Professional Documents
Culture Documents
www.elsevier.com/locate/engstruct
Received 21 March 2007; received in revised form 3 October 2007; accepted 3 December 2007
Available online 22 January 2008
Abstract
The aim of this paper is the development of suitable mechanical models able to reproduce and, hence, to predict the response of masonry
structures. In particular, the study is addressed to the modelling of strengthened masonry structures in which the introduction of new types
of reinforcement, particularly the FRP (fibre-reinforced plastic) materials, strongly affects the structural response through complex interaction
mechanisms between masonry and strengthening elements.
In this paper two different approaches able to model the behaviour of un-reinforced and reinforced masonry structures are proposed. The first
one is based on a micro-mechanical and multiscale analysis combined with the use of the kinematic and static theorems of the limit analysis;
in this case a rigid-perfectly plastic constitutive relationship is considered for the masonry material and for the FRP–masonry interaction. The
second approach is based on the use of macroscopic models. In this case, the constitutive relationship for the masonry material accounts for the
softening effect throughout the use of a smeared crack approach. Moreover, different modelling strategies and constitutive laws are adopted for
the FRP-reinforcement addressing particular regard to the delamination phenomenon.
Numerical computations are developed for un-strengthened and FRP-strengthened masonry panels. The obtained results, in terms of the global
response of the examined panels, are compared with the data available from experimental tests and interesting aspects are remarked.
c 2007 Elsevier Ltd. All rights reserved.
Fig. 1. Model adopted for the homogenized limit analysis approach. (a) Heterogeneous model; (b) un-reinforced masonry homogenization; (c) macroscopic
approach (masonry homogenized + FRP).
always easy, particularly when new structural features must be detailed description of the examined cases, particular regard
simulated. In fact, in these cases, it is important to exploit the has been addressed both to the phase of parameter setting and
capabilities of the codes developing suitable strategies based on to the comparison between experimental and analytical results.
the combination of special elements available in the FE codes. Finally in the Section 5 of the paper, concluding remarks are
In this paper, the commercial nonlinear FE code DIANA reported, reviewing the proposed models features in the light of
9.1 [13] has been used; different numerical models obtained the obtained results.
from the combination of constitutive laws and special features
implemented in DIANA are adopted to simulate the behaviour 2. Limit analysis with homogenization model
of the reinforced masonry, with a critical discussion of the use
of the available elements. In this section, a homogenization approach combined with
In particular, two different models, denoted as MA and limit analysis suitable for the evaluation of the collapse loads
MB, are assumed for the masonry material. Model MA and failure mechanisms of FRP-reinforced masonry panels is
is an orthotropic elastic-perfectly plastic continuum model presented.
with a Hoffmann failure criterion, whereas Model MB is The application of FRP strips on masonry (Fig. 1(a)) is
an orthotropic softening continuum model including a Hill- treated by adopting a simplified multistep approach.
In the first step, denoted here as micro-scale (Fig. 1(b)),
type yield criterion in compression and a Rankine-type yield
un-reinforced masonry, regarded as a periodic heterogeneous
criterion in tension (see [10]). For FRP-strengthened panels,
material, is substituted with a homogeneous macroscopic
only Model MB is adopted for masonry.
Regarding the behaviour of the layer between masonry material using a homogenization technique. In particular, an
and the FRP strengthening, it can be simulated by adopting estimation of the homogenized un-reinforced masonry strength
“special” elements (interface elements) which connect the domain is obtained by means of a micro-mechanical model
nodes of the mesh of panel to the nodes of the FRP elements based on the lower bound theorem of limit analysis, as proposed
or, in a simplified manner, adopting “special” constitutive laws by Milani et al. [8,9].
In the second step, macro-scale (Fig. 1(c)), FRP-
for FRP-strengthening elements. In this paper, for the FRP
reinforcement strips are introduced on the already homogenized
strips, two different models, denoted as MRA and MRB, are
masonry material.
used. Both models consider the reinforcement perfectly glued
Adopting an associated flow rule for the constituent
to the masonry support. In model MRA the reinforcement is
materials, a lower bound approximation of the homogenized
modelled by a linear-elastic constitutive relation, whereas in
strength domain S hom of un-reinforced masonry can be
model MRB a bilinear constitutive relation characterized by
obtained, as proved in Ref. [8], by solving the following
brittle behaviour is adopted, assuming different values of the
problem:
axial strength of the FRP elements. In both cases, the FRP
strips are modelled using truss elements directly connected to Z
1
the nodes of the mesh of the panels. 6 = hσ (y)i = σ (y)dY M (a)
|Y M | Y M
The paper is organized as follows. In Section 2 the
σ (y) · n(y) antiperiodic on ∂Y M (b)
micro–macro model is presented, deriving the upper and lower ∂ S hom = max{Σ }| (1)
divσ (y)
= 0 onblock YM (c)
bounds of the collapse load. In Section 3 the macro modelling
σ (y) ∈ S ∀y ∈ block
(d)
with
approach is described, presenting two material models for the σ (y) ∈ S mortar ∀y ∈ mortar
max
1 4k
X
[Σ11 Σ22 Σ12 ] = λnΣ =
Z h Y k=1
(k) (k) (k)
i
max{λ} σ11 σ22 σ12 dY (a)
∂ S̃ hom = P ≡ [Σ11 Σ22 Σ12 ]T |
such that Y
(k)
σi j = X(k) (y) SiTj y ∈ Y k (b)
(k)
σ ∈ S k k = 1, . . . , 4k max (c)
Box I.
Fig. 2. Subdivision of the elementary cell in subdomains for the micro-mechanical model.
boundary of S hom ; S block and S mortar represent the bricks and subdomain and Si j is the vector of global linearly independent
the mortar strength domains, respectively. unknowns.
Condition (1)(a) defines the macroscopic stress Σ as an Since the polynomial stress field in Eq. (2) is equilibrated
average of the microscopic stress σ on the periodic cell, and satisfies periodicity conditions, problem (1) reduces to the
whereas condition (1)(b) represents the continuity of the micro- (non)linear programming problem given in Box I.
stress vector between two adjacent unit cells. Condition (1)(c) In Box I
imposes micro-equilibrium on the cell and condition (1)(d) • ∂ S̃ hom is the boundary of the approximate homogenized
represents the material admissibility for blocks and mortar, strength domain S̃ hom ;
respectively. T
• P ≡ Σ11 Σ22 Σ12 is a typical point of the surface
An approximate but satisfactory estimation of masonry
∂ S̃ hom ;
homogenized strength domain S̃ hom can be obtained by T
• nΣ = α11 α22 α12 is a unit vector in the macroscopic
adopting the simplified micro-mechanical model [8,9]. One-
stress space Σ11 Σ22 Σ12 , such that λnΣ represents a
fourth of the periodic unit cell is subdivided into nine
geometrical elementary entities (subdomains), as shown in macroscopic stress state belonging to ∂ S̃ hom ;
Fig. 2, so that the whole cell is subdivided into 36 subdomains. • S k stands for the set of admissible stresses of the component
A plane stress state is assumed for masonry and polynomial (unit or mortar) belonging to the kth geometrical subdomain.
distributions of degree m are assumed a priori for the stress A refined piecewise linear approximation of ∂ S̃ hom can be
components within each subdomain. In this way, (m + 1)(m + obtained by collecting a sufficient number of stress states P on
2)/2 unknowns for each subdomain representing polynomial ∂ S̃ hom , evaluated by solving problem given in Box I.
coefficients are introduced. In this way, a lower bound approximation of the
In order to reduce the number of unknowns, continuity of homogenized masonry strength domain is recovered. Such
the stress vector at each internal interface between adjacent strength domain is implemented in lower and upper bound FE
subdomains, equilibrium equations inside each subdomain and limit analysis codes and static and kinematic limit analyses of
anti-periodicity conditions are imposed. In this way, the typical masonry structures reinforced with FRP strips (structural level)
i jth component (i, j = 1, 2) of the micro-stress in the kth are performed.
subdomain can be written as:
2.1. The lower bound approach
(k) (k)
σi j = Xi j (y)SiTj y ∈ Yk (2)
In this section, a 2D lower bound FE limit analysis
(k)
where, for each value of i and j, Xi j (y) is a vector which procedure, based on a modification of the equilibrated
contains only geometrical information of the unit cell and of the triangular element proposed by Sloan [14], is presented.
E. Grande et al. / Engineering Structures 30 (2008) 1842–1860 1847
Fig. 3. Geometry of a typical reinforced masonry wall, FE discretization and geometrical properties of FRP strips.
Fig. 4. Linear stress interpolation inside a triangular element and continuity of the stress vector on the edge between adjacent elements.
A reinforced masonry wall Ω = Ω f ∪ Ωm is considered, actions due to the presence of the reinforcement strips:
with Ωm and Ω f the un-reinforced and reinforced part of Ω
respectively. The reinforcement is constituted by FRP strips; t div σ + tb − 2τ = 0 (3)
(k) (k)
the typical kth strip is characterized by width lw , length ls where b denotes the body forces and τ is the interface shear
(k)
and fibre direction identified by the angle ϑs , as schematically stress vector. As a linear interpolation is chosen for the stress
reported in Fig. 3. The unit vectors that are parallel and state within each element, 2 equilibrium equations can be
orthogonal to the fibre direction of the kth strip are denoted written for each element, one for the x-direction and one for
by s(k) and t(k) respectively. A finite element discretization is the y-direction. Such equations can be re-written in terms of the
performed by ensuring that there are no elements in which the nodal unknown stresses and of the shear actions τ (i) as follows:
masonry is only partially reinforced, as shown in Fig. 3.
(k)
Thus, a typical element E (i) ∈ Ω f is constituted by a central tA(i)
eq σ
(i)
= −tb(i) + 2R ϑ S τ (i) (4)
masonry layer of thickness t, subjected to a plane stress state,
and by two additional external FRP layers of thickness s (k) . where:
It is assumed that only shear stresses τ (i) with direction s(k)
h i
η1 ζ1 η2 ζ2 η3 ζ3
• A(i)
eq =
1
2A(i) 0
0
ζ1 η1 0
0
ζ2 η2 0
0
ζ3 η3 , with
(see Fig. 3) can act at the interface between the masonry and
the reinforcement. In a typical element E (k) ∈ Ωm , the stress ζ i = x j − xk , η i = yk − y j , i = 1, 2, 3; k = 2, 3, 1; j =
is assumed varying linearly, in agreement with the assumption 3, 1, 2;
adopted in Ref. [14]. • A(i) is the area of the ith element;
iT
(k) (k) (k)
h
In this way, 9 unknown nodal stress parameters are • R ϑ S = cos ϑ S sin ϑ S ;
introduced for each element, so that 3 stress parameters
(i.e. σxNx σxNy σ yyN
, see Fig. 4) are associated at each node • σ (i) , b(i) represent the 9 nodal unknown stresses and the
(k) body forces vector relative to the ith element, respectively.
N E of an element. Statically admissible stress discontinuities
can occur between adjacent triangles, assuming only the stress Regarding the external FRP layers of the typical element
vector continuous at interfaces. E (i) ∈ Ω f , only the stress σss parallel to the fibre direction
For each E (i) ∈ Ω f , 2 equilibrium equations for the central is assumed acting, whereas the other stresses are enforced to
masonry layer are imposed by taking into account the shear be zero, i.e. σtt = σts = 0, as reported in Figs. 3 and 5.
1848 E. Grande et al. / Engineering Structures 30 (2008) 1842–1860
Fig. 5. Shear stress distribution on masonry/FRP interface. (a) masonry layer, (b) FRP layer.
Furthermore, the stress component σss is assumed varying programming problem is derived at a structural level:
linearly inside each FRP element. n T T o
max λ|Aeq XT λ = beq , Ain XT λ ≤ bin
Equilibrium is satisfied by imposing the following additional (6)
equality constraint inside each element, that corresponds to
impose the equilibrium along the directions s: where λ is the load multiplier at failure and X contains
h i the (assembled) masonry in-plane stress parameters, the
s (i) η1(i) η2(i) η3(i) σss (1,i) (2,i)
σss (3,i) T = −2A(i) τ (i)
σss reinforcement stress vector σ ss and the interfaces shear
stress vector τ . Matrices Aeq and Ain collect the coefficients
(5)
of the equilibrium equations, previously discussed, and the
where: coefficients of admissibility inequalities, respectively, whereas
(i) beq and bin are the corresponding right-hand sides.
• ηq = t¯k − t¯j , q = 1, 2, 3; k = 2, 3, 1; j = 3, 1, 2
(Fig. 5).
2.2. The upper bound approach
Additional constraints on the nodal stresses of each FRP
element are imposed in order to ensure the continuity of the The upper bound approach adopted in this section is
stress vector along the edges of adjacent triangles. based on the formulation originally developed by Sloan and
It can be demonstrated (Fig. 5(a)) that such condition is Kleeman [15], which assumes a triangular discretization of the
(3,i) (3, j) (2,i) (1, j)
satisfied by imposing σss = σss and σss = σss where domain and possible discontinuities of the velocity field along
i and j are adjacent elements with common nodes (3, i) ≡ the edges of adjacent triangles.
(3, j) and (2, i) ≡ (1, j). For each element E, two velocity unknowns per node i, u ix x
Admissibility conditions for the reinforcement layers are and u iyy (one along x and one along y, see Fig. 6(a)) are
imposed on σss stress for each element in the form − f d−f ≤ introduced.
(r,i)
σss ≤ f d+f ∀i r = 1, 2, 3 where f d+f and f d−f are the Jumps of velocities at interfaces are supposed to vary
failure tensile and compressive strengths of the reinforcement linearly, thus four unknowns collected in the vector ∆u I =
respectively, corresponding to the failure of the FRP or to the [∆v1 ∆u 1 ∆v2 ∆u 2 ]T are introduced at each interface. The
delamination of the FRP–masonry interface. In a similar way components of the vector ∆u represent the normal ∆vi
(i)
shear interface actions τ are assumed to satisfy the inequality and tangential ∆u i jumps of velocities (with respect to the
τ (i) ≤ f b ∀i where f b is the ultimate shear strength of discontinuity direction) evaluated at nodes i = 1 and i = 2
the interface, evaluated according to the indications contained of the interface (see Fig. 6(b)).
in [22] (Appendix B). For any pair of nodes lying on the interface between
Equilibrium constraints, boundary conditions on stresses two contiguous elements m and n, the tangential and normal
and admissibility conditions for masonry, FRP and interfaces velocity jumps can be written in terms of the Cartesian nodal
masonry/FRP are suitably assembled and the following linear velocities of elements m − n; thus, four linear equations for
E. Grande et al. / Engineering Structures 30 (2008) 1842–1860 1849
Fig. 6. Upper bound limit analysis. (a) field of velocities and plastic multipliers rates. (b) jump of velocities on interfaces.
E,ass I,ass I FRP ,ass FRP,ass
min CTE λ̇ + CTI λ̇ + CTFRP λ̇ + CT λ̇
I FRP
( eq eq
à U = b̃
such that
I FRP ,ass
I,ass E,ass FRP,ass
λ̇ λ̇ ≥ 0 λ̇ ≥ 0 λ̇
≥0 ≥0
(9)
E E
u ss,i = ũ M,i + Γ̇iE (8) +(q) −(q)
where λ̇ E FRP and λ̇ E FRP are the plastic multiplier rates of the
I I
E represents the velocity in the brickwork, evaluated
where ũ M,i triangle E associated to node q and corresponding to plastic
along strips directions at node i. dissipation on interfaces when τ (q) = f b and τ (q) = − f b ,
After assemblage operations, a linear programming problem respectively.
is obtained, in which the objective function consists in the In a similar way, the power dissipated in the FRP material
minimization of the total internal dissipated power: can be written as:
1850 E. Grande et al. / Engineering Structures 30 (2008) 1842–1860
Fig. 8. Anisotropic Rankine–Hill model implemented in DIANA [13]. (a) Rankine–Hill yield condition. (b) Hardening–softening law for Hill criterion. (c) Softening
law for Rankine criterion.
Summarizing, the two presented masonry models are where all the symbols are defined in the list of symbols at
characterized by different features. The first model (MA) the end of the paper. Here it is worth noting that γ f d is a
is simpler and it allows us to determine the values of factor that takes into account the modality of the application
the parameters governing the model by means of standard of the reinforcement system (it is assumed to be equal to 1.20
experimental tests. The second model (MB) is more complete if the reinforcement is applied according to the indications
as it accounts for the special orthotropic character of the contained in chapter 2 of [22]), or equal to 1.5 if this condition
masonry and, moreover, it considers the softening behaviour; is not satisfied; γ M is the partial safety factor for masonry
on the other hand, the evaluation of all the parameters (seven material [23], assumed to be equal to 1.0 for q obtaining the
strength parameters and five inelastic parameters) governing E FRP ·tFRP
characteristic value of bond strength; le = 2· f mtm is the
the model requires the development of nonstandard laboratory optimal bond length of FRP corresponding to the minimal bond
tests. length able to carry the maximum anchorage force.
In the code, the term Γ Fk is also introduced and it represents
3.2. FRP-strengthening models: Model MRA and Model MRB the characteristic value of the specific fracture energy of the
FRP strengthened masonry. In particular, when the debonding
One of the most important aspects in the application of involves the first masonry layers, in [22] the following relation
composite materials for strengthening structural elements is is proposed:
the adhesion between the reinforcing and reinforced materials. p
In particular, when delamination from the support occurs, the Γ Fk = c1 f mk · f mtm [ f in N/mm2 ] (14)
effectiveness of the reinforcement vanishes. This phenomenon where all the symbols are defined in the list of symbols at the
is very complex to model, because it involves materials with end of the paper.
different properties (masonry, FRP and glue layer) and depends In the light of these considerations, two different models
on several parameters. Some experimental and numerical have been adopted for the FRP elements in the framework of
investigations have been conducted in the last decades for a macro-modelling approach. In the first model (denoted as
concrete structures but only in the last few years the decohesion MRA) a linear-elastic behaviour has been considered while in
problem has been investigated for masonry structures [18–21]. the second one (denoted as MRB), an elastic-brittle behaviour
Experimental studies demonstrated that the decohesion occurs has been adopted.
because of the failure of the masonry material; in fact, the It is worth noting that, for both the adopted MRA and
delaminated FRP presents a not negligible layer of masonry MRB models, truss elements have been used for FRP,
material on the debonded surface. imposing a “perfect adhesion” between the nodes of trusses
A rigorous methodology to directly take into account in a and the corresponding nodes of the mesh of the panels, so
numerical model the behaviour of the layer between masonry avoiding relative displacements between the masonry and the
and the FRP reinforcement is the use of the interface model reinforcement. The delamination phenomenon between FRP
concept. According to this model, the forces acting on the and masonry support has been accounted in a simplified
interface are related to the relative displacement of the two sides manner, i.e. adopting special constitutive laws for truss
(masonry and FRP). The use of this model is rather complex elements. This choice could be questionable but it simplifies
because it implies the knowledge of the parameters that define the numerical model and, moreover, it allows us to obtain
the behaviour of the FRP/masonry layer. In fact, even if from solutions which are in satisfactory agreement with experimental
the experimental tests available in the technical literature some evidences.
common aspects have emerged, such as for instance the quasi- In principle, as announced above, an enhanced modelling
brittle behaviour and the mechanisms of separation of the FRP could require the use of interface elements to model the
from the bricks surfaces, the heterogeneity of the masonry behaviour of FRP/masonry interaction.
material and the reduced number of performed tests, do not Indeed, several interface constitutive laws are implemented
allow us to have at disposal a sufficiently wide data set to in commercial computer codes, which are commonly based on
completely assess the interface behaviour. the assumption that damage state of the interface elements is
In [22], a simplified approach is proposed to evaluate function of the history of the relative displacements occurring
the delamination phenomenon. In fact, it is suggested to between the nodes of the interface and it does not depend on
evaluate the maximum strength f f dd of FRP elements using the damage evolution of the support material.
the following relation if the bond length lb is smaller than the On the other hand, experimental evidences show that FRP
optimal bond length le : decohesion occurs because of the damage of a consistent layer
of masonry support. In this way, the mechanical properties of
this layer depend on mechanical properties of the masonry and
lb lb
f f dd,rid = f f dd 2− (12) on its actual state, e.g. on the level of damage of the masonry
le le
support. As a consequence, a structural model which does
or, the following relation if lb ≥ le : not consider the coupling between the interface and masonry
s damage appears unreliable.
1 2 · E FRP · Γ F K This implies that sophisticated models affecting the complex
f f dd = √ (13)
γ f d γM tFRP FRP/support interaction mechanisms should be developed.
1852 E. Grande et al. / Engineering Structures 30 (2008) 1842–1860
Fig. 9. Geometric dimensions [mm] and FRP strip arrangement of the examined panels.
Moreover, a further important aspect which is not For all these reasons, the proposed very simple, but quite
well reproduced by classic modelling approaches could be satisfactory, models are adopted throughout this paper to
remarked. Indeed, the damage of the masonry material is reproduce the FRP and FRP–masonry interface behaviour.
generally treated by considering continuous models. For
instance, the masonry model adopted in this paper is based on 4. Numerical applications
the smeared crack approach with strain softening. The main
feature of this method consists of representing fracture in a On the basis of the proposed models and numerical
smeared manner [24]: the effect of infinitely many parallel approaches, several nonlinear analyses have been performed
cracks of infinitely small opening, that are considered to be with reference to un-strengthened and FRP-strengthened
continuously distributed over the finite element, is modelled panels. The obtained results in terms of the global response of
by reducing the material stiffness and strength in the direction the panels, have been compared with the data available from
normal to the cracks after the peak of strength of the material experimental tests.
has been reached.
4.1. Experimental cases
When a certain level of damage occurs in the masonry,
macro-cracks start to develop. The presence of the fractures Five masonry panels with and without CFRP (Carbon FRP)
induces local stress concentrations at the interface layer strips reinforcement are examined (Fig. 9). Two typologies
bridging the crack, leading to local redistributions of the of panels with different dimensions and geometric aspect are
interface stresses. This phenomenon could influence the bond considered: a panel without openings of dimensions 290 ×
strength and, hence, the overall behaviour of reinforcement and 270 mm2 , here denoted as PAN-A and a panel of dimensions
it could be not properly reproduced by considering continuous 416 × 414 mm2 , labelled as PAN-B with a central rectangular
damage models. hole of dimensions 184 × 156 mm2 . Different CFRP strip
As a final comment, it could be emphasized that masonry arrangements characterize both PAN-A and PAN-B: horizontal
structures contain heterogeneities, i.e. discontinuities, which strips (PAN-A1), symmetrical diagonal strips (PAN-A2) and
are due to the presence of mortar joints. Of course, such asymmetrical diagonal strips (PAN-B1), see Fig. 9.
discontinuities cannot be considered adopting a homogeneous For these panels, several results are available from a previous
continuum model. experimental investigation conducted by Milani et al. [25]. The
E. Grande et al. / Engineering Structures 30 (2008) 1842–1860 1853
Fig. 10. Mechanical properties adopted in the homogenisation model for mortar joints reduced to interfaces.
Table 1
Mechanical characteristics of the components of the panels
Table 2
Masonry mechanical parameters (Model MA)
Table 3
Masonry mechanical parameters (Model MB, MRA, MRB)
Young’s Modulus along x-direction (E x x ) 1412 MPa
Young’s Modulus along y-direction (E yy ) 1050 MPa
Poisson’s ratio (νx y ) 0.1762
Shear modulus (G x y ) 367 MPa
Mass density (γm ) 1.6E − 9 N s2 /mm
Tensile Strength along x-direction (σt x ) 8.0E–1 MPa
Tensile strength along y-direction (σt y ) 2.0E–1 MPa
Compressive strength along x-direction (σcx ) 8.0 MPa
Compressive strength along y-direction (σcy ) 6.67 MPa
Fracture energy in tension along x-direction (G f x ) 2.0E − 2 N mm/mm2
Fracture energy in tension along y-direction (G f y ) 2.0E − 2 N mm/mm2
Fracture energy in compression along x-direction (G f cx ) 5.00 N mm/mm2
Fracture energy in compression along y-direction (G f cx ) 10.0 N mm/mm2
Factor that determines the shear stress contribution to the tensile failure (α) 1.73
Factor which couples the normal compressive stresses (β) −1.05
Factor which controls shear stress contribution to compressive failure (γ ) 1.2
Factor that specifies the equivalent plastic strain corresponding to the peak compressive stress (k p ) 8.0E–4
Fig. 12. Schematic representation of the bond length of FRP truss element number 5 and values of the design bond strength along FRP strips.
good agreement between all the admissible stress sets. Some and different values for the strength of the FRP elements. In
differences occur in certain regions particularly for the Hoffman particular, the strength of FRP elements has been evaluated
strength domain. according to the relations (12) and (13) and for each truss
For what concerns the mechanical parameters adopted for element (node-to-node of the mesh of the panel) the bond length
strengthening elements, two different constitutive laws have lb has been assumed equal to the distance between the middle
been considered. In the case of the Model MRA, a linear- section of the truss element and the closest edge of the FRP
elastic behaviour has been assumed with a stiffness value (Fig. 12). In Fig. 12 the values of the bond strength of the
equal to Young’s modulus of the FRP-strengthening (E FRP = FRP elements have been depicted for PAN-A1 and PAN-A2
160,000 MPa) and without limiting the axial strength of reporting on the horizontal axis the FRP element number of the
FRP elements. In the case of model MRB, an elastic-brittle FE model. It is noticeable that the truss elements near the edges
behaviour has been considered by adopting an elastic stiffness of the panel present strength values less than the design value
value equal to Young’s modulus of the FRP strengthening f f dd .
E. Grande et al. / Engineering Structures 30 (2008) 1842–1860 1855
Table 4
FRP-strengthening mechanical parameters (Model MRA, MRB)
Thickness of FRP reinforcement (tFRP ) 0.2 mm
Elastic modulus of FRP reinforcement (E FRP ) 160,000 MPa
Factor c1 [22] (c1 ) 0.015
Factor γ f d [22] (γ f d ) 1.2
Factor γ M [23] (γ M ) 1
Fracture energy [22] (Γ F K ) 0.073 N/mm
Design bond strength [23] ( f f dd ) 164 MPa
Fig. 13. Mesh used in the homogenized limit analysis approach. (a) PAN-A1
series. (b) PAN-A and PAN-A2 series.
Fig. 14. External force-point P vertical displacement. Comparison between
For PAN-B1 (Fig. 9), because of the small length of the numerical and experimental results. (a) PAN-A series. (b) PAN-A1 series. (c)
PAN-A2 series.
diagonal FRP strips, the axial strength of the trusses has been
assumed the same for all the elements of each diagonal strip. The second set of numerical simulations refers to the series
In particular, the average values of the axial strengths of the PAN-B and PAN-B1 (Fig. 9) and they regard a masonry
trusses of each strip have been assumed (i.e. f f dd = 90 MPa wall with a central rectangular opening subjected to a vertical
for the truss elements of the longer strip and f f dd = 79 MPa concentrated load applied on the top edge.
for the truss elements of shorter strip). For what concerns limit analysis simulations, a linearized
4.3. Comparison between experimental and numerical results Lourenço and Rots [29] failure criterion for joints reduced to
interfaces has been used, see Fig. 10 for a synopsis of the
In this section, the accuracy of the numerical results obtained mechanical properties adopted. It is worth noting that joints
by means of both the micro–macro and phenomenological compressive strength f c adopted in the numerical simulations
approaches proposed in Sections 2 and 3, respectively, are is assumed equal to the experimental masonry compressive
assessed through a comparison with experimental results. strength value, because of the fact that both 3D effects and
The first set of analyzed examples refers to the series PAN- brittle behaviour of bricks cannot be reproduced with the
A, PAN-A1 and PAN-A2 (Fig. 9), consisting of masonry panels rigid-plastic homogenization approach at hand. According to
acting as deep beams. experimental data, see Table 1, for bricks a Mohr–Coulomb
1856 E. Grande et al. / Engineering Structures 30 (2008) 1842–1860
Fig. 15. PAN-A, PAN-A1, PAN-A2 series numerical results from the homogenized limit analysis approach: stress distribution at collapse and field of velocities at
collapse. Detail A: crushing failure of the supports. Detail B: shear failure under the point of application of the load. Detail C: diagonal failure. Detail D: flexional
behaviour of the support due to the horizontal reinforcement. Details E: enhancement of panel diagonal strength due to inclined reinforcement.
failure criterion in plane stress has been adopted with f c = primarily involve masonry compressive crushing and shear
15.66 N/mm2 and, as usually assumed, f t = 1/10 f c . It failure. Finally, a linearization of the homogenized failure
is worth underlining that experimental evidences show that surface with 50 planes was used for performing both static and
tensile strength ranges between 1/10 and 1/20 of compressive kinematics limit analyses.
strength. Despite this relatively wide variability, numerical For the macro-modelling approach, the nonlinear static
computations show that, in the worst case, the difference analyses (push-over) have been performed by adopting, for the
between failure loads for the examples analyzed in this paper discretization of the panels, four and three-node isoparametric
obtained by assuming respectively 1/20 and 1/10 is lower plane stress finite elements. In this case, during the incremental-
than 10%. Such a small difference is mainly a consequence iterative procedure, the external load was increased at each
of the failure mechanisms of the analyzed panels, which iteration by directly increasing the external force vector. The
E. Grande et al. / Engineering Structures 30 (2008) 1842–1860 1857
Fig. 17. PAN-B series numerical results. (a) macroscopic model MA, stress distribution at collapse at maximum load level. (b) macroscopic model MA, deformed
shape at maximum load level. (c) homogenized limit analysis approach, stress distribution at collapse. (d) homogenized limit analysis approach, field of velocities
at collapse. Detail A: compressive failure and formation of a compressed pier under the area of application of the load.
As the lower and upper bound FE simulations show, in PAN- For both PAN-B and PAN-B1, the adopted theoretical
A1 series the horizontal strip acts as a tie. Even though the approaches lead to results which are in good agreement
two-strut model of the un-strengthened case remains essentially with the experimental data (the maximum difference between
unchanged, both the compressed sections increase as well as the experimental and theoretical strength values is less than
the intensity. In PAN-A2 series the principal stress distribution 16% corresponding to the lower bound approach adopted for
at collapse and the failure field of velocities result in a PAN-B).
change both of the direction of the compressed struts and In the case of the un-strengthened panel (PAN-B) it
in the failure mechanism. The deformed shape at collapse is interesting to notice that the force–displacement curves
shows compression near the supports, shear under the load and obtained from the macroscopic plasticity models MA and MB
delamination of the diagonal reinforcement. differ both in terms of peak-strength and stiffness values. In
particular, the curve obtained from the Model MA shows a
4.3.2. Panels PAN-B and PAN-B1 stiffness reduction that starts from a load value equal to 43%
Experimental results collected for PAN-B and PAN-B1 of the peak-strength of the panel. This effect was not observed
series show that the introduction of the FRP reinforcement does during experimental tests.
not determine a significant improvement of the load bearing Similarly to PAN-A1 series, in the case of PAN-B1 it is
capacity. In fact, failure occurs in both cases for compression worth noting that the macroscopic anisotropic plasticity model
of the vertical masonry strut under the point of application of with elasto-brittle behaviour of the FRP elements (Model
the load. This is captured by the numerical models adopted in MRB), exhibits a force–displacement curve in good agreement
this paper. with experimental data. On the contrary, the adoption of an
In Fig. 16, the external force–displacement curves of the elastic behaviour for the FRP strips furnishes a limit load of
point of application of the load are reported, as well as the the panel and a corresponding displacement greater than the
failure load obtained by means of the lower and upper bound experimental ones (the difference in terms of load is about 9%
FE limit analyses. while in terms of displacement is about 10%). On the other
E. Grande et al. / Engineering Structures 30 (2008) 1842–1860 1859
Fig. 18. PAN-B1 series numerical results. (a) macroscopic model MRB, stress distribution at collapse at maximum load level. (b) macroscopic model MRB,
deformed shape at maximum load level. (c) homogenized limit analysis approach, stress distribution at collapse. (d) homogenized limit analysis approach, field of
velocities at collapse. Detail A: the inefficient disposition of the reinforcement leads to a stress distribution at collapse similar to the un-reinforced case.
hand, also limit analysis predictions of the ultimate load are in ultimate load have been compared with the experimental ones.
good agreement both with experimental data and Model MRB. Deformed shapes at collapse and principal stress distribution at
Finally, in Figs. 17 and 18 the deformed shapes at collapse the ultimate state have also been reported.
and the principal stress distribution at the ultimate state In all the examined cases, the macroscopic phenomenolog-
obtained both with the limit analysis and the macroscopic ical approach based on an orthotropic plasticity model for ma-
approach are reported for PAN-B and PAN-B1. These results sonry and on an elastic-brittle behaviour for FRP strengthening,
confirm that the introduction of FRP strengthening does not reproduces the experimental behaviour well both for pre- and
affect the stress distribution. post-peak range.
On the other hand, also the adoption of a homogenization
5. Conclusions limit analysis leads to ultimate load values in some cases in
good agreement with the experimental ones.
With the aim of reproducing the structural response of Both experimental failure loads and load–displacement
un-strengthened and FRP-strengthened masonry panels, in curves are satisfactorily reproduced with all the adopted
this paper a homogenized limit analysis and a macroscopic models. However, the post-peak behaviour is captured well
phenomenological approach have been adopted using different only when brittle phenomena are taken into account, i.e. when
constitutive laws and yielding criteria. models MB and MRB are used.
In the modelling process, particular regard has been The proposed models are able to capture the resistant
addressed to the phase of parameter setting and to the modelling mechanism of un-strengthened and FRP-strengthened masonry
strategy adopted to reproduce masonry/FRP interaction panels. In particular, information about the load capacity and
phenomena. Simplified models able to take into account the the post-peak behaviour (i.e. the ductility) can be deduced in
delamination of the FRP from the support have been proposed order to assess the efficacy of the reinforcing system. It is also
both for the macroscopic and the limit analysis approach, also important to underline that the simplified approach proposed to
considering the indications contained in [22]. take into account the contribution of the reinforcement is able
With the aim of assessing the adopted theoretical models, to reproduce the delamination process of the FRP strengthening
the obtained results in terms of force–displacement curves and when the strips are not mechanically fixed to the masonry
1860 E. Grande et al. / Engineering Structures 30 (2008) 1842–1860
support and the delamination process starts from the free edges [12] Bazant ZP. Fracture and size effect in concrete and other quasibrittle
of the strips. In fact, in this case the delaminated parts of the materials. Boca Raton: CRC Press; 1998.
strips are no more in contact with the masonry support and, as [13] DIANA 9.1. Displacement analysis finite element software. Version 9.1.
Delf (The Netherlands): TNO-Building Division; 2000.
a consequence, they are not able to contribute to the strength
[14] Sloan SW. Lower bound limit analysis using finite elements and
of the panel. On the contrary, when the ends of the FRP strips linear programming. International Journal for Numerical and Analytical
were mechanically fixed to the surface of the panels this leads Methods in Geomechanics 1988;12:61–77.
to a resistant mechanism of the FRP strips quite different: the [15] Sloan SW, Kleeman PW. Upper bound limit analysis using discontinuous
parts of the strips that are not in contact with the masonry velocity fields. Computer Methods in Applied Mechanics and Engineering
support continue to contribute to the strength of panel till the 1995;127(1–4):293–314.
[16] Hoffmann O. The brittle strength of orthotropic materials. Journal on
FRP tensile strength is reached. In this case a different model Computational Mathematics 1967;1:200–6.
must be chosen. [17] Hill R. A theory of the yielding and plastic flow of anisotropic materials.
London (UK): Proc. Roy. Soc.; 1947.
Acknowledgement [18] Roko K, Boothby TE, Bakis CF. Failure modes of sheet bonded fiber
reinforced to brick masonry. In: Proceedings of the fourth international
This research has been possible thanks to the financial symposium on fiber reinforced polymer for RC structures. ACI, SP-188.
supports of the project RELUIS from the Italian Ministry of 1999.
[19] Casareto M, Olivieri A, Romelli A, Lagomarsino S. Bond behaviour of
the Civil Protection.
FRP laminates adherent to masonry. In: Proceedings of the international
conference advancing with composites. 2003.
References [20] Ceroni F, Pecce MR, Manfredi G, Marcari G. Experimental bond
behaviour in masonry elements externally reinforced with FRP
[1] van Zijl G. Computational modelling of masonry creep and shrinkage. laminates. In: Proceedings of the international conference composites in
Ph.D. thesis. Delft (The Netherlands): Delft University of Technology;
constructions. 2003.
2000.
[21] Aiello MA, Sciolti SM. Bond analysis of masonry structures strengthened
[2] Berto L, Saetta A, Scotta R, Vitaliani R. An orthotropic damage model
with CFRP sheets. Journal of Construction and Building Materials 2006;
for masonry structures. International Journal for Numerical Methods in
20:90–100.
Engineering 2002;55:127–57.
[22] CNR-DT200. Guide for the design and construction of externally bonded
[3] Pietruszczak S, Ushaksaraei R. Description of inelastic behaviour of
FRP systems for strengthening existing structures. Italy: C N R, National
structural masonry. International Journal on Solids and Structures 2003;
Research Council; 2006.
40:4003–19.
[23] D.M.LL.PP. (1987) Norme tecniche per la progettazione, esecuzione e
[4] Lofti HR, Shing BP. Interface model applied to fracture of masonry
collaudo degli edifici in muratura e per il loro consolidamento. Italy. 1987.
structures. Journal of Structural Engineering, ASCE 1994;120:63–80.
[5] Giambanco G, Di Gati L. A cohesive interface model for the structural [24] Rashid YR. Analysis of prestressed concrete pressure vessels. Nuclear
mechanics of block masonry. Mechanics Research Communications Engineering Design 1968;7(4):334–5.
1997;24(5):503–12. [25] Milani G, Rotunno T, Sacco E, Tralli A. Failure load of FRP strengthened
[6] Alfano G, Sacco E. Combining interface damage and friction in a masonry walls: Experimental results and numerical models. Structural
cohesive-zone model. International Journal for Numerical Methods in Durability and Health Monitoring 2006;2(1):29–50.
Engineering 2006;68(5):542–82. [26] Lurati F, Thurlimann B. Tests in concrete masonry walls. Report no. 8401-
[7] Luciano R, Sacco E. Homogenization technique and damage model for 3. Zurich (Switzerland): Institute of Structural Engineering, ETH Zurich;
old masonry material. Int J Solids and Structures 1997;34(4):3191–208. 1990.
[8] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry [27] van der Pluijm R. Material properties of masonry and its components
walls. Part I: failure surfaces. Computer Structures 2006;84:166–80. under tension and shear. In: Proc. 6th Canadian masonry symposium.
[9] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry 1992.
walls. Part II: structural examples. Computer Structures 2006;84:181–95. [28] van der Pluijm R. Shear behaviour of bed joints. In: Hamid AA, Harris
[10] Lourenço PB. Continuum model for masonry: parameter estimation and HG, editors, Proceedings of the 6th North American masonry conf.,
validation. ASCE Journal of Structural Engineering 1998;124(6):642–52. Philadelphia (PA, USA): Drexel University; 1993.
[11] Marfia S, Sacco E. Numerical procedure for elasto-plastic no-tension [29] Lourenço PB, Rots J. A multi-surface interface model for the analysis
model. International Journal for Computational Methods in Engineering of masonry structures. Journal of Engineering Mechanics ASCE 1997;
Science and Mechanics 2005;6:187–99. 123(7):660–8.