You are on page 1of 30

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

net/publication/257896251

Diagonal compressive strength of masonry samples - Experimental and


numerical approach

Article  in  Materials and Structures · January 2014


DOI: 10.1617/s11527-012-9933-z

CITATIONS READS

13 615

3 authors:

Rui Miguel de Sousa Hipólito Sousa


University of Porto University of Porto
26 PUBLICATIONS   53 CITATIONS    55 PUBLICATIONS   191 CITATIONS   

SEE PROFILE SEE PROFILE

João Miranda Guedes


University of Porto
90 PUBLICATIONS   344 CITATIONS   

SEE PROFILE

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

Parede Eficiente Plus Saint-Gobain Weber/ Efficient Wall Plus Saint-Gobain Weber View project

Wall Cladings and Structural Masonry View project

All content following this page was uploaded by Rui Miguel de Sousa on 16 March 2015.

The user has requested enhancement of the downloaded file.


Diagonal Compressive Strength of Masonry
Samples - Experimental and Numerical
Approach

Abstract

Masonry is a structural material that presents a quite complex behaviour that depends on the
mechanical and geometrical characteristics of the units, the mortar and the link between these two
elements. In particular, the characterization of the shear behaviour of masonry elements involves
proper experimental campaigns that make these analyses particularly expensive. The main
objective of this paper is to present a case study on the characterization of the shear behaviour of
masonry through a methodology that merges a small number of laboratory tests with computer
simulations. The methodology is applied to a new masonry system that has recently been
developed in Portugal, and involves a FEM numerical approach based on micro3D modelling of
masonry samples using nonlinear behaviour models that are calibrated through a small number of
laboratory tests. As a result, the characterization of the masonry shear behaviour trough this
methodology allowed simulating, with reasonably accuracy, a large set of expensive laboratory
tests using numerical tools calibrated with small experimental resources.

Keywords: Masonry; Lightweight concrete units; Diagonal compression;


Laboratory tests; Computer simulations; Sensitivity analysis.

List of symbols
An Net area of masonry sample
b Height of the masonry sample
d Isotropic scalar degradation variable
dc Compressive damage variable
dmax maximum aggregate size
Initial (undamaged) elastic stiffness
dt Tensile damage variable
e Thickness of the mortar parallel joints
E0 Modulus of elasticity
F Compression load
fl Tensile flexural strength
Fmax Maximum compression load
Fmax Maximum compression load
Gp Potential plastic flow
g Full width of the mortar strips
G Shear modulus
GF Fracture energy
GFo Base value of the fracture energy
h Length of the masonry sample
hs Depth of a sample

1
Kc Ratio between the tensile and compressive stress invariants at initial yield
L Distance between measurement points of ∆v and ∆h
n Percentage of gross area of the unit that is solid
Effective hydrostatic pressure
Von Mises equivalent effective stress
R2 Coefficient of determination
sc Weight factor to control the recovery of the compressive stiffness
st Weight factor to control the recovery of the tensile stiffness
t Total thickness of the wall
αi,βi,γi Adimensional parameters
γ Shear strain
γmax Shear strain for the τmax
∆h Horizontal extensions
∆v Vertical shortening
Total strain
εcu Strain for σcu or ultimate strain
εc,limit Limit compressive strain
Elastic strain
εc Compression strain
εmax Strain of masonry for the Fmax
Plastic strain
Multi-axial equivalent plastic strain
Compressive equivalent plastic strain
Tensile plastic strain
εt Tensile strain
ν Poisson coefficient
Cauchy stress
Effective stress
Initial equi-biaxial compressive yield stress
σc Uniaxial compression stress
Compressive effective stresses
Initial uniaxial compressive yield stress
σcu Compressive strength (maximum compression stress)
Maximum principal effective stress (algebraic value)
σt Uniaxial tensile stress
Tensile effective stresses
σto Uniaxial tensile strength
τ Shear stress
τmax Shear strength or maximum shear stress
Dilation angle
Parameter that defines the rate at which Gp approaches the asymptote

1 Introduction

Masonry is a material that presents a quite complex behaviour that depends on


the mechanical and geometrical characteristics of the units, the mortar and the link

2
between these two elements. In particular, this complexity arises from the large
scatter observed in the mechanical characteristics of the materials involved and of
the interface between units and mortar. In fact, experimental results are essential
to characterize the units and the mortar as separated elements, but also the
masonry through samples where both elements and their interactions are taken
into account. The results and the knowledge got from these campaigns have been
essential to define the empirical based rules which have been adopted in most
design codes for masonry and, moreover, to develop and calibrate numerical tools
to simulate masonry structures.
However, this involves expensive resources, even when small size samples are
used. In general, laboratory tests are difficult to prepare and require specific
instrumentation and careful execution. On the other hand, the development of
computer tools and numerical models has made possible to reduce the difficulties
and costs of experimental characterization by means of computer-simulated tests.
Nevertheless, the accuracy and efficiency of these tools depends on the available
experimental information necessary to calibrate the numerical models. After being
properly verified and calibrated through experimental data, numerical tools can be
a very powerful tool to simulate and extrapolate experimental results to other
conditions with lower time consuming and financial costs.
Many behaviour models have been developed during the last decades to
simulate concrete type materials (with tensile brittle behaviour) and, in association
to this, masonry type materials. In this last case, the simulations can be done
either by using refined models, where units, mortar and interfaces are represented
separately through proper behaviour laws calibrated for the different elements, or
by using more global models, where masonry is considered to be a heterogeneous
material ruled by a unique behaviour law that integrates its global behaviour.
This paper describes the characterization of the shear behaviour of a new
masonry system made with lightweight concrete units that has been recently
developed in Portugal. This characterization was made through the application of
an integrated method where numerical analyses are used together with a limited
number of experiments. The results allowed characterizing the shear behaviour of
the masonry system for a set of pre-defined characteristics, but also for controlled
variations on some of the geometrical and mechanical properties of the masonry,
following an extensive sensitivity analysis.

3
2 Literature Review

2.1 New masonry systems

Considerable changes have been reported on masonry wall construction during


the last decades with the introduction or extended use of lightweight materials and
new types of units (Hendry 2001). In more recent years, new concepts for
masonry systems have been developed in some Southern European countries.
These systems are considered innovative, since they are able to accomplish
several functionality aspects (e.g., structural and fire resistance, thermal and
acoustic insulation, water tightness) without the need for complementary
measures. Many of these systems consist on one-leaf walls made of lightweight
materials and large units with high percentage of voids. These new units are
lighter, improving block handling, but keeping structural resistance (Coz Díaz et
al. 2011). The need to understand and to properly characterize the mechanical
behaviour of these innovative systems has motivated experimental and numerical
studies, although the majority of these studies are related to the behaviour of clay
masonry (e.g. da Porto et al. 2011; Mosele and da Porto 2011; Lourenço et al.
2010).

2.2 Masonry Analysis

Masonry, in general, is a structural brittle material made of units jointed


together using, usually, hydraulic mortar that makes the units work together. The
different construction methods, and the large geometrical and mechanical
variations that both units and mortar may exhibit, make the characterization of
masonry a complex issue, in particular if no experimental studies exist to sustain
these analyses. Calderini et al. (2010), for instance, based on observations of
seismic damage on complex masonry walls and on experimental laboratory tests,
report that masonry subjected to in-plane loading has two typical types of
behaviour which are associated to different failure modes:
• flexural behaviour (rocking, crushing);
• shear behaviour (sliding shear failure, diagonal cracking).
4
However, a mixed shear / flexural behaviour may also occur, depending on the
intensity of the axial loading. The behaviour depends on parameters such as: the
geometry of the wall (units and wall aspect ratio and cross-section), the
mechanical characteristics of the masonry materials (mortar joints, units and
interfaces), the boundary conditions and the level of axial load. In fact, most of
the available studies on masonry refer to experimental campaigns involving
masonry samples using different units, mortar and types of joints. Many non-
linear models have been developed or modified to fit the masonry behaviour. In
general, the models are based on a Finite Element (FE) formulation where non-
linear damage constitutive models and friction models are used (e.g. Brasile 2010;
Uva and Salerno 2006; Berto et al. 2004). However, models based on Discrete
Element (DE) formulations have also been reported with success (Lemos 2007).
The simulation of masonry structures through a FE formulation can follow
models with different detailing approaches. The less detailed approach represents
masonry as an equivalent continuum material with an appropriate constitutive
global homogeneous model (e.g. Mistler et al. 2007), while the more detailed
represents the masonry units, the joints and the interfaces explicitly, i.e. using
different FE elements and material models (e.g. Lourenço and Rots 1997;
Chaimoon and Attard 2007). The first approach is more suitable for modelling the
global response of large structures due to its lower time consuming and
computational cost, in opposition to the second approach, which use is only
justifiable when more refined analysis are requested in smaller samples.
Nevertheless, the more detailed analyses have also been used to determine the
necessary parameters to calibrate more global models, or as a complement, or an
alternative to laboratory experiments (Berto et al. 2004).
Concerning the constitutive models, recent studies suggest the use of plastic
damage constitutive models for FE analysis of masonry elements (e.g. Pelà et al.
2011; Berto et al. 2004). In general, these constitutive models were first
developed to simulate other brittle materials, such as concrete or rocks (e.g. Faria
et al. 1998; Lubliner et al. 1989; Tao and Phillips 2005), allowing the
consideration of their main failure mechanisms, i.e. cracking and crushing through
the use of two scalar variables which monitor local damage under tension and
compression.

5
Many studies are found involving the use of experimental data to calibrate and
validate constitutive models that are used in the numerical analyses. However,
only a few involved the prediction of masonry behaviour, being mostly oriented to
the in-plane compression (e.g. Barbosa et al. 2010; Vyas and Reddy 2010) and in-
plane shear behaviour (e.g. Fouchal et al. 2009; Gabor et al. 2006) of masonry
samples.
Regarding shear behaviour, Fouchal et al. (2009) developed a numerical model
based on the adhesion intensity to simulate the mechanical behaviour of the
interfaces in masonry elements. The mechanical characteristics were determined
experimentally using samples made of solids bricks and mortar. The numerical
simulation of masonry shear tests and the comparison to available experimental
data (triplets tested according to European standard EN 1052-3 and diagonal
compression tests based on American standard ASTM E519) gave good results
and provided the coefficients required to model the interfaces: stiffness
parameters and viscosity and friction coefficients.
Gabor et al. (2006) present a numerical and an experimental analysis of the in-
plane shear behaviour of masonry panels made with clay hollow units. A two
dimensional FE model was used adopting an elastic behaviour for the units and an
elastic-perfectly plastic formulation (Drucker–Prager) for the mortar, i.e.
considering the overall non-linear behaviour of the masonry concentrated on the
mortar that was considered perfectly bonded to the units. The material parameters
were determined through compressive tests on masonry and mortar samples and
direct shear tests on masonry triplets (European standard EN 1052-3). The
numerical analysis consisted on the simulation of diagonal compression tests on
masonry panels made in laboratory according to RILEM recommendations. The
model was able to estimate, with good accuracy, the ultimate load and strain, the
plastic strain evolution and the failure mode.
Regardless of the more or less success of the different modelling approaches to
simulate the in-plane behaviour of masonry, the actual investigations are still
looking for a more accurate and generalized model to simulate the different
aspects that can influence the masonry mechanical behaviour.

6
3. Work Methodology

An integrated method was used to predict the in-plane shear behaviour of a


new masonry system that has recently been developed through a national co-
funded research project in Portugal. In particular it aims analysing the influence of
the geometrical and mechanical properties of the constituent materials in the
behaviour of the masonry. The method involves a FE numerical approach based
on a detailed 3D modelling of diagonal compressive tests (standard tests) on
masonry samples, using nonlinear behaviour models that are calibrated through a
small number of laboratory tests, reducing time consuming and financial cost. The
method consists of the following steps:
• characterization of the mechanical properties of the masonry constituents,
units and mortar, through laboratory tests and available literature;
• characterization of the shear behaviour of masonry through laboratory tests
performed on small samples according to American standard ASTM E519
(2002).
• construction of 3D models of masonry samples using a FEM program and
a plastic damage type model to simulate the units and mortar behaviour;
• calibration and validation of the numerical model through experimental
results obtained from laboratory tests;
• characterization of the shear behaviour of the masonry samples through
computer-simulated shear tests, considering different geometrical and
mechanical properties for the masonry constituents.
During the calibration and validation procedures, potential numerical fracture
patterns were determined and compared to the fracture patterns observed in the
laboratory tests.

4 Masonry System - General Characteristics

The masonry system used as case study was developed under a co-funded
national research project, called “Thermal and mechanical optimization of single
leaf masonry (OTMAPS)”, which aimed conceiving a single leaf masonry system
with optimal thermal performance and enough mechanical strength to be used as
structural masonry.

7
This project involved numerical and experimental studies and some results
have already been published (Sousa et al. 2011; Sousa and Sousa 2011; Sousa and
Sousa and Sousa 2010).
The masonry system was developed for single leaf walls with units made of
lightweight concrete with light expanded clay aggregates (Leca®). The units
measure 350x350x190mm3 and have a blind surface for laying mortar (Figs 1a
and 1b). Factory-made lightweight mortar is used in all masonry joints.

Fig.1 Base unit: (a) scheme with unit dimensions (in mm), (b) blind surface for laying mortar
and (c) opposite side.

The bed joints consist on two 10mm thick mortar strips (shell bedded
masonry), 120mm wide each, positioned along the sides of the units. The perpend
joints are represented by three hollow columns defined by the irregular lateral
surfaces of the units in contact, filled in with mortar, forming 40% of the width of
the unit, (Fig.2).

Fig. 2 Masonry assembly details

The shape and the material of the units and, in particular, the dimensions and
distribution of the cavities was decided within the research project to full fill
thermal requirements, but also handling capabilities.
The main properties of the masonry materials (concrete units and mortar joints)
are summarized in Table 1. These results were determined through standard
laboratory tests in the scope of the referred research project (Sousa and Sousa
2010).

Table 1 Properties of the masonry materials (unit and mortar joints)

8
5. Determination of the Masonry Properties

5.1 Materials (units and mortar)

A set of laboratory tests was performed to determine the necessary parameters


to define the constitutive laws of the masonry materials, namely the uniaxial
stress-strain behaviour curves of the units and the joints mortar. These parameters
are essential, not only to characterize the masonry, but also to give input data to
the numerical model.
Given the difficulties associated with the dimensional representation of test
samples (e.g. the concrete samples were extracted from the units webs and shells,
meaning small samples) and the carrying out of the tests (e.g. difficulties to adjust
the loading rate and the displacement measuring capabilities of the test device to
the brittle nature of the masonry constituents), as imposed by the standards, the
properties of the units concrete and the joints mortar were estimated through
laboratory tests done in this study, but also by using expressions in
regulations/codes for concrete structures, i.e. Model Code 90 (CEB 1990), and
experimental data from other studies (Kupfer et al. 1969; Neville 1995; Veiga
1997).
a) Mortar
To characterize the tensile behaviour of the mortar, the direct tensile strength
(σto) and the fracture energy (GF) were estimated from the mathematical
expressions of Model Code 90:

(1)

(2)

and by using the mean flexural and compressive strength obtained from the
flexural and compressive tests performed in this study on the mortar (following
the European standard EN 1015-11 1999): a tensile flexural strength (fl) of
2.45N/mm2 and a compressive strength (σcu) of 11.45N/mm2.
The base values of fracture energy (GFo) are defined in MC90 as a function of
the maximum aggregate size (dmax). A regression technique was used to obtain
GFo as function of dmax.
9
The compressive constitutive law of the mortar (before and after the
compressive peak stress) was also estimated from the Model Code 90
mathematical expressions:
• for compressive strains (εc) lower or equal to than εc,limit;

(3)

• for compressive strains (εc) higher than εc,limit;

(4)

The limit compressive strain (εc,limit) has no significance other than limiting the
applicability of expression (3). The expressions that define the parameter ϕ, as
well as εc,limit, can be found in Model Code 90.
In this case, the mean compressive strength (σcu), the ultimate strain (εcu) and
the modulus of elasticity (E0) had to be previously determined or estimated. The
compressive strength was obtained from the compressive tests on the mortar
samples. The modulus of elasticity was estimated from test data reported in other
experimental studies carried out on similar mortars (Veiga 1997), i.e. the
experimental results of the compressive strength and the modulus of elasticity of
several cement/sand based mortars (12 types of general purpose mortars) were
used to obtain a probable function to calculate E0 (in N/mm2), through a
regression technique:

(5)

Finally, the default value indicated in Model Code 90 for the εcu was adopted
(around of -2x10-3mm/mm).
b) Units
To characterize the tensile behaviour of the lightweight concrete of the units,
the direct tensile strength (σto) and the fracture energy (GF), the expressions (1)
and (2) were also used.

10
However, the average values of the concrete properties needed in the two
previous expressions were obtained from laboratory tests performed in the units,
namely from flexural tests on samples extracted directly from the units webs and
shells (giving fl = 2.0N/mm2), Fig. 3, and from compressive tests performed in the
complete units (giving E0 and σcu, as it will be referred to afterwards).

Fig.3 Examples of samples prepared from webs and shells removed from the units, and the
flexural test set-up

This procedure allowed using more reliable concrete samples and, therefore,
getting more consistent data to include in the numerical model, since the use of
larger samples (more in agreement to the standards) wouldn’t be able to represent,
in such a realistic way, the very particular physical rearrangement of the
constituents of the concrete as they exist within the units. Moreover, the use of
test samples from masonry units takes also into account the influence of the
manufacturing process (vibro-compression moulding) on the mechanical
properties of the concrete. Taking this into consideration, the uniaxial
compressive behaviour of the concrete, in particular the compressive behaviour
before the peak stress (or compressive strength, σcu), was also measured directly
from the units, namely trough laboratory tests performed in two loading directions
(perpendicular and parallel to the bed joints), according to the procedures set out
in standard EN 772-1 (2002). Six units were tested, three for each loading
direction. Two displacement transducers were set on the units to measure the
vertical shortening and the horizontal extension (∆v and ∆h) (Fig. 4).

Fig. 4 Simple compression tests on masonry units: (a) perpend and (b) parallel to the bed joints

The results of these tests enabled to determine the constitutive compression


laws of the units in both loading directions, up to the peak stress. The stresses
were calculated for the net area of the unit’s (surface perpendicular to the loading
direction). The similarity between the results for the two loading directions
allowed estimating, using a regression technique, a single polynomial function to
represent the units compressive law on both directions. This function, that
presents a high correlation to the experimental results (R2 = 0.96), was adopted in
11
the numerical model to simulate the compressive behaviour law before the
compressive strength, σcu (fig.5).

Fig. 5 Compressive behaviour of the concrete used in the masonry units (up to σcu)

To estimate the behaviour of the concrete after the peak stress, the
mathematical expressions (3) and (4) were also used. In these expressions the
properties of the concrete obtained from the compressive behaviour law were
inserted (modulus of elasticity, E0, compressive strength, σcu, and ultimate strain,
εcu). Other experimental data on lightweight concrete available in the literature
were also considered (Neville 1995), in particular on the validation of the
geometrical shape of the compressive stress - strain diagrams after the peak stress
(a constant residual stress for high values of strain is expected for this kind of
concrete).
Notice that the need to characterize the compressive behaviour of both the
concrete and the mortar after the peak stress, i.e. the descending branch of the
diagram for high values of strain, is to prevent convergence problems during the
non-linear numerical simulations.
Figure 6 represents the constitutive compressive laws estimated for the mortar
and the concrete, and Table 2 gives the results estimated for the mechanical
characteristics.

Fig. 6 Constitutive compressive laws estimated for the mortar and concrete

Table 2 Estimated mechanical characteristics of the concrete and mortar

It is stressed that the constitutive law of the concrete used in the units represent
a simplified and integrated way of considering the influence of various factors,
apart from the material itself, on the behaviour and fracture mechanism of the
units. Among those factors, some are particularly complex and difficult to
implement in standard non-linear analysis, such as the manufacturing process,
already referred to, and the potential geometrical instability of the webs and shells
within the units.

12
5.2 Masonry system

Diagonal tension (shear) tests were carried out according to ASTM E519
(2002). This test was chosen because of its versatility on the application to
different types of masonry and of its capability to provide representative values of
the shear mechanical parameters.
The data obtained through these experiments are essential to understand the
mechanical behaviour of the masonry system. This campaign complements the
information got from the previous tests and allows calibrating the masonry as a
whole. Three types of masonry assemblies were built in laboratory conditions,
thus:
• 4 samples with filled perpend joints – assembly A (reference);
• 4 samples with unfilled perpend joints – assembly B;
• 4 samples built with surface finishing made of a 20mm thick layer of
unreinforced rendering mortar (same mortar as used in the joints) and with
filled perpend joints – assembly C.
The dimensions of the masonry samples were adjusted to the laboratory
equipment: 800x800x350mm3 (length x height x width).
The vertical shortening (∆v) and the horizontal extensions (∆h) were measured
by four displacement transducers, two for each side of the samples. A
compression load (F) was applied at a constant speed (0.1N/mm2/minute) to
achieve the maximum strength in an interval of 15 to 20 minutes (Fig.7 and
Fig.8). All measurements were taken up to the maximum load (Fmax).

Fig. 7 Illustration of a diagonal tension (shear) sample used in laboratory tests (dimensions in
mm)

Fig. 8 Diagonal tension test set-up in the laboratory: (a) A and (b) C assemblies

The shear stress and strain diagrams (τ - γ) obtained from numerical regression
techniques applied to the experimental results of each of the three tested
assemblies are represented in Figure 9. The coefficients of determination (R2)
ranged between 0.95 and 0.98.
13
Fig.9 Shear behaviour of the 3 masonry assemblies tested in laboratory (regression analysis
curves)

The main shear characteristics determined through these regression curves are:
• the maximum compression load (Fmax);
• the maximum shear stress (τmax);
• the shear strain for the maximum shear stress (γmax);
• the shear modulus (G).
These characteristics were calculated by means of the following expressions
defined in the American standard ASTM E519 (2002):

(6)

(7)

(8)

(9)

The shear modulus (G) was calculated for strains corresponding to 1/3 of the
maximum shear stress. The main shear characteristics of the masonry assemblies
are given in Table 3.

Table 3 Shear characteristics of the masonry assemblies

In general, the fracture pattern consisted of cracks opening in a direction more


or less parallel to the applied compression load (F). This pattern was particularly
evident in assembly “C” (with mortar rendering - Fig. 10) where the main fracture
line follows an almost straight diagonal line from the loading corners. As for
assemblies A and B, although the fracture pattern tends to follow the same
diagonal, the cracks tag along the joints, near the joint-unit interfaces, contouring
14
the units (Fig.11-12). This behaviour was more evident in the masonry samples
with unfilled perpend joints (assembly B), which exhibited a more brittle
behaviour when compared to the other assemblies (Fig.11).
Local damage in the units near the loading point was observed in all the
assemblies.

Fig. 10 Example of the fracture pattern observed in masonry assembly C (with mortar rendering)

Fig. 11 Example of the fracture pattern in assembly B (unfilled perpend joints)

Fig. 12 Example of the fracture pattern in assembly A (filled perpend joints)

The results from this set of tests, together with the results referred in section
5.1, gave the necessary information to verify and calibrate the numerical models,
providing the necessary confidence to perform the other experimental tests on a
numerical, but reliable basis. More detailed information and conclusions about
this experimental work can be found in (Sousa and Sousa 2011).

6 Numerical Simulations

6.1 Implemented model – General description

The numerical simulations were performed using a micro 3D FE modelling of


masonry samples. In particular, a non-linear constitutive model was used for the
mortar and the concrete, as it will be explained in 6.3. For simplification purposes,
no friction models were adopted at the unit-joint interfaces (e.g., Mohr-Coulomb),
but rigid links instead (Fig.13), as it will be explained in the next section.

Fig.13 Behaviour adopted for the masonry materials and unit-joint interfaces (vertical cut)

15
6.2 Modelling the unit-joint interfaces

The links between the units and the mortar joints were considered rigid, i.e. a full
transmission of tangential and vertical displacements was adopted between the
two elements contact interfaces to overcome convergence problems detected when
using friction models (e.g., Mohr-Coulomb) during non-linear regimes. Although
being a simplification, this was sustained by observations and measurements
obtained from the laboratory tests. During the application of loading, no evident
fracture pattern was observed on the interfaces. Only when the maximum load
was reached the fracture pattern was clearly visible and the maximum
displacement measured parallel to the bed joints was lower than 0,5mm.
In the case of the perpend joints, it was not considered any head-to-head
contact between units. It’s highlighted that the perpend joints have gaps of about 3
to 5mm (see Fig.8a and Fig.12), which are unlikely to be filled by the unit’s
deformability, at least until the maximum load is reached, as it was observed
during the laboratory tests in the masonry assemblies A (reference sample) and B
(sample with unfilled perpend joints). Therefore, for the perpend joints no contact
or friction models were considered in assembly B and a perfect local contact was
introduced in the interfaces between the concrete units and the mortar joints in the
case of assemblies A and C (Fig.14).

Fig.14 Contact model used for the perpend joints for assemblies A and C (horizontal cut)

6.3 Modelling the masonry materials

The tensile and compressive behaviour of the concrete and mortar were
simulated with a plastic damage model developed by Lubliner et al. (1989), later
improved by Lee and Fenves (1998). A full description of the development and
principles of this constitutive model can be found in the relevant references.
However, to contextualize and understand the work carried out in this study, the
main concepts used in this model are briefly described.
This model simulates the non-linear behaviour of brittle materials (e.g.,
concrete, mortar, natural rocks) subjected to monotonic or cyclic loading. For the
concrete and the mortar the main nonlinearities are on the tensile and compression
behaviour curves, which may induce cracking and crushing failures mechanisms.
16
These nonlinearities are simulated in the model through uniaxial constitutive laws,
usually represented by stress-strain behaviour curves in tensile (σt - εt) or
compression (σc - εc) determined experimentally (Fig.15).

Fig. 15 Typical stress-strain diagrams for concrete (uniaxial constitutive laws): (a) tensile and (b)
compressive curves

The uniaxial tensile constitutive laws use the concept of fracture energy (GF),
proposed by Hilleborg (1976). The values of fracture energy, tensions and
extensions can be determined experimentally or, in the case of materials such as
concrete or mortar, estimated from specific literature (e.g. Model Code 90). The
fracture energy can be provided directly to the model and, if properly used, it may
help solving convergence problems in the post-peak tensile regime, in particular
when dealing with concrete elements with little or no reinforcement at all.
This model uses the concepts of classical plastic theory, in particular strain
decomposition, elasticity, and plastic flow.
a) Strain variables
Strain is decomposed into elastic and plastic deformations:

(10)

b) Hardening variables
The failure mechanisms are tensile cracking and compressive crushing that are
characterized by two hardening variables:
–Tensile equivalent plastic strain

– Compressive equivalent plastic strain

The evolution of the yield (or failure) surface is controlled by these hardening
variables.
c) Stress-strain relations
The model considers the reduction of material stiffness through a unique and
isotropic scalar degradation variable (d). This scalar variable is governed by the
hardening variables and the Cauchy stresses in the following generic form:

17
(11)

(12)

If no damage variables are introduced, the model assumes a plastic behaviour.


The plastic formulation is performed in terms of effective stress, calculated in the
following generic form:

(13)

d) Plastic flow
The development of the plastic strains is governed by a potential plastic flow
(Gp), established according to following law:

(14)

The flow potential used is the Drucker-Prager hyperbolic function:

(15)

e) Yield function
The yield surface defined as a function of the hardening variables and the
effective stresses in the following form:

(16)

(17)

The adimensional parameters are related to the material properties, thus


allowing the calibration of the constitutive model, and are calculated in the
following generic form:

(18)

18
on the condition 0 ≤ αi≤ 0,5 (19)

on the condition 0,5 ≤ Kc ≤ 1 (20)

Some studies indicate that this constitutive model is quite capable of simulating
the facture mechanisms in simple and in reinforced concrete structures
(Jankowiak and Lodygowski 2005).

6.4 Implementation of the numerical model

The constitutive model presented in 6.3 is implemented in the finite element


software that was used in this study. The mechanical behaviour of the masonry
was simulated by micro three-dimensional modelling of the masonry samples
tested in the laboratory (Fig.16-17).

Fig. 16 Geometric model used for masonry test sample with and without filled perpend joints
(assemblies A and B)

Fig. 17 Geometric model used for masonry test sample with rendering mortar (assembly C)

Given the complex shape of the masonry units, 4-node tetrahedral finite
elements were used to facilitate the meshing process. The masonry samples were
simulated with total restraint of displacements at the loading corners, therefore
with a higher local confinement effect when compared to the physical samples.
The Newton-Raphson iterative base method was chosen, although the program
can automatically select other methods, depending on the convergence process.
The loading was simulated by imposing, in one of the supporting corners, a
linear smooth variation of displacements in the direction of the compression load
applied to the samples tested in the laboratory. The corresponding compression
load (F) was determined, afterwards, through the sum of the nodal reactions.
The results of the calculations, namely the compression load and the shear
strain diagrams (F - γ), were used to determine the mechanical characteristics of
the masonry through the expressions 1 to 4.

19
7 Calibration and Validation of the Numerical Model

7.1 General aspects

The calibration and validation of the numerical model involved the simulation
of the experimental tests performed on the units and on the masonry (referred in
section 5) to verify the capability of the model to properly represent the masonry
system tested in laboratory conditions. This process involves the comparison
between the test results determined experimentally in the laboratory and the
results obtained in the numerical simulations, in particular the comparison
between the response curves and the mechanical properties of masonry and units.
Moreover, this procedure was split in two phases. The first consisted on
simulating the compression tests performed in laboratory on the concrete units,
and the second the diagonal shear tests on the masonry assemblies. These
simulations allowed verifying, first the units behaviour, i.e. the concrete behaviour
law alone when integrated in the units, and then the masonry system as an
assembly of units, joints and interfaces.

7.2 Calibration of the units’ constitutive model - Phase 1

The calibration of the units compressive behaviour was made by simulating the
compressive tests performed in laboratory. A micro 3D FE modelling of a unit
was constructed and displacements were imposed to simulate the loading. The
other possible displacements of the loading surfaces nodes were restrained; thus a
higher confinement of the unit was expected in the numerical simulations. The
verification and calibration was then conducted with a sensitivity analysis
performed through the variation of some of the parameters of the constitutive
model. The parameters related to the material properties that were obtained
directly from the experimental tests, or indirectly by using experimental values
from the experimental campaign and the expressions from the literature, namely:
the elastic modulus (E0), the Poisson's ratio (ν), the uniaxial tensile and
compressive strength (σto and σcu), the ultimate compressive strain (εcu) and the
fracture energy (GF), were set constant and equal to the values established in
20
Table 2. As for the other model parameters, a range of values was selected
according to following limits:
• ∆ = [15º; 45º];
• ∆ = [0.1; 1];
• ∆Kc = [2/3; 1];
• ∆ = [1; 1.3].

During the simulations, only one parameter was changed at a time, keeping
constant the others parameters for a set of predefined values.
Through direct comparison to the experimental results, it was obtained the best
possible approximation between the numerical and the experimental compressive
stress and strain diagrams. In this case, the best fit gave a small difference,
between 2% and 3%, for εcu and σcu (Fig. 18). Moreover, it was found that Kc and
ψ were the most influential variable parameters. Table 4 presents the values
adopted for the model parameters after the calibration of the units.

Fig.18 Comparison between the experimental and the numerical response curves after calibration
(concrete units in simple compression)

Table 4 Variable model parameters adopted for the concrete units after calibration (best fit)

7.3 Validation of the masonry behaviour – Phase 2

After calibrating the constitutive model of the units, it were performed the
numerical simulations of the shear diagonal tests done in the laboratory. The three
types of masonry assemblies (A, B and C) were simulated, separately, using the
model in section 6.1 and the material properties listed in Table 2 and in Table 4.
This validation was done by comparing the experimental to the numerical
results for each type of masonry assembly, in particular the compression load and
the shear versus strain diagrams (F-γ), and the masonry mechanical characteristics
related to the shear behaviour (Fmax, γ and G).
The first calculations gave good results when compared to the experimental
results, in particular a good estimation of Fmax and γmax, with variations below 9%.
The major difference was found in the shear modulus with variations between
21
10% and 23%, especially in the samples with unfilled joints (assembly B). In spite
of this, it was decided not to make any changes on the model parameters, since the
masonry assemblies B were also those that presented the highest dispersion in the
experimental results (R2 = 0.95) and, thus, the numerical results were considered
to be within an acceptable range of values.
The experimental and numerical diagrams (F-γ) comparing the responses for
each assembly type are presented in Fig. 19. The mechanical characteristics and
the comparative ratios obtained are shown in Table 5. These ratios have the
experimental results as common denominator.

Fig.19 Comparative diagrams of the experimental and numerical results obtained for the three
masonry assemblies after calibrating the units behaviour laws

Table 5 Mechanical properties and ratios obtained after the calibration process – numerical (num.)
versus experimental (exp.)

A detailed analysis of the stresses and the deformations, in particular of the


principal stresses distribution, was undertaken to identify potential
fracture/cracking patterns and, therefore, to further validate the numerical model.
It is assumed that the fracture patterns, or crack directions coincide with the
alignments of the minimum principal stresses, which can then be compared to the
patterns observed in the laboratory tests.
The minimum principal stresses distribution on the assemblies and the
corresponding global deformation (magnified about 250 times) are represented in
Figures 20 to 22 for the maximum compression load (breaking point).

Fig. 20 Minimum principal stresses and global deformation at breaking point – Assembly A
(filled perpend joints)

Fig. 21 Minimum principal stresses and global deformations at breaking point – Assembly B
(unfilled perpend joints)

Fig. 22 Minimum principal stresses and global deformation at breaking point – Assembly C
(rendering mortar and filled perpend joints)

22
The main aspects to be noted in the results of the numerical tests on the
masonry assemblies that are consistent to the experimental results from the
laboratory tests are:
• the compressive stresses are concentrated in an area more or less
coincident with the direction of the compression load (except for assembly
C, the higher stress concentration areas define a “zig-zag” line around the
mortar joints);
• there is a higher concentration of stresses along the bed joints on areas
closer to the perpend joints, especially when these are unfilled (assembly
B);
• there is a high concentration of stresses and deformation in the units
located near the load application points (corners of the masonry samples).

8 Masonry Tests Using Numerical Analysis

8.1 General Aspects

Starting from the calibrated and validated model, it is now possible to perform
other tests on the same masonry assemblies, or on other samples using the same
masonry, but different dimensions or shapes, with low time consuming and
financial cost. In this particular case, the same masonry assemblies were tested
taking into account the influence of the mechanical properties of the materials and
of the geometrical shape of the joints on the mechanical behaviour of the masonry
system. The results were compared to the reference assembly (A) properties listed
in Table 5. The parameters involved in this sensitivity analysis were:
• the compressive strength of the mortar joints; it was adopted the same
initial thickness (10mm), but different compressive strengths (0.5xσcu and
2.0xσcu);
• the compressive strength of the rendering mortar used in assembly C; it
was still considered unreinforced mortar with the same thickness (20mm),
but with a higher compressive strength (2.0xσcu);

23
• the thickness of the parallel joints (e); it was adopted a mortar with the
characteristics of the reference situation (assembly A), but with different
thicknesses (0.5xe and 2.0xe);
• the geometrical configuration of the parallel joints; it was adopted a mortar
with the characteristics of the reference situation (assembly A), but with
discontinuous joints (g/t = 0.685, where "g" is the full width of the mortar
strips and "t" the total thickness of the wall) were replaced by full joints
(g/t = 1.0).
In order to consistently consider the variation of the mortar compressive
strength (σcu) on the global material characteristics, the mechanical properties (E0
and GF) and constitutive laws of the mortar considered in the simulations with half
or twice the mortar compressive strength (0.5xσcu and 2xσcu) were estimated as
described in section 5.1 (results are presented in Table 6). In the case of the
uniaxial tensile strength (σt0), the value was estimated through the compressive
strength (0.5xσcu and 2xσcu) and using the Model Code 90 expression:

(21)

Table 6 Mechanical properties estimated for the mortar used in the sensitivity analysis

In the analysis, only one parameter was changed at a time, being the other
parameters constant and equal to those of the reference sample (assembly A).

8.2 Results

The mechanical behaviour of the new masonry samples with updated


parameters (var.) was compared to that of the masonry assembly A (ref.) through
the final diagrams and the ratios between the mechanical properties determined in
both numerical simulations (var./ref.), namely the maximum compression load
(Fmax) and the corresponding shear strain (γmax) and shear modulus (G) (Table 7).
Figures 23 to 26 show the set of final diagrams (F-γ) that correspond to the
considered variations in relation to the reference sample.

24
Fig. 23 Behaviour of the masonry samples with joints made of mortars with different mechanical
strength

Fig. 24 Behaviour of the masonry samples with rendering mortars with different mechanical
strength

Fig. 25 Behaviour of the masonry samples with continuous (g/t = 1) and discontinuous parallel
mortar joints (g/t = 0.69)

Fig. 26 Behaviour of the masonry samples with joints of different thicknesses (e)

Table 7 Mechanical properties and ratios for the masonry system determined in the sensitivity
analysis

The results show that the use of higher strength rendering mortar and
continuous parallel joints are the factors that mostly improve the shear behaviour
of masonry (in plan), namely its strength and stiffness. These findings seem
consistent with known experimental studies and code specifications. Oliveira and
Hanai (2005) reported an increase of about 55% in the shear strength (diagonal
compression) of the masonry when using unreinforced mortar (with compressive
strength of 4N/mm2) and an increase of about 30% when using higher
compressive strength mortar (23N/mm2). Moreover, although reinforced mortar
was used in the experiments of Hamid et al. (1994) and Drysdale et al. (1999),
these authors refer an overall increase of the shear strength (diagonal
compression) of about 200% and of the stiffness of about 75% in clay masonry
samples when using rendering mortar with higher strength capability.
In the case of continuous joints, the increase of the shear capacity of masonry
is explained by the increase of the bond area between the units and the mortar
joints. This increase in strength is evident in the formulae of Eurocode 6 (EN
1996-1-1 2005), where the shear strength of the masonry depends on the bond
strength (designated as the initial shear strength) and on the coefficient g/t.
On the other hand, increasing the thickness of the parallel joints and (or) the
mechanical strength of the mortar joints have no significant influence on the shear
behaviour of this masonry system. No experimental studies about these aspects in

25
the shear behaviour were accessed or found yet by the authors in order to better
sustain these numerical findings.

9 Conclusions

The present paper describes the study of the shear behaviour of a masonry system
by using an approach that involves the merging of experimental and numerical
results. In particular, it describes the steps followed in the analysis of the shear
behaviour of a new masonry system made of lightweight concrete units that were
recently developed in Portugal. The shear behaviour was studied by loading
masonry test assemblies in compression along their diagonal, as defined in the
American standard test ASTM E519 (2002). A limited number of experimental
tests were performed to produce experimental data to verify and calibrate a
numerical model that was used to simulate a series of new conditions and
possibilities for the new masonry system.
The following conclusions can be drawn from this study:
• it was possible to characterize the mechanical properties of the masonry
using, not only experimental results from standard tests on the masonry
system, but also simplified tests and other data from technical literature,
standards and codes available for the materials;
• despite the simplifications made to model the interfaces between units and
joints, which were considered rigid, the adopted numerical model
characterized fairly well the behaviour of the masonry system in shear
(through diagonal compression) and consistently represented the main
aspects of the failure mechanism observed in the laboratory tests, at least
until the maximum load (breaking point);
• however, the simplification made on modelling the interfaces may result in
important deviations to the experimental results in the post-peak shear
behaviour of the masonry;
• this last aspect can be clarified later by performing laboratory tests after
the breaking point and by using friction models in the joints, although it is
likely that the use of these models can significantly increase the time and
computational resources required for numerical convergence.

26
The results of this study show that is possible to accurately characterize the
shear behaviour of masonry through computer simulations of laboratory tests, in
particular the American standard test defined in ASTM E519 (2002), by using
simple models that can be calibrated using a small number of laboratory tests,
either on the masonry units and materials, or on masonry samples. This practical
study has shown the potential of using computer simulations of laboratory tests,
reducing the need for the involvement of important laboratory resources and costs
in the research and development of masonry systems.

Acknowledgments

The authors gratefully acknowledged ADI- Portuguese Innovation Agency and the company
Maxit-Portugal for the help provided in the OTMAPS research project.

References
Gabor A, Ferrier E, Jacquelin E, Hamelin P (2006) Analysis and modelling of the in-plane shear
behaviour of hollow brick masonry panels. Constr Build Mater 20:308–321.
doi:10.1016/j.conbuildmat.2005.01.032
ASTM E 519-02 (2002) Standard test method for diagonal tension (shear) in masonry
assemblages. American Society for Testing and Materials, West Conshohocken-Pennsylvania,
USA
CEB (1990) Model Code 1990: Design Code. Thomas Telford, Lausanne, Swiss
Vyas ChVU, Reddy BVV (2010) Prediction of solid block masonry prism compressive strength
using FE model. Mater Struct 43:719–735. doi:10.1617/s11527-009-9524-9
Calderini C, Cattari S, Lagomarsino S (2010) The use of the diagonal compression test to identify
the shear mechanical parameters of masonry. Constr Build Mater 24:677–685.
doi:10.1016/j.conbuildmat.2009.11.001
Barbosa CS, Lourenço PB, Hanai JB (2010) On the compressive strength prediction for concrete
masonry prisms. Mater Struct 43:331–344. doi:10.1617/s11527-009-9492-0
Drysdale RG, Hamid AA, Baker LR (1999) Masonry structures. Behaviour and design, 2nd Edn.
The Masonry Society, Boulder, USA
Hendry EAW (2001) Masonry walls: materials and construction. Constr Build Mater 15:323-330
EN 1015-11 (1999) Methods of test for mortar masonry - Part 11: determination of flexural and
compressive strength of hardened mortar. European Committee for Standardization, Brussels,
Belgium
EN 1996-1-1 (2005) Eurocode 6 - Design of masonry structures. Part 1-1: general rules for
reinforced and unreinforced masonry structures. European Committee for Standardization,
Brussels, Belgium

27
EN 772-1 (2000) Methods of test for masonry units. Part 1: determination of compressive strength.
European Committee for Standardization, Brussels, Belgium
Mosele F, da Porto F (2011) Innovative Clay Unit Reinforced Masonry System: Testing, Design
and Applications in Europe. Procedia Eng 14:2109-2116. doi:10.1016/j.proeng.2011.07.265
Faria R, Oliver J, Cervera M (1998) A strain-based viscous-plastic-damage model for massive
concrete structures. Int j solids struct 35 (14):1533-1558
Fouchal F, Lebon F, Titeux I (2009) Contribution to the modelling of interfaces in masonry
construction. Constr Build Mater 23:2428-2441. doi:10.1016/j.conbuildmat.2008.10.011
da Porto F, Mosele F, Modena C (2011) Compressive behaviour of a new reinforced masonry
system. Mater Struct 44:565-581. doi:10.1617/s11527-010-9649-x
da Porto F, Mosele F, Modena C (2011) In-plane cyclic behaviour of a new reinforced masonry
system: Experimental results. Eng Struct 33:2584-2596. doi:10.1016/j.engstruct.2011.05.003
Uva G, Salerno G (2006) Towards a multiscale analysis of periodic masonry brickwork: A FEM
algorithm with damage and friction. Int J Solid Struct 43:3739-3769.
doi:10.1016/j.ijsolstr.2005.10.004
Hamid A, Mahmoud A, Sherif E (1994) Strengthening and repair of masonry structures: State of
the art. In: Proceedings of the 10th International brick and block masonry conference, Calgary,
Canada, pp 485-494
Hilleborg A, Modeer M, Petersson P (1976) Analysis of Crack Formation and Crack Growth in
Concrete by Means of Fracture Mechanics and Finite Elements. Cem Concr Res 6:773-782
Jankowiak T, Lodygowski T (2005) Identification of parameters of concrete damage plasticity
constitutive model. Found Civil Environ Eng 6:53-69. ISSN: 1642-9303
Díaz JJ, Nieto PJ, Rabanal FP, Martínez-Luengas AL (2011) Design and shape optimization of a
new type of hollow concrete masonry block using the finite element method. Eng Struct 33:1-9.
doi:10.1016/j.engstruct.2010.09.012
Lemos JV (2007) Discrete element modelling of masonry structures. Int J Archit Herit 1(2):190–
213. doi:10.1080/15583050601176868
Chaimoon K, Attard M (2007) Modeling of unreinforced masonry walls under shear and
compression. Eng Struct 29(3):2056–2068. doi:10.1016/j.engstruct.2006.10.019
Kupfer H, Hilsdorf HK, Rusch H (1969) Behaviour of concrete under bi-axial stresses. ACI J Proc
66(8):656-666
Lee J, Fenves G (1998) Plastic-damage model for cyclic loading of concrete structures. J Eng
Mech 124(8):892-900
Lourenço P, Rots J (1997) Multisurface interface model for analysis of masonry structures. J Eng
Mech 123(7):660-668
Lourenço PB, Vasconcelos G, Medeiros P, Gouveia J (2010) Vertically perforated clay brick
masonry for loadbearing and non-loadbearing masonry walls. Constr Build Mater 24:2317-2330.
doi:10.1016/j.conbuildmat.2010.04.010
Lubliner J, Oliver J, Oller S, Oñate E (1989) A plastic-damage model for concrete. Int J Solid
Struct 25(3):299-326

28
Pelà L, Cervera M, Roca P (2011) Continuum damage model for orthotropic materials:
Application to masonry. Comput Methods Appl Mech Eng 200:917-930.
doi:10.1016/j.cma.2010.11.010
Berto L, Saetta A, Scotta R, Vitaliani R (2004) Shear behaviour of masonry panel: parametric FE
analyses. Int J Solids Struct 41:4383-4405. doi:10.1016/j.ijsolstr.2004.02.046
Mistler M, Anthoine A, Butenweg C (2007) In-plane and out-of-plane homogenisation of
masonry. Comput Struct 85:1321-1330. doi:10.1016/j.compstruc.2006.08.087
Neville A (1995) Properties of concrete, 4th edn. Longman Scientific and Technical, Harlow
Oliveira FL, Hanai JB (2005) Reabilitação de paredes de alvenaria pela aplicação de revestimentos
resistentes de argamassa armada (in portuguese). Cad Eng Estrut 7(26):131-164. ISSN 1809-5860
Quinteros RD, Oller S, Nallim LG (2012) Nonlinear homogenization techniques to solve masonry
structures problems. Compos Struct 94(2):724-730. doi:10.1016/j.compstruct.2011.09.006
Brasile S, Casciaro R, Formica G (2010) Finite Element formulation for nonlinear analysis of
masonry walls. Comput Struct 88:135-143. doi:10.1016/j.compstruc.2009.08.006
Sousa LC, Castro CF, António CC, Sousa H (2011) Topology optimisation of masonry units from
the thermal point of view using a genetic algorithm. Constr Build Mater 25(5):2254-2262.
doi:10.1016/j.conbuildmat.2010.11.010
Sousa R, Sousa H (2011) Influence of head joints and unreinforced rendering on shear behaviour
of lightweight concrete masonry. In: Proceedings of the 9th Australasian Masonry Conference,
Queenstown, New Zealand, pp 515-522
Sousa R, Sousa H (2010) Experimental evaluation of some mechanical properties of large
lightweight concrete and clay masonry and comparison with EC6 expressions. In: Proceedings of
the 8th International masonry conference, Dresden, Germany, pp 545-554
Veiga M (1997) Comportamento de argamassas de revestimento de paredes: Contribuição para o
estudo da sua resistência à fendilhação (in portuguese). PhD Dissertation, University of Porto
Tao X, Phillips DV (2005) A simplified isotropic damage model for concrete under bi-axial stress
states. Cem Concr Compos 27:716-726. doi:10.1016/j.cemconcomp.2004.09.017

29

View publication stats

You might also like