strain curves is presented in Fig. 8, covering a wide range ser to the average, would have resulted in a slightly better estima-
of results in terms of compressive strength and elastic stiffness. tion of the compressive strength of masonry.
The graphs show the good agreement obtained between the initial For comparison purposes, analyses of the running and Flemish
Young’s modulus of masonry as measured in the experiments and bond cases were performed without taking into account nonlin-
that predicted by the numerical model. The numerical post-peak earities in the interface elements (meso-modeling). The resulting
curve is usually steeper, as in graphs (c) and (f). This is an indica- compressive strength was identical to the one obtained from the
tion that the values assumed for the tensile fracture energy of detailed micro-models, despite the fact that the meso-models
the units and/or the compressive fracture energy of the mortars could not take into account tensile failure of the interfaces
is low. However, some of the experimental case studies were, between the units and the head and transverse mortar joints
indeed, characterized by such steep softening, a behavior which simulated by the micro-models.
was well approximated in the numerical model, as shown in Similarly, the influence of the tensile strength of the mortar was
graphs (a) and (f). investigated in the F1 Flemish bond case while at the same time
Concerning the missing material parameters, the absence of neglecting interface nonlinearities. The existence of both head
values for the tensile strength of the mortar is of very small conse- and transversal mortar joints potentially maximizes the influence
quence for the determination of the compressive strength of of this particular material parameter. Cracking in the mortar occurs
 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28 25

16 10

14 9
8
12
7
10 6

σ [N/mm2]
σ [N/mm2]
8 5

6 4
3
4
2
2 1
0 0
0 0.005 0.01 0 0.002 0.004 0.006 0.008
Strain Strain
(a) (b)
16
16
14
14
12 12
10 10

σ [N/mm2]
σ [N/mm2]

8 8

6 6

4 4

2 2

0 0
0 0.01 0.02 0.03 0.04 0 0.005 0.01 0.015
Strain
Strain
(c) (d)

35 9

30 8
7
25
6
σ [N/mm2]

20 5
σ [N/mm2]

15 4
3
10
2
5
1
0 0
0 0.005 0.01 0 0.02 0.04 0.06 0.08
Strain Strain
(e) (f)
Fig. 8. Experimental and numerical stress–strain diagram comparison. Numerical curves are shown in tinted lines: (a) S17 & R2 [40], (b) S8 & S16 [14], (c) S2 & S7 [32], (d)
S13 [16], (e) S15 [37], (f) S6 [31].

very early in the analysis, appearing in all vertical, transversal and of numerical analyses performed in this paper, two elements were
horizontal joints. In the former two cases the damage propagates used along the thickness of the joint. Two additional meshes were
through the entirety of the joint, while in the latter the damage tested: one with a single element and one with three elements
is limited to an area near the surface of the masonry, which is also along the thickness. The aspect ratio of the resulting finite ele-
the first to fail in compression. However, in terms of overall behav- ments are 1:0.93:0.92, 1:0.62:0.92 and 1:0.41:0.92 for the three
ior, the strength and the elasticity of the masonry composite are cases.
not significantly altered by the modification of the tensile strength The results are presented in Fig. 9. While the compressive
of the mortar: using a value of 10% and 1% of the experimentally strength and initial axial stiffness was unaltered, there is a distinct
derived value of the tensile strength of mortar caused a reduction influence of the fineness of the mesh in the post-peak. The model
of only 2.1% and 3.6% of the compressive strength of masonry. with a single element across the joint exhibits a more ductile
The sensitivity of the results to the fineness of the mesh was behavior. The results of the two and three element models exhibit
studied on the S1 experimental case [41]. Three meshes were similar post-peak behavior.
employed, the defining parameter for refinement being the num- Overall, mesh refinement only affected the post peak behavior
ber of elements along the thickness of the joint. In the main series but not the predicted capacity. The coarser mesh resulted in a more
26 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

25 Table 6
1 Element Experimental results vs. numerical and closed form expression results: Young’s
2 Elements modulus.
3 Elements
20 Case Ref. Ec,exp [MPa] Ec,1D [MPa] Ec,EC6 [MPa] Ec,num [MPa]
S1 [41] – 6351 12,221 6456
S4 [15] – 12,734 4014 13,017
15 S8 [14] 4200 4555 4500 4702
σ [N/mm2]

S11 [15] – 13,177 4014 13,464

S12 [15] – 11,604 4014 10,400
S13 [31] – 1947 6110 1950
10
S14 [13] 1936 854 6340 1475
S16 [14] 5517 5251 4500 5330
S17 [40] 6800 6986 13,055 8294
5 S18 [41] – 9724 17,622 9724
S20 [41] – 14,506 24,315 14,593
S22 [41] – 13,805 21,613 13,788
S24 [13] 1938 1386 6340 2380
0
0 0.002 0.004 0.006 S25 [41] – 9340 15,993 9342
S27 [41] – 13,066 19,616 13,059
Strain
S28 [41] – 10,077 19,825 10,633
S31 [41] – 7977 14,989 8063
Fig. 9. Effect of mesh refinement for S1 [41]: number of elements long the joint
R9 [44] – 18,166 18,866 18,540
thickness.
F1 [45] 1651 3286 7101 3107
F2 [45] 3833 4971 10,738 5002
F3 [45] 4567 6389 19,639 6390
ductile response. Finally, there was very little benefit in using three P7 [13] 661 1386 6340 2007
instead of two element along the thickness of the joint.

5.3. Comparison with closed form expressions The results for the compressive strength and the Young’s mod-
ulus are summarized in Tables 5 and 6 respectively. The Ohler and
The numerical results are compared with a number of closed Hilsdorf models are the most accurate, producing results of compa-
form and empirical expressions for the prediction of the compres- rable accuracy to the numerical model. They are followed by the
sive strength of masonry and its Young’s modulus. These analytical Khoo and Hendry, Francis and Hendry models. However, most
expressions are applied to those cases where the masonry materi- models produce a compressive strength higher than the compres-
als are completely characterized, with the occasional exception of sive strength of the unit for a number of cases, such as the Hendry
the value of the tensile strength of the mortar, which does not model for cases S13, R9, F1, F2 and F3, the Francis model for cases
influence the results for any of the analytical models or for the S18, S20, S28, F2 and F3 and the Khoo and Hendry and Hilsdorf
FEM analyses. models for case F3. The equations provided by the design codes
The analytical models proposed by Hilsdorf [5], Khoo and tend to underestimate the compressive strength, especially for
Hendry [50], Francis [51], Ohler [52] and Hendry [53] are used, masonry composites of higher strength, with the values provided
as well as recommendations by ACI [54] and CEN [21] standards by CEN being slightly closer to the experimental results. The FEM
for the characteristic strength of masonry. The elastic modulus model provides, with a few exceptions, results for the Young’s
of the composite as derived from a simple one-dimensional modulus very similar to the analytical model. The CEN code tends
homogenization method [55] and the CEN European standard are to greatly overestimate this parameter compared to the
similarly compared with the experimental and numerical results. experimental results.

Table 5
Experimental results vs. numerical and closed form expression results: compressive strength.

Case Ref. fc,exp [MPa] fc,Hilsdorf [MPa] fc,Francis [MPa] fc,Khoo & Hendry [MPa] fc,Ohler [MPa] fc,Hendry [MPa] fc,ACI [MPa] fc,CEN [MPa] fc,num [MPa]
S1 [41] 19.70 24.54 29.24 19.81 17.63 9.58 11.56 12.22 19.44
S4 [15] 6.15 8.03 9.61 7.16 6.44 3.75 5.05 4.01 6.83
S8 [14] 4.68 7.79 11.36 7.21 6.50 7.93 5.16 4.50 5.44
S11 [15] 6.32 8.56 10.10 7.67 6.94 3.80 5.05 4.01 7.35
S12 [15] 5.01 6.83 8.28 6.18 5.47 3.59 5.05 4.01 6.21
S13 [31] 14.55 11.89 12.44 12.04 11.78 27.94 5.52 6.11 13.26
S14 [13] 4.50 10.96 10.88 7.99 7.12 9.97 8.76 6.34 3.14
S16 [14] 8.84 9.96 11.87 9.46 8.84 8.13 5.16 4.50 6.91
S17 [40] 11.73 15.88 14.52 12.27 10.31 6.04 15.96 13.06 13.77
S18 [41] 34.70 30.59 47.22 29.05 28.46 10.66 11.56 17.62 36.75
S20 [41] 48.20 43.23 59.75 41.25 43.50 21.17 14.54 24.32 50.59
S22 [41] 40.90 39.48 49.90 35.75 35.50 20.03 14.54 21.61 44.14
S24 [13] 9.33 15.33 15.97 10.55 10.08 12.96 8.76 6.34 5.74
S25 [41] 27.00 28.22 39.86 25.75 24.37 10.33 11.56 15.99 31.32
S27 [41] 32.50 37.33 46.29 32.37 30.56 19.53 14.54 19.62 37.12
S28 [41] 37.70 34.72 71.27 34.98 40.74 11.34 11.56 19.83 43.16
S31 [41] 29.90 33.99 39.91 25.71 23.65 18.47 14.54 14.99 21.12
R9 [44] 30.14 47.12 52.33 40.76 39.08 62.29 14.96 18.87 37.81
F1 [45] 11.00 20.24 25.85 16.28 15.77 50.70 8.14 7.10 4.29
F2 [45] 14.50 21.93 27.34 20.99 20.06 55.80 8.14 10.74 16.70
F3 [45] 17.80 36.58 35.23 76.07 – 72.91 8.14 19.64 29.58
P7 [13] 5.55 15.33 15.97 10.55 10.08 12.96 8.76 6.34 9.12
 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28 27

6. Conclusions References

A systematic numerical simulation of masonry compression [1] Lourenço PB, Rots JG. Multisurface interface model for analysis of masonry
structures. J Eng Mech 1997;123:660.
tests has been performed on fifty cases with available experimen- [2] Lourenço PB, Rots JG, Blaauwendraad J. Continuum model for masonry:
tal results using a detailed micro-modeling technique. The main parameter estimation and validation. J Struct Eng 1998;124:642–52.
purpose of the simulation has been the validation of the proposed [3] Syrmakezis CA, Asteris PG. Masonry failure criterion under biaxial stress state.
J Mater Civ Eng 2001;13:58–64.
micro-modeling technique for the prediction of the compressive [4] Pelà L, Cervera M, Roca P. An orthotropic damage model for the analysis of
strength and elastic modulus of masonry based on the properties masonry structures. Constr Build Mater 2013;41:957–67.
of the constituent materials. [5] Hilsdorf HK. Investigation into the failure mechanism of brick masonry loaded
in axial compression. Designing, engineering and constructing with masonry
The modeling method, resulting from the application of a com- products. Houston (TX); 1969.
bined plasticity and smeared cracking material law in three dimen- [6] Hamid AA, Chukwunenye AO. Compression behavior of concrete masonry
sional masonry models has been shown capable of producing prisms. J Struct Eng 1986;112:605–13.
[7] Reddy BVV, Lal R, Nanjunda Rao KS. Influence of joint thickness and mortar-
reliable results in a wide range of combinations of masonry units
block elastic properties on the strength and stresses developed in soil–cement
and mortar. Its accuracy, however, is strongly dependent on suffi- block masonry. J Mater Civ Eng 2009;21:535–42.
cient mechanical characterization of the individual materials and [8] Anthoine A. Homogenization of periodic masonry: plane stress, generalized
plane strain, or 3D modelling? Commun Numer Methods Eng 1997;13:319–26.
should be tested against carefully executed tests on the resulting
[9] Berto L, Saetta A, Scotta R, Vitaliani R, Berto L, Saetta A, Scotta R, Vitaliani R.
masonry. Overall, no distinct tendency to over- or underestimate Failure mechanism of masonry prism loaded in axial compression:
the compressive strength and elastic modulus of the masonry computational aspects. Mater Struct 2005;38:249–56.
was observed. [10] Barbosa CS, Lourenço PB, Hanai JB. On the compressive strength prediction for
concrete masonry prisms. Mater Struct 2010;43:331–44.
The range of applicability of the proposed technique is far [11] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry walls,
greater than that of plane methods (such as plane strain or plane Part I: Failure surfaces. Comput Struct 2006;84:166–80.
stress applications), as unit–mortar interaction and its out-off-plane [12] Milani G, Lourenço PB, Tralli A. 3D homogenized limit analysis of masonry
buildings under horizontal loads. Eng Struct 2007;29:3134–48.
effects differ greatly for different ratios of elastic moduli between [13] Furtmüller T, Adam C. Numerical modeling of the in-plane behavior of
the materials. Three-dimensional micro-models allow a more historical brick masonry walls. Acta Mech 2011;221:65–77.
accurate and general simulation of the masonry compressive [14] Schlegel R. Numerische Berechnung von Mauerwerkstrukturen in homogenen
und diskreten Modellierungsstrategien. Bauhaus University, Weimar,
effects due to the more realistic prediction of damage patterns Germany. PhD dissertation; 2004.
and their development. [15] Vyas CVU, Reddy BVV. Prediction of solid block masonry prism compressive
In summary, the simulation of phenomena associated with strength using FE model. Mater Struct 2010;43:719–35.
[16] Adam JM, Brencich A, Hughes TG, Jefferson T. Micromodelling of eccentrically
masonry compressive effects using micro-models is much more
loaded brickwork: study of masonry wallettes. Eng Struct 2010;32:1244–51.
accurate and of wider applicability when using models of three- [17] Brencich A, Corradi C, Gambarotta L. Compressive strength of solid clay brick
dimensional geometry due to the more realistic prediction of masonry under eccentric loading. Proc. Br. Mason. Soc. 2002:37–46.
[18] Cecchi A, Milani G, Tralli A. Validation of analytical multiparameter
damage patterns and their development. Abiding to the principle
homogenization models for out-of-plane loaded masonry walls by means of
aims of structural micro modeling, which include the direct, rather the finite element method. J Eng Mech 2005;131:185–98.
than the phenomenological, simulation of structural behavior and [19] Zucchini A, Lourenço PB. A micro-mechanical model for the homogenisation of
failure, and in an effort to establish a generally applicable modeling masonry. Int J Solids Struct 2002;39:3233–55.
[20] TNO. DIANA user’s manual. Delft: TNO DIANA BV; 2012.
approach, three-dimensional micro-modeling appears to be the [21] CEN. EN 1996-1-1: rules for reinforced and unreinforced masonry; 2005.
most promising and advantageous method. [22] Selby RG, Vecchio FJ. Three-dimensional constitutive relations for reinforced
The numerical results have been compared with a number of concrete. University of Toronto, Department of Civil Engineering; 1993.
[23] Feenstra PH. Computational aspects of biaxial stress in plain and reinforced
closed form and empirical expressions for the estimation of the concrete. PhD dissertation, Delft University of Technology, Delft, The
masonry compressive strength and Young’s modulus. Compared Netherlands; 1993.
to the closed form and empirical expression, the numerical mod- [24] Hsieh SS, Ting EC, Chen WF. A plastic-fracture model for concrete. Int J Solids
Struct 1982;18:181–97.
el provides far more accurate estimations. In fact, some of the [25] Vecchio FJ, Collins MP. The modified compression-field theory for reinforced
expressions tested provide only fair estimations. A significant concrete elements subjected to shear. Am Concr Inst J 1986;83:219–31.
discrepancy has been found between the equation proposed [26] Lourenço PB, Pina-Henriques JL. Validation of analytical and continuum
numerical methods for estimating the compressive strength of masonry.
by CEN 2005 [21] for the estimation of the Young modulus
Comput Struct 2006;84:1977–89.
and the experimental or numerically predicted results. According [27] Massart TJ, Peerlings RHJ, Geers MGD, Gottcheiner S. Mesoscopic modeling of
to the comparison, the equation of CEN 2005 largely overestimates failure in brick masonry accounting for three-dimensional effects. Eng Fract
Mech 2005;72:1238–53.
the Young’s modulus.
[28] Pande G, Kralj B, Middleton J. Analysis of the compressive strength of masonry
A particular application of the proposed numerical approach given by the equation fk = K fba fmb. Struct Eng 1994;71:7–13.
can be found in the estimation of masonry average properties [29] Haseltine B. International Rules for masonry and their effect on the UK. Mason
based on the knowledge of specific parameters of the material con- Int 1987;1:41–3.
[30] Sinha BP, Hendry AW. Splitting failure of brickwork as a function of the
stituents. These properties may be later used as input data for deformation properties of bricks and mortar. Br Ceram Res Assoc Tech Note
macro-models utilized in the analysis of large structural parts or No. 86; 1966.
entire structures. [31] Akbarzade A, Tasnimi A. Nonlinear analysis and modeling of unreinforced
masonry shear walls based on plastic damage model. J Seismol Earthq Eng
2011;11:8–12.
Acknowledgments [32] Drougkas A, Roca P, Molins C. Strength and elasticity of pure lime mortar
masonry. Mater Struct 2015. http://dx.doi.org/10.1617/s11527-015-0553-2. in
This research has received the financial support from the Min- press.
[33] Gumaste KS, Nanjunda Rao KS, Reddy BVV, Jagadish KS. Strength and elasticity
isterio de Educación y Ciencia through the research project SUBTIS of brick masonry prisms and wallettes under compression. Mater Struct
(Study of the Sensitivity of Urban Buildings to Tunneling Induced Set- 2007;40:241–53.
tlements, ref. num. BIA2009-13233), the project MICROPAR (Identi- [34] Kaushik HB, Rai DC, Jain SK. Stress–strain characteristics of clay brick masonry
under uniaxial compression. J Mater Civ Eng 2007:728–39.
fication of mechanical and strength parameter of structural masonry [35] Mauro A. Long term effects of masonry walls. MSc dissertation, Roma Tre
by experimental methods and numerical micro-modeling, ref num. University, Rome, Italy; 2007.
BIA2012-32234) and the ERDF (European Regional Development [36] Hossain MM, Ali SS, Azadur Rahman M. Properties of masonry constituents. J
Civ Eng Inst Eng Bangladesh 1997;CE 28:135–55.
Fund).
28 A. Drougkas et al. / Engineering Structures 90 (2015) 15–28

[37] Oliveira DV de C, Lourenço PB, Roca P. Cyclic behaviour of stone and brick [46] Vasconcelos G de FM de, Lourenço PB, Alves CAS, Pamplona J. Ultrasonic
masonry under uniaxial compressive loading. Mater Struct 2006;39:247–57. evaluation of the physical and mechanical properties of granites. Ultrasonics
[38] Panizza M, Garbin E, Valluzzi MR, Modena C. Experimental investigation on 2008;48:453–66.
bond of FRP/SRP applied to masonry prisms. In: 6th International conference [47] CEB-FIP. Model code 1990. Thomas Telford; 1993.
FRP composites civil engineering 2012; 2012. p. 13–5. [48] Chaimoon K, Attard MM. Modeling of unreinforced masonry walls under shear
[39] Petersen RB, Masia MJ, Seracino R. In-plane shear behavior of masonry panels and compression. Eng Struct 2007;29:2056–68.
strengthened with NSM CFRP strips. II: Finite-element model. J Compos Constr [49] Almeida JC, Lourenço PB, Barros J. Characterization of brick and brick–mortar
2010;14:764–74. interface under uniaxial tension. In: Proceedings of 7th international seminar
[40] Vermeltfoort AT, Martens DRW, van Zijl GPAG. Brick–mortar interface effects on structural masonry, Brazil, CEFET-MG; 2002. p. 67–76.
on masonry under compression. Can J Civ Eng 2007;34:1475–85. [50] Khoo CL, Hendry AW. Strength tests on brick and mortar under complex
[41] McNary WS, Abrams DP. Mechanics of masonry in compression. J Struct Eng stresses for the development of a failure criterion for brickwork in
1985;111:857–70. compression. Proc Br Ceram Soc 1973;21:57–66.
[42] Mayorca P, Meguro K. Proposal for an efficient technique for retrofitting [51] Francis A, Horman C, Jerrems L. The effect of joint thickness and other factors
unreinforced masonry dwellings. In: 13th World conference on earthquake on the compressive strength of brickwork. In: Proceedings of the 2nd
engineering, Vancouver, BC, Canada, August 2004. international brick block masonry conference; 1971. p. 31–7.
[43] Page AW. Finite element model for masonry. J Struct Div ASCE [52] Ohler A. Zur berechnung der druckfestigkeit von mauerwerk unter
1978;104:1267–85. berücksichtigung der mehrachsigen spannungszustande in stein und mortel.
[44] Riddington JR, Naom NF. Finite element prediction of masonry compressive Bautechnik 1986;63:163–8.
strength. Comput Struct 1994;52:113–9. [53] Hendry AW. Structural masonry. Macmillan Education, Limited; 1990.
[45] Binda L, Fontana A, Frigerio G. Mechanical behaviour of brick masonries [54] ACI. ACI 530.1-11. Building code requirements and specification for masonry
derived from unit and mortar characteristics. In: 8th International brick block structures and related commentaries; 2011.
masonry conference, vol. 1, Dublin, Ireland; 1988. p. 205–16. [55] Como M. Statics of historic masonry constructions. Springer; 2012.

View publication stats