European Journal of Mechanics A/Solids 50 (2015) 39e58

Contents lists available at ScienceDirect

European Journal of Mechanics A/Solids

journal homepage: www.elsevier.com/locate/ejmsol

Continuous and discrete models for masonry like material: A critical

comparative study
Daniele Baraldi a, *, Antonella Cecchi a, Antonio Tralli b
a
b

Department of Architecture Construction Conservation, University IUAV of Venezia, Dorsoduro 2206, 30123 Venezia, Italy
Department of Engineering, University of Ferrara, Via G. Saragat 1, 44122 Ferrara, Italy

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 31 March 2014
Accepted 20 October 2014
Available online 31 October 2014

The aim of this paper is to present a critical comparative review of different models that may be adopted
for modelling the mechanical behaviour of masonry, with particular attention to microstructured
models.
Several continuous and discrete models are discussed. Such models are based on the following assumptions: i) the structure is composed of rigid blocks; ii) the mortar is modelled as an elastic material
or an elastic interface. The rigid block hypothesis is particularly suitable for historical masonry, in which
stone blocks may be assumed as rigid bodies. For this type of masonry, mortar thickness is negligible if
compared with block size, hence it can be modelled as an interface.
Masonry-like materials may be modelled taking into account their heterogeneity by adopting a heterogeneous Finite Element Model (FEM) or a Discrete Element Model (DEM). The former seems to be
more representative of masonry, but it is computationally onerous and results interpretation may be
difcult; the latter is limited to rigid block assumption and mortar joints modelled as interfaces. For this
reason, continuous equivalent models may be suitable to investigate masonry behaviour. Continuum
equivalent models provide, in an analytical form, constitutive functions, but Cauchy model may be not
suitable to describe masonry behaviour due to not negligible size of heterogeneity (block size) with
respect to masonry panel size. For this reason, micropolar equivalent continuum may be adopted.
By reference to the existing literature, a simple and effective DEM is adopted, in which masonry is
modelled as a skeleton having a behaviour depending on forces and moments transferred between
blocks through the interfaces (mortar joints). Moreover for the micropolar equivalent continuum, an ad
hoc enriched homogenised FEM is formulated by means of triangular elements. The proposed numerical
models represent two possible simple approaches for solving heterogeneous problems. Such models are
developed both by means of fast numerical routines and do not require specic computer codes, whereas
the heterogeneous FEM may be studied by adopting a traditional FE code.
DEM and heterogeneous FEM are adopted to verify reliability and application eld of Cauchy and
micropolar continua. Moreover, sensitivity of micropolar model to the Representative Elementary Volume (REV) chosen is discussed. For these purposes, ad hoc FE models are adopted, with constitutive
functions obtained from an identication procedure (both for Cauchy and micropolar continua). An
extensive comparison between DEM, heterogeneous FEM and equivalent homogenous FEM is presented
in some meaningful cases, taking into account also the effect of heterogeneity size on models behaviour.
2014 Elsevier Masson SAS. All rights reserved.

Keywords:
Masonry-like material
Discrete model
Micropolar continuum

1. Introduction
Masonry is a structural material obtained by composition of
blocks connected or not by mortar joints. Particularity of this

* Corresponding author.
E-mail addresses: danielebaraldi@iuav.it (D. Baraldi), cecchi@iuav.it (A. Cecchi),
tra@unife.it (A. Tralli).
http://dx.doi.org/10.1016/j.euromechsol.2014.10.007
0997-7538/ 2014 Elsevier Masson SAS. All rights reserved.

heterogeneous material is the heterogeneity size (size of block),

that may be not negligible with respect to global size of structural
element as in several composite materials. For this reason, in the
last twenty years, several researchers developed models for
studying masonry-like material adopting different approaches.
With this aim a heterogeneous FE model may be the more
appropriate procedure to investigate this material type. Stafford
Smith and Rahman (1972) were the rst to adopt a rough heterogeneous FE model for determining stresses in brickwork walls;

40

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

then, Page (Page, 1978; Ali and Page, 1988) adopted such type of
model for tting experimental results and taking into account the
non-linear behaviour of mortar joints. However the limit of this
approach lies in the difculty to analyse macro-scale problems. As
expected, the computational effort may be difcult to manage and
the interpretation of numerical results may be not easy.
Then, a discrete model (DEM), based on the assumptions of rigid
block behaviour and mortar joint modelled as interfaces, may be
suitable for investigating masonry behaviour due to the small
number of degrees of freedom (DOFs) involved in the analysis of
masonry panels. These assumptions seem to be appropriate for
historical masonry, in which stone block stiffness is very large if
compared with mortar stiffness, allowing to assume blocks as rigid
bodies, and mortar joint thickness is negligible if compared with
block size, allowing to model mortar joints as interfaces. However,
the assumption of rigid block imposes that boundary conditions
must be referred only to block centres. This aspect may not be
representative of actual mechanical behaviour. In the proposed
DEM, masonry is seen as a skeleton in which the interactions
between rigid blocks are represented by forces and moments that
depend on their relative displacements and rotations. Such model
was adopted in the past by many authors for studying masonry
behaviour in linear and non-linear elds (Masiani et al., 1995;
Formica et al., 2002; Casolo, 2004, 2006). In particular, Cecchi and
Sab (2004, 2009) dened a simple and effective DEM for studying
the three-dimensional behaviour of masonry panels and for
modelling random brickwork. Recently such model has been
extended to the viscoelastic eld by Baraldi and Cecchi (2014).
Discrete or distinct element models are widely adopted in other
scientic elds such as rock mechanics (see for example the pioneering works of Cundall and Strack, 1979; Cundall, 1988). Limits in
DEM approaches lie in the assumptions mentioned above, hence
during the last decades the original model has been modied for
taking into account the deformability of elements by introducing
additional parameters or by introducing FE discretisations (Itasca,
1989). Some examples of evolution of the DEM are represented by
commercial or open source codes (Itasca, 2000; Munjiza, 2004;
Mahabadi et al., 2012) that are characterised by a larger computational effort with respect to the original DEM. Recently, a comparison between such models and a simple DEM has been carried on for
studying masonry linear behaviour (Baraldi et al., 2013). Moreover,
an exhaustive description of discrete models and their improvement
up to recent years may be found in the work of Lemos (2007).
Although the DEM requires a small computational effort with
respect to the heterogeneous FE model at micro-scale level and
panel size level, it may be still unsuitable for studying masonry
behaviour at macro-scale level. For the above mentioned reasons,
continuous material equivalent to masonry were proposed. Among
continuous models, homogenisation-identication procedures
represent a consistent part of research. Indeed, homogenisation
procedures allow to take into account different mechanical assumptions for blocks and mortar. Standard Cauchy continuous
models are obtained applying periodic homogenisation techniques
and considering the elastic behaviour of both brick and mortar
(Anthoine, 1995; Cecchi and Sab, 2002; 2004). In the non-linear
eld some models exist in which blocks are assumed to be elastic
and mortar is modelled with a coupled damage-friction behaviour
(Milani et al., 2006; Sacco, 2009). In the non-linear eld, both block
and mortar may also display a non-linear behaviour (De Buhan and
De Felice, 1997; Gambarotta and Lagomarsino, 1997; Pegon and
Anthoine, 1997; Luciano and Sacco, 1998; Formica et al., 2002;
Massart et al., 2007; Wei and Hao, 2009).
On the other hand, micropolar or higher order continua have
also been adopted for masonry study. For micropolar continuum
see for example Masiani et al. (1995), Masiani and Trovalusci

(1996), Boutin (1996), Sulem and Mhlhaus (1997), Smyshlyaev

and Cherednichenko (2000), Forest et al. (2001), Casolo (2009),
Salerno and De Felice (2009), Addessi et al. (2010), De Bellis and
Addessi (2011) and Pau and Trovalusci (2012). For higher order
continuum see for example Stefanou et al. (2010), Bacigalupo and
Gambarotta (2012) and Trovalusci and Pau (2014).
A crucial problem with the choice of homogenisationidentication procedures is not only how kinematic, dynamic,
and constitutive prescriptions of a discrete system are transferred
to the continuous one, but also which continuum may be more
appropriate. Hence, constitutive functions of the continuous system may be different (Lofti and Benson Shing 1994, Loureno and
Rots, 1997; Del Piero, 2009).
The aim of this paper is to present a deep investigation of
different models that may be adopted for modelling the mechanical
behaviour of masonry, with particular attention to micro-structured
models. At microscopic level, blocks are assumed to be rigid and
mortar joints are modelled as elastic interfaces; a Cauchy standard
continuum and a micropolar model e based on two different REVs e
are considered. Then, an identication between the block structure
and a plane continuum model is carried out by equating the mechanical work in the two models for a class of regular motions. Due
to the hypotheses of the discrete model, the identication procedure
turns out to be simpler than a homogenisation procedure and leads
to the same results if blocks are assumed to be rigid. Hence, the
constitutive function of the two-dimensional (2D) model is obtained
from actual geometry and constitutive function of the discrete
model. Such compatible identication procedure is adopted for all
the models, in order to obtain equivalent continuous macroscopic
constitutive functions, that turn out to be orthotropic starting from
isotropic constitutive behaviour of block and mortar due to the
arrangement of masonry texture.
At panel size level this paper presents a comparison between
DEM and FEM in which constitutive continuum e Cauchy and
micropolar e functions are obtained from an identication procedure. Furthermore, a FE heterogeneous model is taken into account,
where constitutive functions of mortar and block are isotropic and
where Young modulus of block is 104 time larger than Young
modulus of mortar such as to simulate rigid block assumption. For
representing the micropolar continuum and performing examples
that can not be solved in analytical form, an enriched FE model is
adopted, with triangular elements that take into account rotations
as degrees of freedom. Recently in this eld, several enriched FE
models have been developed for studying the behaviour of generic
micropolar elastic materials (Zhang et al., 2005, 2012; Beveridge
et al., 2013). In particular, Providas and Kattis (2002) developed
an enriched triangular FE model and proposed several patch tests.
The paper is organised as follows: in Section 2, a description of
the 2D model is given and the mechanical power spent in its middle
plane is dened. In Section 3, in a dual manner, the discrete model
is described and the mechanical power, expanded to a generic
couple of blocks at the interface, is dened. In Section 4a correspondence between a class of regular motions is dened for two
portions of 2D and discrete models having the same size, and their
mechanical power is equated. In this way, the stress measure in the
plane is described as a function of the stress measure both for
standard Cauchy model and for micropolar model. A constitutive
linear isotropic elastic function for the mortar interface is adopted.
Consequently, the above mentioned compatible identication leads
to a constitutive orthotropic function. This procedure is applied to
the case of a masonry panel with a running bond pattern and in
Section 5 explicit formulas for this case are dened with reference
to two different REVs. It must be noted that this methodological
identication approach is embedded in linearised elasticity but
may be extended to the non-linear case.

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

In Section 6, the enriched triangular FE is described and Section

7 presents some examples of masonry panels subject to various
load and restraint conditions, together with a critical comparison
between the proposed models and taking into account the effects of
size of heterogeneity on models behaviour. In this section the
works of Salerno and De Felice (2009), Pau and Trovalusci (2012)
and Trovalusci and Pau (2014) are chosen as benchmark. In these
papers the problem of Cauchy and micropolar equivalent continua
is dealt with. The DEM proposed by Salerno and De Felice (2009) is
a 2D FEM with periodic brickwork modelled as a Lagrangian system
of rigid bodies interacting by point elastic interfaces (Salerno et al.,
2001; Formica et al., 2002), whereas Cauchy and micropolar continua are modelled by rectangular high continuity FEs. Pau and
Trovalusci (2012) and Trovalusci and Pau (2014) adopt commercial FE codes both for the continua and the DEM. In the latter case
the code represents each block by a rectangular plane element with
four nodes, constrained so that each element behaves as a rigid
body, with mortar layers modelled as linear elastic springs. Differently than these works, this paper describes an ad hoc DEM that
requires a small computational effort with respect to other similar
models due to the small number of DOFs (block displacements)
involved in the analysis. Similarly, the enriched FEM used for the
comparison is simply obtained by adding an additional nodal DOF
to traditional constant strain triangular (CST) FEs.

41

points belonging to S. In the plane case, following the notation of

Cecchi and Rizzi (2005), the static counterpart is fully described by
the eld N, collecting the in-plane actions, and by the eld M,
representing the microcouple:

N : S/V;

M : S/Skw V

(3a,b)

The static model envisages spatial elds of forces and couples

(N, M) and eld of body forces and couples (b, B). The balance
equations for the in-plane case are:

divN b 0;

divM  2SkwN B 0

(4a,b)

where div is divergence operator dened on S. For the continuum,

set N and M actions, the mechanical work on S may be written as:

P N$grad u U Me3 $grad U;

(5)

where grad represents the gradient operator on S. If the adopted

continuum follows Cauchy's hypotheses, the in-plane couple is
assumed equal to zero (Sulem and Mhlhaus, 1997; Stefanou et al.,
2008) and the total internal work can be thus evaluated as the
power expended by membrane actions:



PCauchy N$sym grad u :

(6)

2. Plane continuum model

A plane model is dened, hence reference is made to a 2D
continuum identied by its S middle plane in a Euclidean coordinate system (y1, y2) of normal vector e3 along the y3 coordinate
direction (Fig. 1).
The kinematic descriptors of a generic point belonging to the 2D
continuum are represented by the following elds: u(y), U(y), that
are respectively the translation vector and rotation tensor of a
generic point y. u(y) is a vector with two components, namely
u1(y1,y2) and u2(y1,y2), whereas the rotation is a skew tensor U(y)
with one component, dened as follows:

U

0
u3

u3
0


(1)

From the above considerations, the generic displacement is

described by the elds:

u : S/V;

U : S/Skw V;

(2a,b)

which completely describe the V space of displacements-translation and skew (Skw) or antisymmetric part i.e. rotation-of all

3. Discrete model
A standard running bond periodic masonry is considered and
Fig. 2a shows a Representative Elementary Volume (REV) having a
block Bi,j surrounded by six blocks. Block plane dimensions are: a
(height) and b (width), whereas s represents block and panel thickness. Assuming rigid block hypothesis, the displacement of each
block Bi,j is a rigid body motion referred to the motion of its centre
and it is dened by the following expression (Cecchi and Sab, 2004):

 


ui;j y ui;j Ui;j y  yi;j ;

(7)

i;j T
i,j
where ui;j fui;j
1 u2 g and U are the translation vector and rotai; j
tion skew tensor having one component u3 of Bi,j, respectively, and
i,j
y is the position of its centre in the Euclidean space.
Following the procedure described by Cecchi and Sab (2009), a
generic couple of blocks Bi,j and Bik1 ; jk2 (Fig 2b) is considered.
Considering p as the centre of the Sk1 ; k2 interface between such
blocks, the displacement of the material points y of Bi,j and
Bik1 ; jk2 in contact in a generic position x2Sk1 ; k2 , may be written
as follows.

 
 


ui;j x ui;j p Ui;j x  p
 
 


ui k1 j k2 x ui k1 j k2 p Ui k1 j k2 x  p

(8a,b)

The measure of deformation may be written as a function of the

jump of displacement eld d(x) between Bi,j and Bik1 ; jk2 in a point
x2Sk1 ; k2 :

 
 
 
d x uik1 ;jk2 x  ui;j x
 
 




uik1 ;jk2 p  ui;j p Uik1 ;jk2 x  p  Ui;j x  p


up Up x  p
(9)
Fig. 1. 2D Continuum model.

42

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 2. Discrete model: running bond REV (a), generic couple of blocks, (b).

where up uik1 ;jk2 p  ui;j p and Up Uik1 ;jk2  Ui;j . The

kinematic adopted in this discrete model will be adopted both for
the DEM (Cecchi and Sab, 2004) and for the following compatible
identication with a micropolar continuum.
Let be ti;j x; tik1 ;jk2 x the forces respectively acting on Bi,j and
Bik1 ; jk2 blocks, the balance equation provides ti;j x
tik1 ;jk2 x. Hence, the work of the contact actions at the interface
becomes:

Z
P

tx$dxdA
S

Z h

 
 
 
 i
ti;j x $ui;j y tik1 ;jk2 x $uik1 ;jk2 x dA;

where dA is an innitesimal portion of the interface Sk1 ; k2 .

Assuming tik1 ;jk2 x tx, the work of contact actions become:



 
 
 
t x uik1 ;jk2 x  ui;j x dA

tp $up Up

Skwtx5x  pdA;

(11)

where the symbol 5 represents the tensor product. Hence,

Pp tp $up
where tp

R
S

1
Mp $UP ;
2

txdA; Mp 2

(12)
R

Skwtx5x  pdA.

By assuming block as rigid bodies and mortar joints modelled as

linear interfaces, the constitutive function that denes interaction
between block Bi,j and Bik1 ; jk2 is

tx Kdx:

A portion P of a masonry panel with the same dimensions of the

REV is considered (Fig. 3a). This portion is chosen so that its centre
yi,j coincides with the centre of the REV (Fig. 2a). A portion of panel
H, having the same edge is considered, so that the y point of H
coincides with yi,j.
From the above considerations, it is possible to assign a correspondence between a class of regular displacement in P and H. In
particular, it is assumed that the translation and rotation of the
centre of the block Bi,j in the discrete system are equal to the
translation and rotation of the centre of the REV in the continuum
model: ui,j(y) u(y), Ui,j(y) U(y); hence

(10)

4. Compatible identication

 
 
 

ui k1 ;j k2 y u y grad u y yi k1 ;j k2  y ;
 
 
 

Ui k1 ;j k2 y U y grad U y yi k1 ;j k2  y :

where yi,j and yik1 ; jk2 are the centres of Bi,j and Bik1 ; jk2 2 P
generic couple of blocks and a rst order Taylor approximation
(rst order identication) in translation and rotation is used. According to the kinematic description adopted, the vector tp {t1p
t2p}T denotes in-plane tractions (projection on S). Taking in
consideration correspondent displacement tests, from Eqs. (14) and
(12) may be split into two parts for the sake of clearness (Salerno
and De Felice, 2009) and re-written as follows:

 


tp $up tp 5 yik1 jk2  yi;j $ grad u U
n
 
h

grad U tp 5 p  yik1 jk2 5 yik1 jk2  y
 
io

 p  yi;j 5 yi;j  y
(15a)

(13)

If e is the vector orthogonal to plane of interfaces, the constitutive function becomes K 1=emI m le 5e , where e is
s conthe actual thickness of mortar joint and m, l are the Lame
stants of mortar. The proposed formulation for the discrete model is
coincident to that proposed by Salerno and De Felice (2009).
Furthermore, the proposed formulation allows to dene the work
of contact actions at interfaces (Eq. (10)) in terms of block degrees
of freedom ui,j and Ui,j. Then, forces and couples between blocks are
obtained by differentiating the work at the interfaces and the
equilibrium problem of the panel subject to in-plane actions is
solved numerically by adopting a molecular dynamics method.
More details for the DEM description and its solution may be found
in Cecchi and Sab (2004, 2009).

(14)

1
Mp $Up
2

Z 
S




tp 5vp  vp 5tp $grad U yi k1 j k2  yi;j dA

Z 

 

v1p t2p  v2p t1p 5 yi k1 j k2  yi;j grad UdA

(15b)
where the distance vector vp {v1p v2p}T can be dened as
vp x  p (Masiani et al., 1995). At this stage, for a chosen REV and a
given class of regular displacements, it is imposed that the mechanical work spent by the contact actions on P and H coincides.
Hence, forces and couples of the 2D continuum may be expressed
in terms of the forces acting in the discrete model:

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

43

Fig. 3. Identication between REV and 2D continuum.



1 X
tp 5 yik1 jk2  yi;j
2A n

Z


 
1 X
Me3
v1p t2p  v2p t1p 5 yik1 jk2  yi;j dA
2A n
S
 
h

tp 5 p  yik1 jk2 5 yik1 jk2  y
 
i

 p  yi;j 5 yi;j  y

(16a)


 nh
i
KC E; grad U U; U ; ui;j E$yi;j vi;j ; Ui;j
o
h
i
ui;j ; v; u 2L2

(16b)
where A is the area of the chosen REV and the symbol Sn indicates a
summation extended to the n interfaces of the chosen REV. The 1/2
coefcient appearing in the above expressions for N and M is
relative only to the external interfaces of the REV, because such
interfaces are shared by contiguous REVs. Under these assumptions, if the Cauchy continuum is chosen, only the symmetric part of
P is considered, hence the membrane tensors N and M may be
expressed as a function of the vector tp, i.e. as a function of the
measure of the stress in the micro-mechanical model. It must be
noted that the part of P associated to Skw(grad U) is not taken into
account in the Cauchy continuum. In fact, in the adopted 2D model
such kinematic elds characterise neutral (rigid) motions. Then,
applying Eq. (13), that becomes tp K up, into Eq. 16a,b it is possible
to obtain constitutive equivalent functions.
5. Model for rigid blocks connected by elastic interfaces:
running bond masonry
The compatible identication procedure is based on the same
geometry of discrete system shown in Fig. 2, where the texture
pattern may represent a running bond brickwork; moreover, the
block Bi,j together with the six surrounding blocks, form a Representative Elementary Volume (REV). The displacement of the block
Bi,j is the rigid body displacement dened by Eq. (7). The constitutive
law for any interface between adjoining blocks, Sk1 ; k2 , is supposed to
be a linear elastic relationship between the tractions t over the block
surfaces and the jump in displacement d across Sk1 ; k2 (Eq. (13)).
Upper bounds for the strain energy of the equivalent medium
may be obtained using a suitable kinematic eld over the REV.
Assuming E as the macroscopic in-plane strain tensor in the
equivalent medium, the continuum equivalent in-plane tensor A
and micro-couple tensor L are obtained by solving the following
minimisation problem (Cecchi and Sab, 2009):

1
1
$A$E grad U$L$grad U minU;U2KCE;grad
2
2

where z is the strain energy averaged over the REV, U and U

represent any strain-periodic rigid body displacement of the blocks,
kinematically compatible with E. The set KC of E-kinematically
compatible (U, U) is introduced:

(18)

where [v, u] is the in-plane rigid displacement of block Bi,j e

respectively translation and rotation. Hence, uniform boundary
displacement and rotation are applied to REV.
The approximate equivalent elastic coefcients, denoted by Aijkl
and Lij are such that:

1
1
1
1
A
E 2 A2222 E22 2 A1212 E12 2 A2121 E12 2
2 1111 11
2
2
2
L11 u23;1 L22 u23;2  z;
(19)
where E11 u1,1, E22 u2,2, E12 u1,2 u3 and E21 u2,1  u3.
Following the procedure proposed above, it is possible to dene
the equivalent micropolar continuum. As well known, for the
Cauchy continuum, Eq. (19) becomes:

 2
1
1
Cauchy
C
A1111 E11 2 A2222 E22 2 2A1212 E12
z
2
2

(20)

C EC 1=2u
where E12
1;2 u2;1 .
21
Two different REVs may be considered. REV1 (Fig. 4a) is characterised by a block at the centre of the Cartesian coordinate system, with four horizontal interfaces and two vertical interfaces far
from the centre of the cell, whereas REV2 (Fig. 4b) is characterised
by four horizontal interfaces and a vertical interface at the centre of
the Cartesian coordinate system.
The constitutive functions of the components of N and C Me3
on S plane are the same obtained by Salerno and De Felice (2009)
adopting a 2D discrete model and following the same compatible
identication procedure.

N A$grad u U; C L$grad U

(21a,b)

The components of tensor A are the same for both REVs

considered are given by the following expressions (see Appendix
A1 for more details):

 .  
 
h
i.h  . 
i
A1111 4Kv eh a b=a Gh ev =a
4 eh a ev =b ;

U z

(17)

(22a)

44

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

A2222 Kh

. . 
eh a ;

(22b)

A1122 0;
A1212

(22c)
6. Finite element formulation

 . 
Gh a eh ;

A2121 Kh b2

(22d)

.



4aeh Gv b=ev :

(22e)

where a and b represent, respectively, height and width of the

block. e represents the thickness of mortar joints and K and G are
the bulk and shear moduli of mortar. h and v superscripts and
subscripts are referred to horizontal and vertical interfaces,
respectively. In the following, for simplicity, horizontal and vertical
mortar joints are assumed to have the same thickness (ev eh e)
and the same elastic properties (Kv Kh K, Gv Gh G).
The equivalent moduli for the Cauchy continuum do not depend
on the REV chosen and they are given by Eq. (22aec), moreover it
can be demonstrated that
Cauchy

A1212

Stefanou et al. (2008, 2010) that compared the dispersion functions

of the discrete model with respect to those of the equivalent
continuum.

A1212 $A2121
:
A1212 A2121

(23)

The components of diagonal tensor L are strictly dependent on

the REV. Expressions 24a,b are related to REV1, whereas Eq. 25a,b
are related to REV2:

h
.
.

 .
. i
aeh 12Gh a eh
L11 b2 192 16Kv a2 bev 4Kh b2
(24a)
h

.
i
.
L22 b2 Kh a 12eh Gh a3
4b2 eh ;

(24b)

h
.
.

 .
. i
aeh 12Gh a eh ;
L11 b2 192 16Kv a2 bev Kh b2
(25a)
h
i
.
L22 b2 Kh a 48eh :

(25b)

It is worth noting that REV1 is characterised by a larger mechanical work than REV2 (Eq. (16)). In fact, REV1 has two vertical
interfaces far from the centre of the cell, whereas REV2 has one
vertical interface at the centre of the cell. Hence, coefcients Lij of
REV1 are larger than those of REV2. According to Salerno and De
Felice (2009), both REVs are centre-symmetric. However, REV1 is
centred at the centre of the block and REV2 is centred at the centre
of the vertical joint. The stiffness coefcients of Eq. 24a,b are larger
than those of Eq. 25a,b due to the tangential contributions of vertical joints in REV1, that do not appear in REV2, in which the centre
of vertical joint coincides with REV centre. The better performance
of REV2 with respect to REV1 has been also demonstrated by

An enriched Constant Strain Triangular (CST) nite element is

taken into account for performing analysis of micropolar models.
j
j
j
Nodal degrees of freedom are u1 , u2 , u3 with j 1, 2, 3 (Fig. 5), that
correspond to the independent elds u and u3, that are discretised
over the triangular area by the linear polynomials commonly
adopted in standard triangular FEs (Zienkiewicz and Taylor, 1989):

u/u1

X j
u1 N j ;

u2

X j
u2 Nj ;
j

u3

X j
u3 N j ;

(26)

where Nj a0 a1 y1 a2 y2 is the common bilinear shape function for a triangular element. The displacement and rotational degrees of freedom of the element may be collected in a vector qel
having nine components. Then, the stiffness matrix Kel of the
triangular element is determined as usual starting from the potential energy Pel of the triangular element and introducing the
expressions for stresses and strains of the micropolar continuum. In
a generic micropolar elastic continuum, force and couple stresses
within the element are related to deformations by six elastic constants (Eringen, 1966). In this case, for modelling a running bond
masonry panel, the six constants are determined towards the
identication procedure described in section 5 and consequently,
two different stiffness matrices are obtained, based, respectively, on
REV1 and REV2. The CST element is adopted in order to follow an
existent literature dedicated to the analysis of periodic structures
modelled as micropolar continua (Masiani et al., 1995; Masiani and
Trovalusci, 1996; Providas and Kattis, 2002; Trovalusci and Masiani,
2003; Wheel, 2008). Moreover, the triangular element allows to
perform an exact integration for the determination of the stiffness
matrix; further details may be found in Appendix A2.2. However, it
is clear that elements better than the CST are available (Cook et al.,
2001) without increasing the complexity of the problem. In the
following numerical tests, a symmetric discretization made of
triangular FEs is adopted for reducing the directional stiffness bias
(Logan, 2012). Elements superior than the CST, such as the quadrilateral ones, may be adopted in further developments of the
present work.

7. Numerical tests
A numerical experimentation is performed in order to evaluate
a) the opportunity of adopting a micropolar continuum instead of a
traditional one, b) the effectiveness of the proposed triangular FE
for determining the behaviour of masonry-like panels, with
particular attention to the rotations u3 of the discrete system and of
the micropolar continuum, c) the more appropriate REV for

Fig. 4. Representative Elementary Volumes (REVs) considered for the identication procedure.

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 5. Constant Strain Triangle (CST) with three degrees of freedom per node.

representing running bond masonry behaviour by means of the

micropolar continuum.
In order to evaluate also the effect of size of heterogeneity on the
behaviour of discrete and micropolar continuum models, different
scale factors r L/b are adopted, taking into account panel size L
with respect to block size b.
The numerical experimentation is performed on several models,
that are resumed in the following list:
i)
ii)
iii)
iv)
v)

Discrete Element Model (DEM);

Heterogeneous FE model (FEM Hetero);
Micro-polar FE model based on REV 1 (FEM REV1);
Micro-polar FE model based on REV 2 (FEM REV2);
Cauchy orthotropic FE model (FEM Cauchy)

The DEM (i) has been briey described in section 3; differently

than experimentations performed by other authors (Salerno and De
Felice, 2009; Pau and Trovalusci, 2012; Trovalusci and Pau, 2014),
the solution for this model is simply obtained by using a standard
computer code for matrix analysis. It is worth noting that this
model allows to study masonry behaviour with a smaller computational effort with respect to the other models considered, due to
the small number of DOFs taken into account.
A heterogeneous FE model (ii) is adopted in order to verify the
effectiveness of the DEM. This model is briey described in
Appendix A2.1, together with a convergence test performed for
determining the more appropriate mesh renement, in order to
obtain accurate results and, contemporarily, avoid a large number
of DOFs involved in the analysis. Such convergence test shows that
differences between displacements are not evident if a rough mesh
(Fig. 6a) is adopted instead of a ne mesh (Fig. 6b). Moreover, a ne

45

mesh for a panel with a large scale factor r may require a huge
computational effort, hence in the following, the ne mesh is
adopted for determining the behaviour of panels having a small
scale factor, whereas the rough mesh is adopted for evaluating the
behaviour of the models increasing the scale factor.
The FE models for the micropolar continuum (iii and iv) are
dened by subdividing panel length and height into nel,1 and nel,2
subdivisions, respectively; then 4 nel,1  nel,2 triangular elements
are dened (Fig. 7b, with nel,1 nel,2 8) in order to obtain a
symmetric mesh. The elastic moduli dened in section 5 (Eqs.
(22)e(25)) are assumed as elastic properties of the enriched
triangular FEs. In addition, a Cauchy continuum (v) is taken into
account. In this case and a traditional FE model is developed by
adopting the FE mesh in Fig 7b and by assuming the moduli in Eq.
22a,b and Eq. (23) as elastic properties of the triangular FEs. Such
model is obviously not able to furnish in-plane rotations, then, inplane translations obtained by all the models are also compared.
In the equivalent continuum and heterogeneous FE models,
boundary conditions e loads and restraints e are applied along the
edges of the panel, whereas in the DEM such conditions are
applied at block centres. In the following, uniform load distributions along the top edge of the panel are taken into account,
together with load distribution over small areas, in order to evaluate the effect of the load wavelength respect to the size of the
representative volume.
7.1. Rectangular panel
In the rst three examples a masonry panel with a running bond
texture pattern is considered (Fig. 7a). The panel is composed by
UNI bricks (b 250 mm, a 55 mm, s 120 mm), with bed and
head mortar joints having the same thickness e. The panel is
characterised by 6 blocks in horizontal direction and 16 courses in
vertical direction. The mechanical characteristics of the mortar are
Em 1 GPa and nm 0.2 and a standard joint thickness e 10 mm is
assumed. Hence, the overall dimensions of the panel are: length
L 1550 mm, height H 1030 mm and thickness s 120 mm.
In order to evaluate the effect of the size of heterogeneity on the
behaviour of the models, the number of blocks along both plane
directions is increased by introducing a geometry scale factor r L/
b and maintaining xed block width-to-height ratio and panel dimensions (L, H, s) as it was done by Salerno and De Felice (2009).
Hence, the panel in Fig. 7a is characterised by r 6. With this
approach, the order of magnitude of results does not vary if
different scale factors are taken into account; furthermore, the
elastic parameters of the Cauchy continuum are not affected by
such scale factor, hence results obtained with this model do not
vary for increasing number of blocks.
Three numerical examples are carried on varying load and restraint conditions for each case considered. Fig 8 shows the three
examples together with details of applied loads and restraints for

Fig. 6. Detail of the heterogeneous FE model: (a) rough mesh, (b) ne mesh.

46

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 7. Masonry panel considered for the numerical examples: (a) discrete model, (b) FE model for micropolar and Cauchy continua.

the panel with r 6. In rst and second example, the applied

horizontal load distribution is q 1000/L KN/m and in the third
example the applied force is F 1000 KN.
In order to compare the results given by the different models
taken into account, the following relative difference is dened:

dDEM  dFEM
 100;
dDEM

(27)

where d may represent a displacement or a rotation of the discrete

model or the equivalent continuum (u1, u2 or u3) and the superscript FEM may represent one of the four FE models e heterogeneous or continuous-taken into account (iiev). Relative differences
evaluated at several points, such as ends and mid-point of sections
AeA and BeB (Fig. 8), are collected in Table 1 and are represented in
Fig. 20aec for increasing scale factor r.
7.1.1. Example 1: panel subject to a horizontal shear load
In the rst example the panel is subject to a horizontal shear
load along the top edge and it is simply supported along the other
edges (Fig. 8, rst column). In this case horizontal translations u1
linearly proportional to y2 and negligible vertical translations u2 are
expected. Fig. 9 collects results relative to the panel having a scale

factor r 6. Due to the asymmetric load, asymmetric horizontal

translations and symmetric rotations are obtained.
The rst row of Fig. 9 shows horizontal translations obtained
with the DEM and the continuous FE models (iiiev). In this case, inplane translations determined with both micro-polar models and
with the Cauchy FE model turn out to be coincident. Moreover, the
maximum horizontal displacement at the top of the panel is equal
to 2.202 mm, which is coincident to the analytic solution of the
problem (umax
qH=ACauchy
1
1212 ).
Considering the in-plane rotations (second row of Fig. 9), the
discrete model furnishes non-uniform values over the panel,
with smaller values for the blocks along the edges of the panel,
where the boundary conditions are applied. Both micropolar FE
models give coincident rotations (u3 1.450  104 rad) and
their order of magnitude is acceptable if compared to the results
of the DEM.
Fig. 10a,b shows the in-plane rotations along section AeA and
BeB, respectively, obtained by modelling the panel with r 6 with
the different models. Along both sections, the values obtained with
the DEM and the heterogeneous FEM are not constant, whereas the
values obtained with the micropolar FE models are constant.
Considering a panel with a larger scale factor with respect to the

Fig. 8. Numerical examples for the rectangular masonry panel.

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

47

Table 1
Results obtained with DEM and relative differences obtained by adopting other models for modelling a rectangular panel subject to three different load and restraint conditions
(Fig. 8).
Ex.

Displacement

u1 (L/2, H)
u1 (L/2, H/2)

u3 (L/2, H)
2

u1 (L/2, H)

u3 (0, H)
u3 (0, H/2)
3

u2 (L/2, H)

u3 (0, H)
u3 (0, H/2)

5
30
5
30
5
30
5
30
5
30
5
30
5
30
5
30
5
30

DEM

2.082 [mm]
2.169 [mm]
1.145 [mm]
1.129 [mm]
2.308E-04 [rad]
1.462E-04 [rad]
3.594 [mm]
3.945 [mm]
4.806E-03 [rad]
6.113E-03 [rad]
3.108E-03 [rad]
3.946E-03 [rad]
1.662 [mm]
2.882 [mm]
8.528E-04 [rad]
7.839E-04 [rad]
4.707E-04 [rad]
3.606E-05 [rad]

standard case (r 30), Fig. 10c,d shows the in-plane rotations along
section AeA and BeB, respectively. In this case the values obtained
with the DEM and the heterogeneous FEM are almost constant
along the sections, except at section ends. Moreover with this scale
factor, results given by DEM and Heterogeneous FEM tend to be
closer to equivalent continuum solutions with respect to the case
characterised by r 6.

d [%]
FEM Hetero

FEM REV1

1.55
0.80
0.16
0.35
4.893
13.81
0.12
1.64
2.138
13.91
0.96
1.77
2.05
2.34
2.10
2.17
1.06
0.88

7.57
2.51
3.84
2.48
37.2
0.82
5.69
1.23
32.45
14.89
18.01
6.33
16.34
6.23
24.71
1.78
12.95
4.39

FEM REV2

FEM Cauchy

8.82
1.57
16.08
9.33
4.96
3.03
32.13
1.36
11.76
1.79
4.03
1.75

e
e
11.88
1.93
e
e
e
e
118.69
26.12
e
e
e
e

In order to evaluate the effect of size of heterogeneity on panel

displacement, with particular attention to the behaviour of
discrete, heterogeneous and micropolar models with respect to the
traditional Cauchy model, the horizontal displacement at the mid of
the top edge is evaluated increasing the scale factor (Fig. 11a)
together with the potential energy spent by the models (Fig. 11b). In
this example, due to the simple loading condition, the

Fig. 9. Horizontal translation (rst row) and in-plane rotation (second row) for the panel subject to a horizontal shear load, having r 6. Results given by DEM (rst column) and
microp. FEM (second column).

48

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 10. Rotations u3 along section AeA (a,c) and the BeB (b,d) obtained with the different models (ieiv) for a panel having r 6 (a,b) and r 30 (c,d) subject to a horizontal shear
load.

displacement and the energy obtained with continuum models are

constant for increasing r, whereas the values obtained with DEM
and heterogeneous FEM tend to converge to the results given by the
continuum models. Hence, as can be expected, the behaviour of
models i and ii converge to that of a traditional continuum for
increasing number of blocks. The heterogeneous FEM appears to be
more rigid than the other models due to the rough mesh adopted
for modelling the panel with increasing r.

Fig. 11. Horizontal displacement at the mid of the top edge (a) and expended energy
(b) obtained with the different models (iev) for a panel subject to a horizontal shear
load and increasing the scale factor.

7.1.2. Example 2: panel with xed base subject to a horizontal shear

load
In this example the panel is xed at the base and subject to a
horizontal shear load (Fig. 8, second column) in order to obtain a
exural behaviour characterised by non uniform rotations over the
panel. Fig. 12 shows the contours of horizontal translations and
rotations over the panel obtained with the DEM and micropolar FE
models, with r 6. Similarly to the previous example, asymmetric
horizontal translations and symmetric rotations are obtained.
In this case results given by models iii and iv are not coincident
and the micropolar FEM based on REV2 appears to be slightly more
deformable than that based on REV1. Fig. 13a,b shows the in-plane
rotations along sections AeA and BeB, respectively, for the panel
having r 6. The values obtained with the DEM are not constant
along both sections and the values obtained with the heterogeneous FEM are quite close to DEM results. Considering the micropolar FE models, the results obtained with model based on REV2
are quite close to those obtained with the DEM along both sections,
whereas results obtained with the model based on REV1 are quite
different respect to those obtained with the DEM, especially at the
ends of each section. Considering the panel with r 30, Fig. 13c,d
shows in-plane rotations along sections AeA and BeB, respectively.
In this case the results given by the different models are closer to
each other with respect to the case with r 6; however, rotations
given by model iv (REV2) appear to be closer to DEM rotations with
respect to those given by model iii (REV1), especially at section
ends.

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

49

Fig. 12. Horizontal translation (rst row) and in-plane rotation (second row) for the panel subject to a horizontal shear load, having r 6. Results given by DEM (rst column),
microp. FEM based on REV1 (second column), microp. FEM based on REV2 (third column).

Fig. 13. Rotations u3 along sections AeA (a,c) and BeB (b,d) for a panel with a xed base having r 6 (a,b) and r 30 (c,d) and subject to a horizontal shear load obtained with the
different models (ieiv).

50

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 14. Vertical stresses along section BeB obtained with different models.

Fig. 15. Horizontal displacement at the mid of the top edge (a) and expended energy (b) obtained with the different models (iev) for a panel with increasing scale factor r with xed
base and subject to a horizontal shear load.

Fig. 16. Vertical translation (rst row) and in-plane rotation (second row) for the panel subject to a symmetric vertical load at the top edge, having r 6. Results given by DEM (rst
column), microp. FEM based on REV1 (second column), microp. FEM based on REV2 (third column).

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

51

Fig. 17. Rotations u3 and vertical displacements u2 along sections AeA (a,c) and BeB (b,d) for a panel with a simply supported base having r 6 (a,b) and r 30 (c,d) and subject to
a symmetric vertical load at the top edge.

In this example the determination of the internal developed

stresses by the different models may be of particular interest.
Fig. 14a,b shows vertical stress along section BeB for panel with
r 6 and 30, respectively. In both cases results given by the numerical models considered do not show relevant differences.
However for r 30 the stresses obtained with the Cauchy model
approximate better those obtained with DEM with respect to the
micropolar models.
Similarly to the previous case, the horizontal displacement at the
mid of the top edge and the energy spent by models are evaluated
increasing the scale factor r (Fig. 15a,b). In this example, due to the

different restraint condition with respect to the previous case, only

displacement and energy obtained with the Cauchy continuum
model are constant for increasing r, whereas the values obtained
with models ieiv tend to converge to the results given by the Cauchy
model. As it has been found in Fig 12, the micropolar model based on
REV2 is more deformable than that based on REV1 and u1(L/2,H)
given by the micropolar model based on REV2 is closer to the constant value given by the Cauchy model with respect to the values
given by the micropolar model based on REV1. The heterogeneous
FEM appear to be more rigid than other models due to the rough
mesh adopted for modelling the panel with increasing r.

Fig. 18. Vertical stresses along section BeB obtained with different models.

52

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

relevant differences and the corresponding results obtained with

r 30 are not showed for simplicity.

Similarly to the previous example, Fig. 18a,b shows vertical
stress along section BeB for panel with r 6 and 30, respectively. In
both cases micropolar models slightly underestimate stresses obtained with DEM, however for r 6 the stresses obtained with the
micropolar model based on REV2 approximate better those obtained with DEM with respect to the micropolar model based on
REV1.
The effect of size of heterogeneity is taken into account in this
case by determining the vertical displacement at the mid of the top
edge for increasing r (Fig. 19). Models ieiv converge to Cauchy
solution more slowly with respect to the previous examples due to
the small wavelength of the applied load.
Fig. 19. Vertical displacement at the mid of the top edge obtained with the different
models (iev) for a panel with increasing scale factor r with xed base and subject to a
horizontal shear.

7.2. Square panel subject to a force acting on its diagonal

7.1.3. Example 3: panel subject to a symmetric vertical load

In this example the panel is simply supported at the base and
subject to a vertical force applied at the mid of the top edge (Fig. 8,
third column). This problem may represent a masonry wall loaded
by a beam of a wooden oor at the top edge, hence the size of the
load distribution is similar to the size of the REV and it is quite
smaller than the size of the panel.
Fig. 16 shows the maps of in-plane vertical translations and
rotations over the panel obtained with the DEM and micropolar
FE models, with r 6. In this case, due to the symmetric load,
symmetric vertical translations and asymmetric rotations are
obtained.
With particular attention to in-plane rotations, the contour
maps obtained with the FE model based on REV2 (third column of
Fig. 16) appear to be very close to DEM results (rst column of
Fig. 16). Fig. 17a,b shows in-plane rotations obtained with the
different models and evaluated along sections AeA and BeB of the
panel having r 6. In this case DEM and heterogeneous FEM results
are in excellent agreement; moreover Fig. 17a,b shows clearly that
the micropolar FE model based on REV2 is closer to the DEM than
the micropolar FE model based on REV1 along both sections
considered. If the panel with r 30 is taken into account, Fig. 17c,d
conrm that the micropolar FE model based on REV2 approximates
better the DEM. In this case, Fig. 17d shows a detail of in-plane
rotations in the middle of section AeA in order to appreciate that
results given by the micropolar model based on REV1 are far from
other results. Far from the loaded region (section BeB), in-plane
rotations obtained with the different models do not show

In order to have a further validation of the discrete and micropolar continuum models, part of an existing numerical experimentation carried out by Pau and Trovalusci (2012) and Trovalusci
and Pau (2014) is taken into account as benchmark solution. In
particular, analyses are limited to the cases a1, a2 and a3 of both
papers, characterised by scale factors r L/b equal to 5, 10 and 20,
respectively, and having block width-to-height ratio b/a 4, which
is quite close to that of UNI bricks of the previous examples. The
panel is square, with length
pand height L 8000 mm and it is
subject to a force F 1000 2 KN acting along the diagonal of the
panel (Fig 21). In this example mortar is modelled accordingly to
Pau and Trovalusci (2012). The thickness of horizontal and vertical
mortar joints is equal to 10 mm and it is kept constant and not
inuenced by the scale factor, in order to have the same joint
stiffness for varying blocks' size.
Fig. 22aec shows the contour maps of rotations over the panels
modelled by DEM; as expected, the larger the scale factor, the
greater the deformability of the panel in terms of in-plane rotations. The order of magnitude of results is in quite good agreement
with benchmark results that may be found in Pau and Trovalusci
(2012). Differently than benchmark results, contour lines in
Fig. 22 highlight the masonry texture pattern of each panel. Then,
following the work done for the benchmark results, rotations
(Fig. 23, rst column) and vertical translations (Fig. 23, second
column) given by the different models ieiv are evaluated along the
left edge of the panel (y1 0) for the three scale factors considered.
Considering in-plane rotations, the micropolar model based on
REV2 ts DEM results better than the model based on REV1.
However, differently with respect to benchmark results, the Cauchy

Fig. 20. Relative differences obtained by adopting continuous FEMs and Heterogeneous FEM in comparison with DEM for determining displacements at several points of a
rectangular panel subject to three different load and restraint conditions and increasing scale factor r.

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 21. Masonry square panel having r 5 subject to a force acting along its diagonal.

model is more deformable than other models and displacements

obtained with micropolar model converge to those obtained with
the Cauchy model for increasing scale factor.
8. Conclusions
In the present work, a deep investigation of different models
that may be adopted for studying the mechanical behaviour of
masonry is presented, with particular attention to the opportunity

53

of using a micropolar model instead of a traditional Cauchy continuum model.

The discrete model introduced by Cecchi and Sab (2004), which
considers masonry as a skeleton with rigid blocks and elastic
mortar interfaces, is adopted for studying the in-plane behaviour
of the panels and results are taken as reference solution. Moreover,
an identication between the discrete model and a 2D continuum
is carried out by equating the mechanical work in the two models
for a class of regular motions. Two different REVs are taken into
account for representing the micropolar continuum and a triangular FE in plane stress state enriched by rotational nodal degrees
of freedom is adopted for modelling the micropolar continuum
and representing the behaviour of in-plane loaded periodic
brickworks.
The discrete model allows to calculate the rotations of the blocks
in masonry walls and allows to compare the rotations determined
by the micropolar FE models (based on two REVs). In addition, a
traditional heterogeneous FE model is taken into account in which
mortar joints are modelled as a continuous material.
The proposed examples investigate the behaviour of masonry
panels subject to both concentrated and distributed loads, in order
to evaluate the effect of the load wavelength with respect to panel
size. Moreover, the proposed examples investigate the effects of
size of heterogeneity on masonry behaviour by introducing a scale
factor r L/b.

Fig. 22. Contour maps of in-plane rotations for square panels subject to a diagonal force and with r equal to 5 (a), 10 (b) and 20 (c).

54

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

Fig. 23. Comparison of rotations (rst column) and vertical displacements (second column) along vertical line y1 0.

The DEM model turns out to be more deformable than the

heterogeneous FE model, however results in terms of in-plane rotations and displacements appear to be in very good agreement.
Then, this comparison conrms that the DEM can be the reference
solution for evaluating the effectiveness of the micropolar FE
models.
Comparing results in terms of rotations given by the micropolar
FE models with respect to the DEM, the micropolar FE model based
on REV2 turns out to be more appropriate than that based on REV1.
Results given by the FE model based on REV2 turn out to be
generally closer to those given by the DEM than those given by the
FE model based on REV1. In particular, considering rotations close
to the loaded regions, the micropolar FE model based on REV1
completely fails to approximate DEM results.

It is worth noting that the micropolar FE model based on REV2 is

quite more deformable than that based on REV1. Then, considering
results in terms of translations, the behaviour of micropolar FE
model based on REV1 appears to be closer to that of DEM with
respect to the model based on REV2. However the differences of
model based on REV2 with respect to the DEM are acceptable.
As can be expected, the FEM based on Cauchy continuum turns
out to be more deformable than DEM, heterogeneous FEM and both
micropolar models. It is worth noting that increasing panel size
with respect to block size, the behaviour of micropolar models
converge rapidly towards that of the Cauchy model, whereas
convergence is slow for the DEM and heterogeneous FEM.
Then, the micropolar model turns out to be suitable and effective in modelling the in-plane behaviour of masonry panels

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

characterised by a small or quite small number of blocks. The REV2,

characterised by four horizontal interfaces and a vertical interface
at the centre of the cell, represents the best choice for performing a
homogenisation or identication of the micropolar continuum. The
micropolar FE model adopted, based on a simple enrichment of a
traditional triangular element, is simple and effective for representing the behaviour of masonry panels subject to in-plane loads
and turns out to be more efcient with respect to a traditional
heterogeneous FE model, that requires a large computational effort.
Acknowledgements
The research has been carried out thanks to the nancial support of PRIN 2010e2011 (under grant 2010NRBMTP, project
Models and algorithms for the nonlinear analysis of structures and
the validation of performance-based design rules).
Appendix
A1. Compatible identication

Pv

1
2

55

dxT Kv dxdA

8
9
h
i=
1<
Gv =ev Sv D2 2 Kv =ev Sv D1 2 Iv3 d3 2 :

;
2:

(A.1b)

Where Sh b s/2 represents the area of the horizontal interface

and Ih3 b3s/96 is its inertia with respect to y3 axis. Sv a s is the
area of the vertical interface and Iv3 a3s/12 is its inertia respect
to y3 axis. Kh and Kv are the interfacial stiffness matrices. It is
worth noting that the jumps of displacements are functions of
block translations and rotations (see for details Cecchi and Sab,
2004).
The total work PREV spent by the REV is obtained by adding
together the works of horizontal and vertical interfaces and averaging the total sum over the REV. PREV is function of (grad u U)
and grad U. According to the identication between discrete and
continuous systems, by derivation respectively versus.

The components of equivalent in-plane tensor A and microcouple tensor L may be obtained by dening the work of contact



v2 PREV u1;1 ; u2;2 ; u1;2 ; u2;1 ; u3 ; u3;1 ; u3;2
vu21;1


v2 PREV u1;1 ; u2;2 ; u1;2 ; u2;1 ; u3 ; u3;1 ; u3;2
vu21;2


v2 PREV u1;1 ; u2;2 ; u1;2 ; u2;1 ; u3 ; u3;1 ; u3;2
vu23;1



v2 PREV u1;1 ; u2;2 ; u1;2 ; u2;1 ; u3 ; u3;1 ; u3;2
vu22;2


v2 PREV u1;1 ; u2;2 ; u1;2 ; u2;1 ; u3 ; u3;1 ; u3;2
vu22;1


v2 PREV u1;1 ; u2;2 ; u1;2 ; u2;1 ; u3 ; u3;1 ; u3;2
vu23;2

actions at the interfaces generated by simple deformation states

and then differentiating the sum of the works of the REVs with
respect to the deformation components.
The deformation states considered are: i) horizontal extension
E11, ii) vertical extension E22, iii) horizontal shear E12  u3, iv)
vertical shear E21 u3. For each deformation state, the relative
displacements and rotations of blocks are evaluated over the REV
and Eq. (10) is applied for all the interfaces of the REV, taking into
account their constitutive relationship (Eq. (13)). It is worth noting
that REV1 is characterised by eight horizontal interfaces shared
with contiguous REVs and two vertical interfaces, whereas REV2 is
characterised by four horizontal interfaces shared with contiguous
REVs and one vertical interface.
The work spent over a horizontal and a vertical interface may be
written in terms of the jump of displacements (D1, D2) and rotations
(d3) between adjacent blocks as follows (Cecchi and Sab, 2004):

Ph

1
2

 
dxT Kh d x dA

8
2
3
2
39
=
1<
2
2
2
Gh =eh 4Sh D1 5 Kh =eh 4Sh D2 Ih3 d3 5 ;

;
2:
(A.1a)

(A.2a,b)

(A.2c,d)

(A.2e,f)

the homogenized coefcients of Eqs. (22), (24) and (25) may be

obtained.

A2. Finite element models for masonry

A2.1. Heterogeneous FE models
The heterogeneous FE model is represented by a traditional 2D
FE model where standard quadrilateral FE elements in plane stress
state are used and mortar joints and blocks are distinguished by
adopting different elastic parameters. Such type of model was
widely used in past by many authors for developing analysis in both
elastic and inelastic eld (Stafford Smith and Rahman, 1972; Page,
1978; Ali and Page, 1988). In the rst three examples of the present work, materials are isotropic and the elastic modulus of the
blocks is assumed 104 times larger than Em in order to simulate the
innite rigidity of blocks respect to mortar joints. Such model allows to apply the same boundary conditions applied to continuum
models. However, the limit of this model is represented by the huge
number of degrees of freedom needed for modelling a masonry
panel, in particular if a ne mesh renement is adopted.
Then, a numerical test is carried on in order to dene the most
correct mesh renement of the heterogeneous FE model for representing the in-plane behaviour of masonry panels with regular
texture. It is clear that quadrilateral elements adopted for

56

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

modelling mortar joints may be slender if compared to elements

adopted for blocks, due to the small joint thickness with respect to
block dimensions. Then, the mesh renement of mortar joints may
be crucial for modelling the overall masonry behaviour correctly.
Considering the masonry texture introduced in paragraph 7.1,
Fig. A1 shows eight different mesh renements, starting from a
coarse one with one subdivision along the thickness of both blocks
and mortar joints and concluding with a very accurate mesh
characterised by sub-elements having each side equal to 1/4 of
mortar joint thickness.
Convergence tests are performed on the panel described in 7.1,
having scale factor r 6, and considering load and restraint conditions of Fig. 8. Table A1 collects the DOF number of the panel
modelled by adopting the eight different mesh renements of
Fig. A1. It is worth noting that the DOF number of the panel with
r 6 modelled by DEM is 312, hence it is clear that the heterogeneous FE model requires a large computational effort with respect
to the DEM even if a coarse renement is considered. For this
reason, an accurate mesh renement may be unsuitable for
modelling panels with large scale factors. Errors committed in
determining panel translations and rotations with different mesh
renements with respect to DEM results (Eq. (27)) are evaluated for
a panel with r 6. Fig. A2 show errors for translations e horizontal
for examples 1 and 2 and vertical for example 3 e and rotations
evaluated at several points on the panel for increasing DOF number.
Considering results in terms of displacements, the rst row of
Fig. A2 shows that errors of the heterogeneous FE models with
respect to DEM ones decrease for increasing mesh renement. As
can be expected, displacements obtained adopting four subdivisions along joint thickness are close to DEM ones. However,
adopting one subdivision along mortar joint thickness, errors are
close to 2%, whereas with two subdivisions along mortar joint
thickness, errors are close to 1%. The second row of Fig. A2 shows
that rotations obtained with heterogeneous FEMs do not converge
to DEM in rst and second example. This aspect is motivated by the
differences of load and restraint conditions adopted by the models,
that are more evident for low values of the scale factor r. In fact in
the DEM, loads and restraints are applied at block centres, whereas
in the heterogeneous FEM loads and restraints are applied along
panel boundaries. In the third example, due to the load applied over
a small area, differences between DEM and heterogeneous FEM are
less evident and the errors in determining rotations converge to
zero for increasing mesh renement and two subdivisions along
joint thickness appear to be sufcient for obtaining an error close to
1%.
Considering the number of DOFs collected in Table A1, it is clear
that accurate mesh renements require a huge computational
effort if large scale factors r are considered. Then, for the present
analyses with the heterogeneous FEM for increasing scale factor
r L/b, mesh renement of type b or c may be adopted, whereas for
individual analyses with small values of r, mesh renement of type
d may be adopted.
A2.2. FE model for micropolar continuum
With reference to Fig. 5, translations u1, u2 and rotations u3
within the triangular element may be related to the nodal degrees
of freedom by means of a matrix N*, which collects the shape
functions:

8 9 2
3
< u1 =
N1 0 0 N2 0 0 N3 0 0
1
2
3
u
4 0 N 0 0 N 0 0 N 0 5qel N* qel
: 2;
u3
0 0 N1 0 0 N2 0 0 N3

(A.3)

Then, the strain-displacement relations for the micropolar

continuum may be written in matrix form as follows:

el f11 22 12 21 c13 c23 gT

T
u1;1 u2;2 u1;2 u3 u2;1  u3 u3;1 u3;2 IDN qel B qel
(A.4)
where ij and cij represent strains and curvatures within the
element, respectively; ID is the matrix of the derivatives that and B*
is the strain-displacement matrix. Differently than the standard
CST element (Zienkiewicz and Taylor, 1989), the straindisplacement matrix B* does not have constant elements everywhere, due to the uj3 terms that are needed to dene the shear
strains. However, matrix B* becomes a constant matrix if the shape
functions N j representing the terms uj3 are replaced by their values
at the centre of the triangle.
In a generic micro-polar elastic continuum, force and couple
stresses within the element are related to deformations by six
elastic constants (Eringen, 1966). In this case, for modeling a
running bond masonry panel, the six constants are determined
towards the identication described in section 5. Hence, the
stressestrain relationship may be written as follows.

sel f s11

s22

s12

s21

m13

m31 gT

A
0


0 el
Del
L
(A.5)

Where the matrix A collects the Aijhk moduli (Eq. 22aee) and L
collects the Lij moduli (Eq. 25a,b). Due to Eq. (22c), matrix D turn
out to be diagonal.
Then, the stiffness matrix Kel of the triangular element may be
determined as usual starting from the potential energy Pel of the
triangular element and introducing the expressions for the stresses
and the strains:

Pel

1
2

Z 
Ael

sel

T

el dA

1 el T el el
q
K q ;

1
2

Z 

qel

T

B*T DB* qel dA

Ael

(A.6)

where Ael is the area of the triangular element. It is worth noting

that Eq. (A6) is carried on by applying a full integration over the
area of the element, in fact the integration of the linear and bilinear
terms of the shape functions N j over a triangular area does not
require particular effort.
The stiffness matrix of element adopted turns out to be equal to
that of the linear triangle proposed by Providas and Kattis (2002),
that considered nodal displacement and rotation vectors separately
and dened consequently three different strain displacement
matrices, obtaining a stiffness matrix having four sub matrices, in
order to identify the one corresponding to displacements which is
equal to the well known stiffness matrix of the CST. It is worth
noting that the proposed triangular element satises the patch
tests proposed by Providas and Kattis.

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

57

Fig. A1. Different mesh renements taken into account for modelling the behaviour of masonry with regular texture.

Fig. A2. Relative differences in determining translations and rotations at several points of the panel obtained with the heterogeneous FE models, increasing mesh renement, with
respect to DEM results.

Table A1
Number of degrees of freedom (DOFs) of the different heterogeneous FE models considered for the convergence test for a panel having r 6.
Mesh renement type

DOFs (r 6)

1536

3456

5776

17,280

22,610

68,370

89,586

271,986

58

D. Baraldi et al. / European Journal of Mechanics A/Solids 50 (2015) 39e58

References
Addessi, D., Sacco, E., Paolone, A., 2010. Cosserat model for periodic masonry
deduced by nonlinear homogenization. Eur. J. Mech. A Solids 29, 724e737.
Ali, S.S., Page, A.W., 1988. Finite element model for masonry subjected to concentrated loads. J. Struct. Eng. ASCE 114 (8), 1761e1784.
Anthoine, A., 1995. Derivation of the in-plane elastic characteristics of masonry
through homogenization theory. Int. J. Solids Struct. 32 (2), 137e163.
Bacigalupo, A., Gambarotta, L., 2012. Computational two-scale homogenization of
periodic masonry: Characteristiclengths and dispersive waves. Comput. Meth.
Appl. Mech. Eng. 213e216, 16e28.
Baraldi, D., Reccia, E., Cazzani, A., Cecchi, A., 2013. Comparative analysis of numerical discrete and nite element models: the case of in-plane loaded periodic
brickwork. Compos. Mech. Comput. Appl. Int. J. 4 (4), 319e344.
Baraldi, D., Cecchi, A., 2014. Discrete element model for in plane loaded viscoelastic
masonry. Int. J. Multiscale Comp. Eng. 12 (2), 155e175.
Beveridge, A.J., Wheel, M.A., Nash, D.H., 2013. The micropolar elastic behaviour of
model macroscopically heterogeneous materials. Int. J. Solids Struct. 50 (1),
246e255.
Boutin, C., 1996. Microstructural effects in elastic composites. Int. J. Solids Struct. 33
(7), 1023e1051.
Casolo, S., 2004. Modelling in-plane micro-structure of masonry walls by rigid elements. Int. J. Solids Struct. 41 (13), 3625e3641.
Casolo, S., 2006. Macroscopic modelling of structured materials: relationship between orthotropic Cosserat continuum and rigid elements. Int. J. Solids Struct.
43 (3e4), 475e496.
Casolo, S., 2009. Macroscale modelling of microstructure damage evolution by a
rigid body and spring model. J. Mech. Mater. Struct. 4 (3), 551e570.
Cecchi, A., Rizzi, N.L., 2005. Modelli 2D con microstruttura per pannelli di muratura
in 3D. In: Proc. XVII congresso AIMETA di meccanica teorica e applicata, Firenze,
11e15 sett..
Cecchi, A., Sab, K., 2002. A multi-parameter homogenization study for modeling
elastic masonry. Eur. J. Mech. A Solids 21 (2), 249e268.
Cecchi, A., Sab, K., 2004. A comparison between a 3D discrete model and two
homogenised plate models for periodic elastic brickwork. Int. J. Solids Struct. 41,
2259e2276.
Cecchi, A., Sab, K., 2009. Discrete and continuous models for in plane loaded
random elastic brickwork. Eur. J. Mech. A Solids 28 (3), 610e625.
Cook, R.D., Malkus, D.S., Plesha, M.E., Witt, R.J., 2001. Concepts and Applications of
Finite Element Analysis. Wiley.
Cundall, P.A., Strack, O.D.L., 1979. A discrete numerical model for granular assemotechnique 29 (1), 47e65.
blies. Ge
Cundall, P.A., 1988. Formulation of a three-dimensional distinct element model e
part I: a scheme to detect and represent contacts in a system composed of many
polyhedral blocks. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 25 (3), 107e116.
De Bellis, M.L., Addessi, D., 2011. A Cosserat based multi-scale model for masonry
structures. Int. J. Multiscale Comp. Eng. 9 (5), 543e563.
De Buhan, P., De Felice, G., 1997. A homogenization approach to the ultimate
strength of brick masonry. J. Mech. Phys. Solids 45 (7), 1085e1104.
Del Piero, G., 2009. On the method of virtual power in continuum mechanics.
J. Mech. Mater. Struct. 4 (2), 281e292.
Eringen, A.C., 1966. Linear theory of micropolar elasticity. J. Math. Mech. 15 (6),
909e923.
Forest, S., Pradel, F., Sab, K., 2001. Asymptotic analysis of heterogeneous Cosserat
media. Int. J. Solids Struct. 38, 4585e4608.
Formica, G., Sansalone, V., Casciaro, R., 2002. A mixed solution strategy for the
nonlinear analysis of brick masonry walls. Comput. Meth. Appl. Mech. Eng. 191,
5847e5876.
Gambarotta, L., Lagomarsino, S., 1997. Damage models for the seismic response of
brick masonry shear walls part ii: the continuum model and its application.
Earthq. Eng. Struct. Dyn. 26, 441e462.
Itasca, 1989. UDEC (Universal Distinct Element Code) Version ICG1.5 User's Manual.
Itasca, 2000. UDEC Universal Distinct Element Code Manual. Itasca Consulting
Group Inc.
Lemos, J.V., 2007. Discrete element modeling of masonry structures. Int. J. Arch.
Herit. 1, 190e213.

Lofti, H.R., Benson Shing, P., 1994. Interface model applied to fracture of masonry
structures. J. Struct. Eng. ASCE 120, 63e80.
Logan, D.L., 2012. A First Course in the Finite Element Method. Cengage Learning.
Loureno, P.B., Rots, J.G., 1997. On the use of homogenisation techniques for the
analysis of masonry structures. Mason. Int. 11 (1), 26e32.
Luciano, R., Sacco, E., 1998. Variational methods for the homogenization of periodic
heterogeneous media. Eur. J. Mech. A Solids 17 (4), 599e617.
Mahabadi, O.K., Lisjak, A., Munjiza, A., Grasselli, G., 2012. Y-Geo: a new combined
nite-discrete element numerical code for geomechanical applications. Geomechanics 12, 676e688.
Masiani, R., Rizzi, R., Trovalusci, P., 1995. Masonry as structured continuum. Meccanica 30 (6), 673e683.
Masiani, R., Trovalusci, P., 1996. Cauchy and Cosserat materials as continuum
models of brick masonry. Meccanica 31 (4), 421e432.
Massart, T.J., Peerlings, R.H.J., Geers, M.G.D., 2007. An enhanced multi-scale
approach for masonry wall computations with localization of damage. Int. J.
Numer. Methods Eng. 69 (5), 1022e1059.
Milani, G., Loureno, P.B., Tralli, A., 2006. Homogenised limit analysis of masonry
walls, part I: failure surfaces. Comput. Struct. 84 (3e4), 166e180.
Munjiza, A., 2004. The Finite/Discrete Element Method. John Wiley and Sons,
Chicester.
Page, A.W., 1978. Finite element model for masonry. J. Struct. Div. 104 (8),
1267e1285.
Pau, A., Trovalusci, P., 2012. Block masonry as equivalent micropolar continua: the
role of relative rotations. Acta Mech. 223 (7), 1455e1471.
Pegon, P., Anthoine, A., 1997. Numerical strategies for solving continuum damage
problems with softening: application to the homogenization of masonry.
Comput. Struct. 64, 623e642.
Providas, E., Kattis, M.A., 2002. Finite element method in plane Cosserat elasticity.
Comput.. Struct. 80, 2059e2069.
Sacco, E., 2009. A nonlinear homogenization procedure for periodic masonry. Eur. J.
Mech. A Solids 28 (2), 209e222.
Salerno, G., Bilotta, A., Porco, F., 2001. A nite element with micro-scale effects for
the linear analysis of masonry brickwork. Comput. Meth. Appl. Mech. Eng. 190
(34), 4365e4378.
Salerno, G., De Felice, G., 2009. Continuum modeling of periodic brickwork. Int. J.
Solids Struct. 46 (5), 1251e1267.
Smyshlyaev, V.P., Cherednichenko, K.D., 2000. On rigorous derivation of strain
gradient effects in the overall behaviour of periodic heterogeneous media.
J. Mech. Phys. Solids 48, 1325e1357.
Stafford Smith, B., Rahman, K.M.K., 1972. The variation of stresses in vertically
loaded brickwork walls. ICE Proc. 51 (4), 689e700.
Stefanou, I., Sulem, J., Vardoulakis, I., 2008. Three-dimensional Cosserat homogenization of masonry structures: elasticity. Acta Geotech. 3, 71e83.
Stefanou, I., Sulem, J., Vardoulakis, I., 2010. Homogenization of interlocking masonry
structures using a generalized differential expansion technique. Int. J. Solids
Struct. 47, 1522e1536.
Sulem, J., Mhlhaus, H.B., 1997. A continuum model for periodic two-dimensional
block structures. Mech. Cohesive Frict. Mat. 2, 31e46.
Trovalusci, P., Masiani, R., 2003. Non-linear micropolar and classical continua for
anisotropic discontinuous materials. Int. J. Solids Struct. 40, 1281e1297.
Trovalusci, P., Pau, A., 2014. Derivation of microstructured continua from lattice
systems via principle of virtual works: the case of masonry-like materials as
micropolar, second gradient and classical continua. Acta Mech. 225 (1), 157e177.
Wheel, M.A., 2008. A control volume based nite element method for plane
micropolar elasticity. Int. J. Numer. Meth. Eng. 75 (8), 992e1006.
Wei, X., Hao, H., 2009. Numerical derivation of homogenized dynamic masonry
material properties with strain rate effects. Int. J. Impact Eng. 36, 522e536.
Zhang, H.W., Wang, H., Liu, G., 2005. Quadrilateral isoparametric nite elements for
plane elastic Cosserat bodies. Acta Mech. Sin. 21, 388e394.
Zhang, H.W., Xie, Z.Q., Chen, B.S., Xing, H.L., 2012. A nite element model for 2D
elasticeplastic contact analysis of multiple Cosserat materials. Eur. J. Mech. A
Solids 31 (1), 139e151.
Zienkiewicz, O.C., Taylor, R.L., 1989 The Finite Element Method, fourth ed., vol. 1.
McGraw-Hill.