1 s2.0 S0141029622011920 Main

Engineering Structures 274 (2023) 115116
Contents lists available at ScienceDirect
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
An effective numerical modelling strategy for FRCM strengthened curved

masonry structures
Anna Castellano a, Aguinaldo Fraddosio b, Daniel V. Oliveira c, Mario D. Piccioni b,
Eleonora Ricci b, *, Elio Sacco d
a
Department of Mechanics, Mathematics and Management, Polytechnic University of Bari, Via Orabona, 4, 70125 Bari, Italy
b
Department of Civil Engineering Sciences and Architecture, Polytechnic University of Bari, Via Orabona, 4, 70125 Bari, Italy
c
Department of Civil Engineering, University of Minho, ISISE, Campus de Azurém, 4800-058 Guimarães, Portugal
d
Department of Structures for Engineering and Architecture, University of Naples Federico II, Via Claudio 21, 80125 Naples, Italy
A R T I C L E I N F O A B S T R A C T
Keywords: Fabric Reinforced Cementitious Matrix (FRCM) composites are currently considered a very effective solution for
Masonry strengthening masonry constructions. However, the mechanical interactions governing the response and the
Vault strength of FRCM reinforced masonry structures are very complex, especially in the case of curved structures.
FRCM
Moreover, these interactions involve several interfaces between different materials. Thus, the development of
Numerical modelling
accurate numerical models for curved FRCM reinforced masonry structures comes up against several difficulties,
and models too complex for practical applications can be obtained. In addition, several mechanical parameters
needed for the calculations are generally inaccessible by conventional experimental tests.
Here, a suitable numerical modelling strategy for FRCM strengthened curved masonry structures is proposed
to combine the accuracy in simulating the actual behaviour in terms of stiffness, strength and collapse mecha
nisms with a reasonable simplicity, making the proposed approach usable also by practitioners, by adopting
commercial codes and at a moderate computational effort. The relatively small number of mechanical param
eters characterizing the model can be determined by ordinary experimental tests on materials or by literature
formulations.
The proposed modelling strategy is validated with respect to experimental data found in literature concerning
a FRCM reinforced masonry barrel vault, and then is employed for studying the seismic capacity of the vault
through a pushover analysis. A broad sensitivity analysis sheds light on the effect of variations of the mechanical
parameters on the predicted overall behaviour, showing the robustness of the results obtainable through the
proposed approach concerning inaccuracies in the determination of the parameters often very difficult to
determine by ordinary experimental tests on masonry structures.
1. Introduction composites are known as Fabric Reinforced Cementitious Matrix

(FRCM) or Textile Reinforced Matrix (TRM) composites; in what fol
Fiber reinforced composites have been widely used in the last de lows, the denomination “FRCM” will be adopted.
cades for strengthening masonry structures because of their high The mechanical behaviour and the failure modes of FRCMs signifi
strength-to-weight ratio, low invasiveness and relatively easy installa cantly differ from that of the most established Fiber Reinforced Polymers
tion procedure. Recently, composites made of an inorganic matrix (FRP) composites due to a different kind of matrix. First, the stress–
generally reinforced with short fibers and embedding continuous fibers strain response of FRCM composites under uniaxial tension is mainly
(glass, carbon, PBO fibers and steel cords) in form of fabric gained trilinear, with each branch corresponding to a different damage level
relevance as strengthening materials, especially for historic construc (undamaged, cracking of the matrix and final damage of the fiber net)
tions [1,2], in addition to other conventional techniques [3]. These [4–6], whilst the FRP behaviour is mainly linear up to failure. Second,
* Corresponding author.
E-mail addresses: anna.castellano@poliba.it (A. Castellano), aguinaldo.fraddosio@poliba.it (A. Fraddosio), danvco@civil.uminho.pt (D.V. Oliveira), mariodaniele.
piccioni@poliba.it (M.D. Piccioni), eleonora.ricci@poliba.it (E. Ricci), elio.sacco@unina.it (E. Sacco).
https://doi.org/10.1016/j.engstruct.2022.115116
Received 13 May 2022; Received in revised form 11 September 2022; Accepted 9 October 2022
Available online 1 November 2022
0141-0296/© 2022 Elsevier Ltd. All rights reserved.
A. Castellano et al. Engineering Structures 274 (2023) 115116
the fabric-matrix bond strongly depends on the inorganic matrix pene simulation are presented and discussed in Section 6, including a com
tration between the filaments of the fabric, hindered by the presence of parison with the experimental results. A broad sensitivity analysis is
binder grains with a large diameter. Ineffective penetration might lead performed in Section 7 to study the overall response of the reinforced
to the so-called “telescopic failure” of the composite in tension, not structure. Indeed, a numerical model for a complex structural system
observed for FRP composites [7,8]. Finally, it has been experimentally like a FRCM-reinforced masonry requires inevitably several mechanical
observed that the friction among grains provides the matrix-substrate parameters. In principle, in practical applications, it is needed to
interface of residual strength, also not observed for FRP composites experimentally determine all of these parameters with the greatest
[9,10]. For the above, models developed for FRP composites cannot be possible accuracy. Anyway, for some parameters, this can be challenging
used to properly describe the mechanical behaviour of FRCM if not unfeasible. On the other hand, not all the material parameters have
composites. the same influence on the predicted structural response: therefore, the
Experimental and numerical study on the application of FRCM sensitivity analysis is useful for understanding which of the mechanical
composites for strengthening masonry structures, especially for arches parameters needs to be carefully determined to have representative
and vaults, is still the object of ongoing research. In particular, different numerical results; while a rough estimate starting from literature values
numerical procedures for representing FRCMs mechanical behaviour or formulations can be adopted for the remaining without significantly
have been proposed in the literature to date, based either on macro- or affecting the accuracy of the results. Finally, in Section 8 the proposed
micro-modelling approaches. A feature generally common to both ap modelling strategy has been applied to the evaluation of the seismic
proaches is the modelling of the FRCM-substrate interaction using non- capacity of the examined reinforced masonry vault through a pushover
linear interface elements. analysis.
In some recent research works on FRCM reinforced masonry, a
macro-modelling approach is adopted both for the masonry substrate 2. Modelling FRCM-reinforced curved masonry structures: An
and the composite. For example, in [11] a glass FRCM system is overview
modelled with shell elements using the Total Strain Crack model and
assuming experimental curves (multilinear in tension and parabolic in Experimental campaigns on FRCM retrofitted masonry arches and
compression). In [12] a similar approach is used, and plane stress ele vaults [17–23] show that due to the presence of the reinforcement, high
ments were employed for the composite, bonded to a panel by inelastic compressive stresses may develop in curved masonry structures, leading
interface elements to represent the possible debonding of the FRCM. to the crushing of masonry, very uncommon for unreinforced structures
A micro-modelling description has been proposed in [13] and [14]; and, at high load levels, sliding of masonry blocks along mortar joints
the fiber net and the mortar layers constituting the matrix are modelled might occur since FRCM reinforcements also allow for the development
separately. The mechanical behaviour of the matrix is described by a of substantially higher shear stresses.
smeared cracking approach, namely the Concrete Damage Plasticity The bond at the composite-substrate interface plays a crucial role in
model in [13] and the Total Strain Crack model in [14]. A relevant the collapse mechanism. Debonding is here understood as a partial
difference is that in [13] fibers and matrix are separately modelled, damage process that weakens the bond between different materials
whereas in [14] a special shell element formulated for reinforced con (like, e.g., the masonry substrate and the reinforcement layer), while
crete and embedding bar elements is used. Moreover, in [13] an elastic- detachment means the complete loss of bond, involving the separation
perfect plastic behaviour is assumed for the fibers while in [14] the between the materials. The failure of arches and barrel vaults
behaviour of the bars is considered linear elastic up to the failure. Notice strengthened at the extrados or intrados can be due to the debonding at
that both in [13] and [14] no interfaces are considered between the the matrix-substrate interface, associated with the cracking of the matrix
composite and the masonry substrate. where the reinforcement prevents the formation of hinges typical of the
The complexity of the non-linear phenomena to be described, often collapse mechanism of unstrengthened arches. As the load increases,
strictly interconnected, yields the risk of developing models too complex transversal cracks appear and propagate from the external matrix layer
for practical applications, useable only by the research community. to the inner layer up to the masonry substrate; then, tangential stresses,
Indeed, for the needs of practitioners, numerical models should depend τnt, develop at the matrix-substrate interface. In addition, stresses
on a reasonable number of mechanical parameters to be evaluated by normal to the reinforced masonry surface, σn, develop to radially
conventional experimental tests and/or deduced by acknowledged equilibrate the shear stresses τnt (see Fig. 1). In particular, for re
values or empirical laws in literature. Furthermore, it is very important inforcements at the intrados, these normal stresses negatively affect the
to know in advance how inaccuracies in the determination of these composite-support bond capacity, thus facilitating the debonding, which
parameters can affect the final results in terms of the overall response of can lead to the composite detachment. This phenomenon doesn’t occur
the structure. In this vein, the purpose of this research is to develop a for reinforcements applied at the extrados, where the effect of the cur
modelling strategy for FRCM reinforced curved masonry structures able vature is beneficial to the overall strength.
to combine the suitability of use by practitioners with a sufficiently In particular, [24] reports experimental and analytical studies about
accurate description of the mechanical behaviour in terms of stiffness, the influence of the masonry substrate curvature on the bond capacity of
peak loads, failure modes and post-peak behavior, generally strongly carbon FRCM and Steel Reinforced Grout (SRG). The study is conducted
dependent from the interactions between reinforcement and masonry on suitable specimens having curvatures simulating intrados re
support in terms of stress transfer and of bond integrity. To this aim, it is inforcements. It is shown that, as the curvature increases, both the bond
relevant not only the ease of the model, but also the use of constitutive strength and the ultimate displacement decrease because of the normal
laws characterized by a relatively small number of mechanical param stresses developing at the matrix-substrate interface (the reduction is
eters, consistent with that evaluable by conventional experimental tests, more marked for SRG).
or by new experimental approaches, still under investigation [15,16]. In general, the failure of arches strengthened at the extrados is due to
The paper contains a review of the modelling approaches for ma the sliding along joints and the detachment of the reinforcement at one
sonry strengthened with FRCMs and a brief discussion on the mechan of the abutments [18,19,25]. The failure is also associated with the
ical behaviour of masonry arches and barrel vaults reinforced with cracking of the matrix and the debonding at the matrix-substrate
FRCM composites (Section 2), being the main features of the developed interface where the reinforcement prevents the formation of hinges.
modelling strategy described in Section 3. Then, a representative case The breaking of the reinforcement fiber grid has been reported for a
study is considered from the literature (Section 4) and used as the structure where steel anchor plates were used [26] to fix the strength
reference for validating the modelling approach and for discussing its ening system at the masonry support. On the other hand, the failure of
effectiveness (Section 5). The results obtained by the numerical arches and barrel vaults strengthened at the intrados is generally
2
Fig. 1. Stress transfer between the arch/vault and reinforcements applied at the intrados (a) or extrados (b).
characterized by the debonding of the reinforcement near the load For a model to be sufficiently representative of the mechanical
application point [27]. The use of spike anchors or steel anchors pre behaviour, it should consider the following key aspects: debonding of
vents the reinforcement detachment but might result in the fracture of the composite at the composite-substrate interface; cracking of the
the matrix and the rupture of the fibers [19]. matrix; sliding of the fiber at the matrix-fiber interface; sliding and
Although arches and vaults represent fundamental structural ele crushing of the blocks; influence of the substrate curvature on the
ments in masonry construction, the development of simple and effective overall behaviour. Also, the presence of anchors increasing the bond
numerical models of these elements reinforced with FRCM composites between FRCM reinforcements and masonry, or of the infill are aspects
appears limited in the literature. Among them, [28] proposes a numer to be conveniently represented in simulations.
ical model for FRCM reinforced masonry arches, represented by a set of
rigid plates interacting through unidirectional links and connected to 3. Proposed modelling strategy
the substrate employing interface elements. In [29], a SRG strength
ening layer for a masonry arch was modelled by equivalent two-node Based on the analysis of the literature, here an effective modelling
truss elements perfectly bonded at the extrados of the arch. Although strategy is proposed for curved masonry structures strengthened with
the above numerical models succeed in describing the contribution of FRCM composites, aimed at representing the relevant mechanical as
the reinforcement in terms of increased load carrying capacity and pects at a relatively low computational cost. This strategy is formulated
ductility and in reproducing the debonding of the reinforcement, they in a way suitable for being adopted in commercial codes, and thus to
were not able to model the cracking of the matrix observed near the make the model available for practical applications. To this aim,
collapse. In [30] the mechanical behaviour of an arch externally modelling choices have been oriented towards the limitation of the
strengthened with PBO-FRCM composites, experimentally studied in number of the required mechanical parameters. The proposed modelling
[25], was numerically simulated by modelling the composite as a system strategy has the advantage to be suitable for the implementation in some
of two layers of mortar matrix held together by an inelastic interface and commercial codes, like DIANA FEA, here considered, and therefore can
connected to the arch through another interface. This way, both the bring immediate advantages for improving the accuracy of calculations
matrix-fibers delamination and debonding at the composite-substrate in practical applications concerning FRCM-reinforced masonry struc
can be described. Moreover, for the matrix, a smeared cracking model tures. Moreover, it allows determining the main mechanisms inducing
was employed to consider the occurrence of cracks. Schemes of the the failure of a reinforced arch and the material models to adopt for
aforementioned numerical approaches for FRCM reinforced masonry simulating its response.
arches and vaults are reported in Fig. 2. Moreover, in [31] the influence
of the values of mechanical parameters on the limit horizontal load for
masonry arch bridges reinforced at the intrados by FRCM composites is 3.1. Modelling approach
discussed. Finally, in [32] also the influence of settlements is studied.
The idea is to adopt a simplified micro-modelling approach for the
Fig. 2. Numerical approaches in literature for FRCM reinforced arches: (a) [29], (b) [33], (c) [30].
3
substrate to describe also the sliding of the blocks, and of considering the through the whole cross-section parallel to the × axis, as reported in
composite as a continuum (representing the matrix) reinforced with bars Fig. 3d. The S2 cross-section was located at the center of the sample;
representing the textile, and having the same cross-section and elastic anyway, numerical experiments performed show that moving the posi
modulus of the textile. No relative displacements are considered be tion of S2 provided that there is sufficient distance from the basis does
tween the textile and the matrix. The bond of the FRCM strip to the not influence the results.
curved structure (from now on termed as “vault”) is reproduced by A 2D plane stress model is adopted to investigate the propagation
interface elements. process of the damage in the cementitious matrix and the effect of the
Although the FRCM components (matrix and textile) are modelled fiber mesh presence inside the composite under axial tension. The
separately, the assumption of coupling between the matrix and the sample is supposed to be fixed at one end and a prescribed displacement
embedded bar prevents one of the most complex ingredients of micro- is imposed at the opposite end to force the composite to stretch up to
modelling approaches, the representation of the matrix-fiber in 0.02 mm. Two modelling approaches are used and compared: a macro-
terfaces. Therefore, the adopted approach for the composite cannot be modelling approach (Fig. 3a), with the fiber mesh modelled as a bar
strictly defined as a micro-model. Indeed, even though the slippage of embedded in a continuum characterized by the mechanical properties of
the fibers inside the matrix cannot be directly reproduced, the effects of the composite, and a micro-modelling approach (Fig. 3b), with the
debonding phenomena occurring at the matrix-fiber interface on the composite modelled as two layers of matrix bonded through interfaces
overall structural behaviour can be still indirectly taken into account. As to the reinforcement bars, which represent the fiber mesh. Indeed, here
an example, consider the FRCM composite beam element shown in Fig. 3 the goal is not that of validating the modeling strategy concerning the
subjected to axial tensile load. In the configuration adopted for deter results of an experimental tensile test, but that of comparing the two
mining the tensile capacity of the FRCM, evaluated as described in [23], numerical strategies.
a test set-up with clamping-grip configuration is employed [34,35]. In both the macro-model and the micro-model, the matrix is
Assuming the presence of initial localized damage, as depicted in Fig. 3a, modelled with the Total Strain Rotating Crack model assuming the
the fracture propagates towards the reinforcement bar and localizes mechanical properties listed in Table 6 (see Section 3.2 for more details).
around it when the beam is axially elongated. Cracks appear in the The bar is linear elastic with a Young modulus equal to 78900 MPa. In
matrix generally where the reinforcement hinders the opening of hinges. the macro-model no bond-slip relation is considered to model the con
The cracks progressively propagate towards the reinforcing fiber net and tact at the matrix-bar interface, while in the micro-model the matrix is
localize around it. Therefore, debonding phenomena at the matrix-fiber connected to the reinforcement bar using two non-linear interfaces
net interface can take place. Indeed, this kind of mechanism happens in modelled using the Discrete Cracking model. The mechanical parame
FRCM composites applied on vaults. ters assumed for the matrix-bar (MB) interfaces are listed in Table 1.
From the numerical point of view, if the FRCM composite is modelled Fig. 4 shows the Cauchy total stresses recorded in the simulations for
as a continuum embedding a bar, using a smeared cracking model for the a point of the matrix (P1) and a point of the reinforcement bar (P2) for
matrix, it can be shown that assuming a perfect bond between the bar both models. It is easily seen that these results are practically
and the matrix, or introducing matrix-fiber grid interfaces, should pro
vide comparable results in terms of stresses, displacements and failure
Table 1
mode. To this aim, a sample of FRCM composite 60 mm × 30 mm × 10
Matrix-Bar (MB) interface mechanical parameters.
mm made of a lime mortar with a glass fiber mesh embedded is
considered. To reproduce the presence of localized fracture, the cross- MB Normal Shear Tensile Tensile Reduced
stiffness stiffness strength fracture shear
section labelled S2 (Fig. 3d) is characterized by a reduced area (150
energy modulus
mm2) with respect to any other typical cross-section located at the dis
kn [N/ ks [N/ ft [N/mm2] Gft [N/mm] GMB [N/mm2]
tance z from the reference cross-section, like S1 in Fig. 3c, having the
mm3] mm3]
area 300 mm2. The fractures are positioned symmetrically with respect 5000.0 2000.0 3 1.5 1
to the z axis and are 2.5 mm in length (along the y axis). They run
Fig. 3. a) Macro-modelling approach; b) micro-modelling approach; c) transversal section S1 of the undamaged solid; d) damaged transversal section of the solid S2.
Units: mm.
4
Fig. 4. Cauchy total stresses vs displacement of the reinforcement bar and the matrix for both models.
superimposed. As the load gradually increases, the stresses in the matrix 3.2. Simplified micro-modelling for the masonry substrate
and the bar also increase and cracks appear in the matrix in the prox
imity of the ending tip of the fracture on both sides of the sample. When In the framework of simplified micro-modelling approaches, the
the axial displacement reaches the value of 0.01 mm, the cracks have mortar joint is not represented with its actual thickness (see Fig. 6), but
already developed in the matrix up to the reinforcement bar, where the it is considered as a zero thickness interface where cracks can potentially
stresses suddenly increase, while the stresses in the matrix slightly occur and open [36]. The benefit of this simplification in terms of
decrease and then remain almost constant, due to the stress transfer computational costs is high since the number of interfaces needed is
from the matrix to the bar. Fig. 5a and 5b show the distribution of cracks halved.
at the final step of the calculations, for the prescribed displacement For unreinforced masonry structures, bricks are often modelled as
equal to 0.019 mm. linear elastic blocks since the stresses hardly ever overcome the
It is worth noting that some cracks in the macro-model run parallel to compressive nor the tensile strength, and the collapse of the structure is
the bar (see the zoom of Fig. 5a), revealing that, as a consequence of the
propagation of the fracture from the external layer of the composite
toward the bar, cracks can occur at the bar-matrix contact surface and
debonding phenomena can take place. A similar result is obtained for the
micro-model (see Fig. 5b), where rotated cracks appear near the rein
forcement bar and shear relative displacements increase at the bar-
matrix interface. Thus, it is possible to claim that the results provided
by the two models are well comparable and that the assumption of a
perfect bond between the bar and the matrix does not fail in indirectly
reproducing also the debonding phenomena, which may occur in FRCM
composites at the fiber mesh-matrix interface, possible especially when
the reinforcement is applied at the extrados.
In conclusion, this modelling choice combines the use of mechanical
parameters that can be reasonably determined by practitioners in real
applications with the capability of taking into account both the loss of
bond at the composite-substrate (directly) and the loss of bond at the
matrix-textile (indirectly).
Fig. 6. Modelling of the reinforced vault.
Fig. 5. Cracks distribution for the macro- model (a) and the micro-model (b).
5
mainly due to the opening of fractures at the joints. Since the rein the sum of the elastic deformation and of a contribution which takes into
forcement can yield substantially higher stresses, here bricks are account non-elastic effects during unloading of the material adjacent to
modelled by using a smeared cracking damage model, suitable to the crack surfaces, and δ0 is the crack opening for which stress can no
describe distributed crack patterns, when the specific location of frac longer be transferred. The crack opening δ0 is determined starting from
tures cannot be predicted, as for the case of crushing or tensile failure of the interface tensile strength ft and the mode I fracture energyGIft .
bricks. In particular, the Total Strain Rotating Crack model [37] is The shear stress τ is supposed to reduce after cracking. Denoting by w
adopted. the crack width and s the crack slip, it is assumed that.
Following [38], the response in compression is represented by a ⎧ σ
parabolic constitutive law, governed by the compressive strength fc and ⎪
⎨ τ = kt s
⎪ for w ≤
kn
by the compressive fracture energy Gfc. The response in traction is (5)
⎪ σ
assumed linear elastic up to the tensile strength ft; by further increasing ⎪
⎩ τ = βkt s for w > ,
kn
the tensile strains, an exponential softening phase follows [39], ruled by
the following expression: with β the shear retention factor varying between 0 and 1. In
( ) ( ) particular, for β = 1 the shear response after cracking is still described by
σ cr cr
nm εnm εcr (5.1), without any reduction of the shear stresses. On the contrary, for β
= exp − nm
(1)
ft εcr
nm,ult = 0, no shear stresses are transmitted after cracking, as for the case of
( cr ) smooth crack surfaces. When β assumes a value between 0 and 1, a
nm εnm is the crack stress, εnm is the crack strain and εnm,ult is
where σ cr cr cr
reduced shear modulus is assumed after cracking, and the shear stresses
the ultimate crack strain (see Fig. 7a). never go to zero describing the effect of the sliding friction between the
Eight-node square plane stress elements (CQ16M) are used (Fig. 7b) crack surfaces that ensures stress transferring residual capacity to the
for the masonry units. interface.
The Discrete Cracking model is adopted to describe the brick-brick Six-node interface elements, labelled CL12I, are used in the adopted
(BB) interface mechanical behaviour. This interface model is based on mesh for describing BB interfaces (see Fig. 8b-c).
the total deformation theory where initiation, Mode-I behaviour and
Mode-II behaviour are independently specified (uncoupled modes). [41]
An exponential stress–strain law is assumed for modelling the 3.3. FRCM modelling
decreasing of the stress as the cracks opening grows when the material is
subjected to tractions. This stress–strain law is well suited also for The strengthening FRCM composite, made of a cementitious mortar
masonry-like materials since an exponential behaviour is reported in reinforced by short fibers and embedding a fiber net, is described as a
[40] for the cracks opening in mortar joints of masonry panels in continuum reinforced by suitable bar elements representing the fiber
traction. net. The primary goal is to reproduce the overall effects of the cracking
The Discrete Cracking model is ruled, in the elastic phase, by the of the matrix and the tensile strength provided by the combination of the
normal stiffness (kn) and tangent stiffness (kt) with respect to the matrix and the fiber net. Therefore, the same smeared cracking damage
interface plane according to the following relation: model employed for the brick, the Total Strain Rotating Crack model, is
[ ] considered here for reproducing the nucleation and evolution of cracks
t= n
k 0
Δu, (2) in the continuum (the matrix). The tensile softening behaviour is
0 kt described by the JSCE model [42], considering a stress plateau after
cracking, followed by a softening phase ruled by the equation
where t={σ, τ}T is the traction vector and Δu is the vector collecting
(ε )c
the relative interface displacement. The tensile softening of the interface
σ = ft tu , (6)
response (see Fig. 8a) is described by an exponential function depending ε
on the interface tensile strength and the tensile fracture energy [41]: with σ the tensile stress, ε the total tensile strain, ft the tensile
σ δ strength, εtu the tensile strain corresponding to the end of the plateau,
= f (δ) − f (δ0 ), (3) and c an exponent, usually set as 0.4 for unreinforced concrete and as 0.2
ft δ0
for reinforced concrete elements. It has been shown in [14] that the JSCE
where model is capable to simulate the response of FRCM composites in trac
[ ( )3 ] ( ) ( ) tion; in fact, the plateau before the softening phase reproduces the
δ δ δ
f (δ) = 1 + 3 exp − 6.93 − 28 exp( − 6.93), (4) tension stiffening effect observed in the matrix of FRCM composites
δ0 δ0 δ0
during the crack propagation phase, when cracks develop in the matrix
σ is the stress at the interface, ft is the interface tensile strength, δ is and the load is progressively transferred from the matrix to the fiber
the crack opening, obtained by subtracting from the total deformation mesh. Indeed, eq. (6) represents a residual tensile strength and a
smoother reduction of the tensile stresses, typical of FRCM composites, if
compared to other softening models like eq. (1), more suitable for un
reinforced masonry. For the applications of interest of the present paper,
the description of the FRCM matrix behaviour in compression has no
practical interest: therefore, a simple linear elastic response can be
considered.
To model the mechanical response of the FRCM, a special element
implemented into DIANA FEA is adopted. It is an eight-node plane stress
element with reinforcement bars embedded inside. Since the displace
ments and the strains of the bars and the continuum elements are fully
coupled, it results in a contribution of the bars in terms of stiffness,
tensile strength and ductility of the continuum. Moreover, the bar ele
ments play a leading role in withstanding the tensile stresses transmitted
Fig. 7. Total Strain Rotating Crack model constitutive law: exponential soft by the vault to the composite when the matrix cracks.
ening in traction and parabolic response in compression (a) and CQ16M The input data for reinforcement bars comprise the Young modulus
element (c). (Ef), the stress–strain law under uniaxial traction and the geometrical
6
Fig. 8. Tensile behaviour for interface elements according to [41] (a). 3 + 3 nodes interface element: displacements (b) and stresses (c).
properties, i.e., the cross-section area. means of a hydraulic jack connected to a load cell with a maximum
capacity of 100 kN. Four loading and unloading cycles were performed
3.4. Brick-FRCM interface modelling on the FRCM retrofitted vault, the first two useful for the settling of the
experimental setup and the further two for investigating the beginning
The interaction between the FRCM composite and the substrate is of the non-linear phase of the vault response. In Fig. 10 the
described by an interface, labelled CS (composite-substrate), using again load–displacement experimental curves for UV and RV are reported.
the Discrete Cracking model described in Section 3.1, also suitable for The above described experimental provision concerning the load
reproducing the detachment of the composite from the substrate, which application and the execution of load cycles have not been reproduced in
might occur due to the debonding at the composite-substrate interface. numerical simulations because of the complexity in accounting for the
contact surface imperfections in the numerical model. Therefore, nu
4. Case study and determination of the mechanical parameters merical analyses for both UV and RV are numerically performed under
displacement control according to a monotonically increasing
4.1. Reference case-study displacement history.
The unstrengthened vault (UV) collapsed as a four-hinge mechanism.
The proposed numerical approach is applied for simulating the According to [23], the first hinge (H1) appeared at the extrados, under
experimental behaviour observed in an in-situ test of a masonry barrel the load application point. The second hinge (H2) opened at the intrados
vault retrofitted at the extrados with glass FRCM composites. The results at a symmetric position with respect to the first hinge; the third and the
of the experimental tests are presented in [23], where the load-carrying fourth hinges (H3, H4) occurred at the left and right abutments,
capacity of the reinforced vault is compared to that of an unstrengthened respectively (see the scheme in Fig. 10). No information is provided
vault having the same geometrical and mechanical features. For con about the load levels corresponding to the opening of the hinges.
venience of the reader, the main experimental results in [23] are briefly The collapse mechanism of the reinforced vault (RV) was charac
summarized. terized by the formation of one hinge (H1) under the load application
The examined structure is a barrel vault characterized by the area for about 1800 N, while the second hinge (H2) occurred at about 3/
geometrical data reported in Table 2 and subjected to a concentrated 4 of the vault span, at a load of 2100 N and at a symmetric position
load at a distance of 556 mm from the left abutment (approximately-one concerning H1. The occurrence of H2 was revealed by the cracking of the
fourth of the span). composite that prevented the complete opening of the hinge. When the
In Fig. 9a and b, the geometry of the model is reported along with the third and the fourth hinges formed at the abutments, the vault collapsed
position of the hinges opened during the in-situ test of the FRCM rein due to slippage phenomena at the fiber–matrix interface and the
forced vault. Notice that the hinges corresponding to angles α1 and α4 debonding of the reinforcement at the left support (see schemes in
are not located at the impost of the vault, as one would expect. This Fig. 10). The third and the fourth hinges formed after the peak load.
discrepancy can be explained by Fig. 1(c) of [23]. This figure shows that
the cuts made on the existing vault for obtaining the part to be rein 5. Reference mechanical parameters
forced by FRCM, and then to be tested, do not reach the two timber
beams constituting the imposts. Therefore, the actual span of the tested The mechanical characterization of the masonry and the glass FRCM
FRCM reinforced vault is reduced with respect to the internal span of reinforcement was carried out by means of experimental tests on the
2555 mm between the two timber beams (see also Fig. 2(b) of [23]). In bricks, the mortar, the cementitious matrix, the fibers and the whole
particular, by considering the actual position of the hinges at the ex reinforcement system. The main experimental data are listed in Table 3,
tremity of the vault, an actual span of 1934 mm has to be considered for where the subscripts have the following meanings: “b” stands for
calculations (see Fig. 9c). “brick”, “mj” for “mortar joint”, “f” for “fiber”, “lm” for “lime mortar”;
Due to the small thickness of the vaults, the test on the unstrength moreover, “c” stands for “compressive”, “t” for “tensile”, “s for “shear”,
ened structure (from here on indicated as “UV”) was performed under “d” for “debonding”, and are used for the characterization of the me
load control by sequentially putting sandbags on the extrados of the chanical parameters governing the numerical model, as specified below.
vault that weighed 4–7 kg each, up to the collapse. The reinforced vault It is worth noting that the test reported in [23] was performed on a
(from here on indicated as “RV”) was tested in displacement control by masonry specimen cut from the unreinforced vault; thus, it is reasonable
Table 2
Geometrical data for the vault.
Center Thickness [mm] Width [mm] Span [mm] Rise [mm] Mid-line radius [mm]
C(0, − 1164.98) 60 300 2555 461 2025
7
Fig. 9. a) Tested vault. b) Hinges position for the in-situ tested FRCM reinforced vault. c) Numerical model of the tested FRCM reinforced vault.
Fig. 10. Load-displacement curves of the unstrengthened (UV) and strengthened (RV) vault. Schemes of the collapse mechanisms [23].
8
Table 3 average Young’s modulus of the bricks Eb = 2016 MPa, the average
Experimental data on the mechanical properties of materials [23]. Young’s modulus of the mortar obtained from a test on a masonry
Test n. Data specimen cut from the unreinforced vault Emj = 321.5 MPa, and the
sample average thickness of the joint hmj = 10 mm. Therefore, eq. (7) yields a
Compressive test on 8 Young’s modulus Eb 2016 normal stiffness of the BB interface kBBn = 38.25 N/mm3. Assuming the
bricks MPa tangent stiffness kBBt of the BB interface to be 40 % of kBBn, one obtains
Compressive strength fbc 10.7 kBBt = 15.3 N/mm3.
MPa Notice that the pictures reported in [23] show that the thickness of
Brazilian test on bricks n.a. Tensile strength fbt 0.8 MPa
Compressive test on 1 Mortar Young’s modulus Emj 306/
the joints is very variable along the vault; this could affect somewhat the
masonry 337 reliability of the obtained values of kBBn and kBBt. Nevertheless, in
MPa absence of direct experimental results on the stiffness of the joints, the
Compressive strength fmjc 3.5 MPa value determined by eq. (7) is considered.
Tensile test on dry glass 5 Weight per unit area 223.4
ρf
To verify the reliability of the above reported values of kBBn and kBBt,
fibers kg/m2
Young’s modulus Ef 75.43 a numerical simulation of the load–displacement behaviour of the un
GPa reinforced vault is performed (see the blue continuous curve in Fig. 12,
Tensile strength fft 1.442 labelled UV). In the same figure, the experimental load–displacement
GPa curve is the red continuous line. The upper bound for the collapse load
Three-point bending test n.a. Young’s modulus Elm 6080
on lime mortar for MPa
obtained by the kinematic theorem of Limit Analysis with reference to
FRCM composite Compressive strength flmc 7.48 the actual hinges position at the collapse, see Fig. 8, is also reported
MPa (dashed black line, corresponding to 254 N). The considered values of
Flexural strength flmt, 3.16 kBBn and kBBt give a response comparable with the experimental results
MPa
b
[23], and compatible with Limit Analysis results, since the latter are
Pull-off test on FRCM 3 Maximum bond stress
composite Reinforcement applied σbd 0.17 obtained neglecting the tensile strength of joints [46].
to the brick MPa The interface tensile strength, ft, which rules the onset of the inter
Reinforcement applied σbd face opening, does not correspond to the tensile strength of the mortar
to both brick and mortar 0.05 and depends on the bond between the bricks and the mortar joint, which
joint MPa
is generally lower. In absence of experimental results, in [47] the tensile
Shear test on FRCM 5 Debonding stress σsd 335
composite MPa strength fmjt = 0.02 MPa for the BB interface has been determined for the
Tensile test on FRCM n.a. Trilinear stress–strain (see Fig. 11) same UV under investigation through a large number of numerical
composite curve calibration experiments. Therefore, the above value was adopted for the
interface tensile strength.
Experimental data for the tensile fracture energy Gmjft are not
to assume that the cutting processing of the specimen from the structure
available. It is experimentally observable that the tensile fracture energy
may have induced a disturbance in the material.
tends to increase when the tensile strength increases. Therefore, the
The composite tensile properties plotted in Fig. 10 are the schema
tensile fracture energy Gmjft = 0.012 Nmm/mm2 suggested in [48] is
tization of the results of tensile tests performed by [23] using the
considered.
clamping-grip configuration reported in [34,35]. In the following Sub
Finally, as explained in Section 2.1, assigning a value of the reduced
sections, the results of the experimental tests are employed for charac
shear modulus corresponds to providing a value of the shear retention
terizing the constitutive laws used in the proposed model strategy. When
factor β, and therefore assuming that the shear stress between the blocks
experimental results are not available, suitable literature values or for
never vanishes. In absence of specific experimental results, it was
mulations are considered.
deemed appropriate not relying on the residual capacity of transmitting
shear stresses after cracking, thus adopting a very low value of the
5.1. BB interface reduced shear modulus GBB = 0.01 MPa. This value is close to zero, but
not zero for avoiding numerical problems. In summary, the values of the
According to the Discrete Cracking model, the mechanical parame BB interface mechanical parameters considered for the computations are
ters required to define the behaviour of the BB interface are the normal reported in Table 4.
stiffness kBBn, the tangent stiffness kBBt, the tensile strength fmjt, the
tensile fracture energy Gmjft and the reduced shear modulus GBB.
The normal and tangent stiffness cannot be determined from the 5.2. CS interface
experimental tests usually performed for the mechanical characteriza
tion of materials. In [36] a relation is proposed for the definition of the The CS interface describes the contact interactions between the
normal stiffness of BB interfaces as a function of the Young’s moduli of substrate bricks and the cementitious matrix. The mechanical parame
the bricks and the mortar of the joints, Eb and Emj respectively, and of the ters required for characterizing the Discrete Cracking model for CS
thickness hmj of the mortar joint: interface are the normal stiffness kCSn, the tangent stiffness kCSt, the
tensile strength fit, the tensile fracture energy Gift and the reduced shear
EE modulus GCS.
kBBn = ( b mj ) (7)
hmj Eb − Emj For what concerns the normal stiffness kCSn, no experimental data are
Notice that according to [43] eq. (7) gives accurate results only if the
Young’s modulus of the brick is sufficiently higher than that of the Table 4
mortar, otherwise an unrealistically high normal stiffness is obtained. A BB interface mechanical parameters.
similar formula is also proposed in [36] for the determination of the BB Normal Tangent Tensile Tensile Reduced
tangent stiffness kBBt as a function of the shear modulus of the bricks and stiffness stiffness strength fracture shear
energy modulus
mortar, Gb and Gmj, respectively. When the latter data are not available,
kBBt can be assumed equal to about 40 % of the normal stiffness kBBn, as kBBn [N/ kBBt [N/ fmjt [MPa] Gmjft [N/mm] GBB [MPa]
suggested in the literature [44,45]. mm3] mm3]
38.25 15.3 0.02 0.012 0.01
In the present case, the reference paper [23] reports (see Table 3) the
9
available. Anyway, since the matrix is characterized by a higher Young’s employed FRCM is assumed, while the mass density of the composite ρc
modulus with respect to the mortar used for the joints and to the bricks is determined starting from the mass density of the fibers and the matrix
(see Table 3), the CS normal stiffness is expected to be higher than the BB and taking into account the mass percentage of each component in the
normal stiffness. Applying eq. (7) to the matrix-substrate system and composite.
considering the average Young’s moduli of the bricks Eb = 2016 MPa, The tensile strength of the composite fct does not coincide with that of
the Young’s modulus of the lime mortar Elm = 6080 MPa, and a thickness the mortar constituting the cementitious matrix since it accounts for the
hi = 10 mm for the composite layer, a rounded value kCSn = 300 N/mm3 mutual collaboration between the matrix and the fiber net. Anyway,
is obtained. Again, the tangent stiffness is conventionally fixed in the 40 these two values are quite similar. Notice that also the Young’s modulus
% of kCSn, resulting kCSt = 120 N/mm3. of the FRCM composite Ec = 6500 MPa is slightly different from that of
Regarding the tensile strength of the CS interface fit, in [23] the bond the matrix Ematrix = 6080 MPa. In Table 4 of [23] are reported data
strength σ bd between the reinforcement and the substrate has been related to the tensile behaviour of the FRCM system under consider
experimentally evaluated by means of pull-off tests on 3 masonry ation, represented in Fig. 10. In particular, the stress over the whole
specimens extracted from the vault after the collapse of the structure, section of the reinforcement (matrix and fiber net) at the end of the so-
with reference to two different configurations. In the first configuration, called phase I is indicated. Starting from this value, the tensile strength
providing σbd = 0.17 MPa, the reinforcement was applied only on the of the composite fct = 2.08 MPa has been determined in [23].
brick, while in the second configuration the reinforcement was applied The JSCE tension stiffening model requires the definition of the
both on the brick and the mortar joint. In the latter case, which provides tensile strain corresponding to the end of the plateau, εcu, and the power
a much more realistic situation, a far lower bond strength σ bd = 0.05 parameter c. For the latter, experimental results for FRCM composites
MPa has been obtained (see Table 3), being here assumed for the tensile are well represented choosing c in the range 0.4 ÷ 0.8. In particular, the
strength of the CS interface. former value is suggested in [14] for a fitting experimental response;
The tensile fracture energy Gift of the CS influences the global therefore, c = 0.4 is assumed. Since the experimental behaviour shown
behaviour of the composite with respect to debonding phenomena and, in Fig. 10 indicates the absence of an appreciable stress plateau, the
thus, the post-peak behaviour of the structure. In particular, a low value value εcu = 0.00032 representing the deformation at the end of the
of Gift compared to the value assumed for fit leads to a brittle behaviour elastic range is considered.
of the interface and, consequently, to a sudden increase in the relative No data are available regarding the compressive strength of the
displacements of the interface. Here Gift is evaluated for a value of the FRCM; therefore, the compressive strength of the lime mortar used for
ductility index diu = Gift/fit = 0.145 mm, determined according to the the matrix fcc = 7.48 MPa is considered. Similarly, no experimental
indication for composite-masonry interfaces reported in [49]. This way, values are available for the compressive fracture energy Gcc. Thus, Gcc is
the value Gift = 0.0072 N/mm is obtained. estimated by following the Model Code 90 indications for concrete [50].
Finally, for the same arguments reported in Section 4.3 for BB In particular, if fcc < 12 MPa, Gcc can be determined assuming a ductility
interface, a very low value of the reduced shear modulus GCS = 0.01 MPa factor dcc = Gcc/fcc = 1.6 mm, thus obtaining Gcc = 11.97 N/mm. In
is considered. In summary, the values of the CS interface mechanical conclusion, the mechanical parameters considered for characterizing
parameters reported in Table 5 are considered. the Total Strain Rotating Crack model adopted for the FRCM composite
are summarized in Table 6.
5.3. FRCM composite 5.4. Masonry bricks
The FRCM matrix is modelled as a continuum and the embedded The mechanical tests on the masonry carried out in [23] provide
fiber net is represented by bar elements with no degree of freedom on most of the parameters required for the numerical model (see Table 3).
their own. Therefore, even if the FRCM components are distinctly rep In particular for the bricks, the Young’s modulus Eb = 2016 MPa and the
resented, the fibers and the matrix behave as a single element having compressive strength fbc = 10.7 MPa, along with the tensile strength
mechanical properties representative of the whole composite. Therefore, obtained by Brazilian tests fbt = 0.8 MPa, are considered. The Poisson’s
the behaviour of the embedded bars corresponds to that experimentally ratio is set νm = 0.2, a typical value for clay bricks [51]. The mass density
observed for the composite regarding the third phase of the trilinear of the clay bricks is considered 1800 kg/m3.
stress–strain relation in [23] (see also Fig. 11). In particular, the bar is For the determination of the compressive fracture energy, the Model
modelled as linear up to failure, having the stress–strain relation shown Code 90 recommendations can be used; in particular, for fbc < 12 N/
by the blue dashed curve in Fig. 11, with Young modulus equal to that of mm2 a compressive fracture energy Gbc = 17.12 N/mm is obtained. For
the composite in the third phase of the average trilinear stress–strain what concerns the tensile fracture energy Gbft, as reported in [52] in case
curve (blue continuous curve in Fig. 11; Ef = 78900 MPa). The ultimate of clay bricks and the absence of specific information, the average value
tensile stress is identified with the average ultimate stress reached by the 0.029 mm can be assumed for the ductility index dbu = Gbft/fbt. There
composite in tension (σ f = 595 MPa). Notice that in Table 4 of [23] both fore, the value Gbft = 0.0232 N/mm is assumed. All the mechanical
data on the Young modulus and the tensile strength are referred to the parameters considered for the masonry bricks are listed in Table 7.
cross-section of the fiber grid.
In Table 6 the mechanical properties considered for the Total Strain 6. Main case study results
Rotating Crack model describing the behaviour of the FRCM composite
are listed. In particular, the Young’s modulus Ec was experimentally Fig. 13 shows the main results in terms of the load–displacement
determined in [23]. For the Poisson’s ratio νc, a typical value for the curves; the experimental results are plotted with a dashed line and the
numerical results obtained by the proposed model strategy are repre
Table 5 sented with a continuous line (blue lines for UV and red lines for RV).
CS interface mechanical parameters. The numerical simulations of the strengthened vault (RV) yields results
CS Normal Tangent Tensile Tensile Reduced consistent to the experimental ones both in terms of collapse load and
stiffness stiffness strength fracture shear displacement at the collapse, intended as the displacements at the peak
energy modulus
load, see also Table 8. Moreover, the proposed modelling strategy gives
kCSn [N/ kCSt [N/ fit [MPa] Gift [N/mm] GCS [MPa] representative results also for the case of the unreinforced vault (UV),
mm3] mm3] although the displacement at the collapse is quite different.
300.0 120.0 0.05 0.0072 0.01
As recalled in Section 3, the numerical model is composed of 928 8-
10
Table 6
FRCM composite mechanical parameters (Total Strain Rotating Crack model).
Young’s Poisson’s Mass Tensile Tensile strain at the end of the Power Compressive Compressive fracture
modulus ratio density strength plateau exponent strength energy
Ec [GPa] νc [-] ρc [kg/m3] fct [MPa] εcu [-] c [-] fcc [MPa] Gcc [N/mm]
6500 0.2 1900 2.08 0.00032 0.4 7.48 11.97
experimental tests on the materials or determined according to literature

formulations were considered for the calculations. No calibrations of the
parameters in view of reproducing the (known) experimental response
of the structure have been made.
6.1. Reinforced vault (RV)
The numerical behaviour of the reinforced vault is practically linear

up to a load of about 1300 N, slightly lower than the value at which the
formation of the first hinge has been experimentally detected. Then, as
the load increases, a slight loss of stiffness occurs because of the gradual
opening of the first hinge located under the loading application point
(H1 in Fig. 14), and of the appearance of the first crack at the abutments.
After the peak (2204 N), the load suddenly decreases to 2060 N, cor
responding to a 3.4 mm displacement of the load application point. At
the same time, a sudden increase in the tangential relative displacement
Fig. 11. Tensile test on FRCM composite [23]. for the CS interface at the left abutment is observed. Then, the load
gradually decreases as a progressive debonding of the reinforcement
from the left abutment occurs (H3 in Fig. 14), and more cracks appear in
the matrix in the centre-right of the vault, where the reinforcement
hinders the opening of the hinge H2 (see Fig. 14). This behaviour is
comparable to that observed in the experiments. Also, the softening
branch of the load–displacement curve obtained from the numerical
analyses follows with a good approximation of the experimental curve.
In the sequence, the reinforcement in correspondence of H2
completely cracked and the debonding at the left abutment extended to
all the abutment length, causing the detachment of the FRCM strip from
the left abutment. In this phase, a displacement of about 7.5 mm is
obtained, and the load tends to remain almost constant as the
displacement increases.
At the end, the reinforcement starts to detach also from the right
abutment (hinge H4 in Fig. 14) because of the debonding at the CS
Fig. 12. Experimental and numerical load–displacement curves for the unre interface. In this last phase, the numerical curve is slightly more sus
inforced vault, compared to the limit load evaluated by kinematic theorem of tained than the experimental one, but differences are not relevant.
Limit Analysis. Table 8 reports the numerically and experimentally collapse load
(peak load) for the reinforced vault along with the corresponding
node plane stress elements of 15x10 mm2 and 276 6-node interface el displacement of the load application point. It is seen that the numeri
ements of 15x15 mm2 (termed as mesh 1). Further refinements of the cally estimated collapse load is practically coincident with the experi
mesh yield a very limited variation of the numerical results despite mental one (the difference is of about 0.2 %).
higher computational costs. As a representative example of the sensi For what concerns the displacement of the load application point at
tivity analyses performed, Fig. 12 shows also the results obtained by the collapse, it is worth recalling that the experimental testing consid
considering 8-node 7.5 mm × 3.3 mm plane stress elements for the ered four loading–unloading cycles, two in the elastic range and two out
matrix and 8-node 7.5 mm × 7.5 mm plane stress elements for the bricks of the elastic range. In particular, the last cycle has been performed until
for a total number of 4096 8-node plane stress elements and 552 6-node the appearance of the first hinge, and produced a residual displacement
interface elements (termed as mesh 2). It is seen that the curves corre of 0.65 mm. These loading–unloading cycles are not numerically
sponding to “mesh 1” and “mesh 2” are practically superimposed up to reproduced for the reasons explained in Section 4.
the peak load and modestly differ in the post-peak phase. However, In Table 8, the difference between the experimental and numerical
“mesh 2” required more than twice the time for the analysis. displacement of the load application point at the collapse for the RV is
It is worth recalling that the approach followed here is that of 0.66 mm. This difference corresponds almost exactly with the above
identifying the values of the mechanical parameters by a “blind” pro reported residual displacement experimentally observed in the last
cessing of the available data; only values directly estimated by loading–unloading cycle. Therefore, it is reasonable to think that the
Table 7
Masonry bricks mechanical parameters.
Young’s modulus Poisson’s ratio Mass density Compressive strength Compressive fracture energy Tensile strength Tensile fracture energy
3
Eb [MPa] νb [-] ρb [kg/m ] fbc [MPa] Gbc [N/mm] fbt [MPa] Gbft [N/mm]
2016 0.2 1800 10.7 17.12 0.8 0.0232
11
Fig. 13. Load-displacement curves for the unreinforced (UV) and reinforced vault (RV).
stiffness of the simulated response with respect to the experimental

Table 8 response can be explained since, as recalled in Section 4.2, the elastic
Comparison between the numerical and experimental results.
parameters of the masonry measured in the experiments and then
UVEXP UVNUM RVEXP RVNUM_1 RVNUM_2 assumed in the numerical model correspond to that of a damaged ma
Collapse load Fmax [N] 390 380 2170 2204 2212 terial, because of the way the samples were extracted.
Displacement of the 6.41 1.54 4.08 3.43 3.74 For the CS and the BB interfaces described by the Discrete Cracking
loaded point at the model, the distribution of relative displacements in the direction normal
collapse load δmax [mm]
to the interface cracks are represented in Fig. 14a. The figure clearly
shows that the highest displacement values are concentrated in corre
difference is only because numerical simulations do not reproduce the spondence of the first hinge H1 and at the left abutment, where the
cyclic part. debonding of the composite takes place (H3). The distribution of the
A good agreement between the numerical and the experimental (smeared) cracks in the materials described by the Total Strain Rotating
response curves of the RV in Fig. 13 can be noted, also considering that Crack model, i.e., the bricks and the matrix, can be represented by the
the loading and unloading cycles performed in the experimental tests normal stresses in the local direction across the crack, see Fig. 14b. This
have not been reproduced in numerical simulations. The lower initial figure reveals that the deformation of the vault before the collapse
Fig. 14. Reinforced vault (RV): relative displacements in the direction normal to the interface cracks at the collapse (a); local stresses in the direction normal to the
crack (b).
12
caused the cracking of the composite at a position symmetric with the left abutment, and that the residual strength of the structure is
respect to H1; moreover, FRCM cracks also at the right support and in the substantially due to the fact that the reinforcement prevented the
proximity of the area near H3 and H4. This was also observed during the opening of hinge H2 holding the bricks together. Notice that the cracking
in-situ tests in the position H2 in Fig. 10. Some cracks formed also in the of the matrix starts before reaching the peak load, and evolves up to the
masonry blocks at the abutments. collapse of the vault; therefore, it does not correspond to a particular
In Fig. 15a the load is related to the number of the (smeared) cracks point of the curves in Fig. 15a.
in the bricks and the reinforcement matrix obtained in the numerical The above considerations allow showing that the proposed model is
simulations (red line). The curve is compared to that obtained by plot capable of successfully reproducing the stress-transfer mechanisms be
ting the tensile stress in the cross-section of the bar elements, repre tween the different components of the reinforced structure and its
senting the fibers embedded in the composite, versus the above-defined collapse mechanism.
number of cracks (blue line). The tensile stress is obtained by dividing
the textile forces by the textile cross-section area. In particular, the 6.2. Unreinforced vault (UV)
tensile stress is evaluated in correspondence to a point near the joints
where the hinge H2 in Fig. 14 is supposed to open. In fact, as the matrix Table 8 shows that for UV there is a small difference between the
progressively cracks in that area, the opening of the hinge H2 is hindered numerically estimated and experimentally evaluated collapse load
by the reinforcement, and tangential stresses developing at the CS (intended as the peak load). On the other hand, a quite large difference
interface are transferred to the bar elements through the matrix. between experimental and numerical displacements of the loaded point
Consequently, the tensile stress developing in the bars increases mostly at the peak load is observed. This discrepancy can be likely ascribed to
in this range. Indeed, the tensile stress in the bar elements abruptly in the adopted experimental setup. In fact, experimental tests have been
creases from 25 MPa to 135 MPa at point A, corresponding to a vertical performed in load control that probably motivates the quite irregular
load of 1776 N, and then up to 252 MPa at point B, corresponding to a trend of the experimental response characterized by the lack of the
vertical load of 1787 N. Subsequently, when the load transfer process softening phase. Thus, the experimental peak load is coincident with the
between matrix and fiber net is completed, the axial forces in the bars last recorded load value (correspondingly for the displacement). On the
progressively increase with the number of cracks, towards the final point other hand, the numerical analysis allowed to capture the evolution of
C. Notice that here the tensile stress reach the value of 347 MPa, the softening phase.
consistent with the debonding stress experimentally determined in [23]. About the numerical collapse mechanism for the UV, the typical four-
On the red line in Fig. 15a, red dots mark the load values at which hinge mechanism, also observed during the experiment, has been ob
each hinge opens. In particular, the opening of hinges H1 and H3 is tained. In particular, the first hinge H1 appeared under the load appli
associated with a relative displacement in the direction normal to the cation point, followed by the second hinge H2 at the right abutment and
interface cracks of 0.1 mm for the BB and CS interface. For detecting the the third hinge H3 at the left abutment. The collapse is reached when the
opening of the hinge H2, the load corresponding to the sudden increase fourth hinge H4 opened at a symmetrical position with respect to H1. The
of the axial forces in the reinforcing bar elements is considered. Finally, position of the hinges is visible in Fig. 16, where the relative displace
the formation of the hinge H4 is identified by the cracking of the matrix ments in the direction normal to the interface cracks are displayed in
and by a slight debonding of the reinforcement at the right abutment, correspondence to the collapse load.
which occurred practically simultaneously with the formation of H2. If the numerical collapse mechanism of the UV (see Fig. 15) is
Comparing the diagram with the cracked configurations in Fig. 15b, compared to the experimentally determined one (see Fig. 9), it can be
it is easily seen that the progressive cracking of the composite corre noticed that the position of the hinges perfectly corresponds. However,
sponds to an increase in the axial forces developing in the bar elements. it should be pointed out that the order of opening is slightly different; the
After the maximum load is reached, the number of cracks increases hinge symmetrical to the load application point opens as the last in the
under a constant value of axial force in the reinforcement (before the numerical simulations, while it is the second to open in the experiments.
sudden increase leading to point A in the blue curve in Fig. 15a). This
means that the bond between the composite and the substrate is lost at
Fig. 15. a) Load (red curve) and stress in the fibers (blue curve) vs number of the (smeared) cracks in the bricks and in the reinforcement matrix. b) Distribution of
the (smeared) cracks in the bricks and in the reinforcement matrix in the three different conditions, A, B and C, marked on the blue curve in the diagram on the left.
13
Fig. 16. Unreinforced vault (UV): relative displacements in the direction normal to the interface cracks at the collapse.
7. Sensitivity analysis of the mechanical parameters has been increased or decreased five
times, while all the other parameters are kept constant.
Parametric analyses are now performed to investigate the influence The results of sensitivity analyses are discussed by grouping the
of each mechanical parameter required to describe the mechanical investigated parameters. The discussion is eased using figures reporting
behaviour of the structure (masonry vault and FRCM composite) on the the experimental data (dotted curve), the response obtained with the
global response of the reinforced structure. For each analysis, only one choice of the considered mechanical parameters reported in Section 5
Fig. 17. Influence of variations of BB interface normal stiffness kBBn (a), tangential stiffness kBBt (b), tensile fracture energy Gmjft (c), shear modulus GBB (d).
14
(red curve), the response corresponding to the 5 times increased mechanical properties of the CS interface.
parameter (light blue curve), and the response corresponding to the 5 The obtained load–displacement curves are shown in Fig. 19a -
times decreased parameter (green curve). Possible relevant differences Fig. 20. It is seen that variations of the CS mechanical parameters mostly
in the stress distribution or the collapse mechanism due to the variation influence the post-peak behaviour, related to the debonding of the
of one of the mechanical parameters are highlighted. composite from the substrate at the left abutment.
Fig. 19a shows that the interface normal stiffness kCSn has a small
7.1. BB interface influence on the response before the load peak, and also on the value of
the load peak. After the peak, significant reductions of kCSn do not seem
Fig. 17 and Fig. 18a show the results in terms of load–displacement to influence the response, whereas too high values of kCSn might lead to a
curves obtained from parametric analyses of the influence of the BB post-peak phase much more sustained than that experimentally
interface mechanical parameters on the global behaviour of the rein observed. Moreover, for kCSn = 1500 N/mm3, 5 times higher than the
forced vault. Recall that the reference values are collected in Table 4. reference value, not only the reinforcement prevents the opening of a
The numerical results are mostly affected by the interface normal hinge near the left support, but also no debonding takes place from the
stiffness kBBn (Fig. 17a) and, secondarily, by the interface tensile substrate at the left end of the vault. Indeed, as shown in Fig. 19b, where
strength fmjt (Fig. 18a). Indeed, 5 times increasing or decreasing of the the relative displacements in the direction normal to the interface cracks
interface tangent stiffness kBBt (Fig. 17b), the interface tensile fracture at the collapse are depicted, the debonding of the composite anchored
energy Gmjft (Fig. 17c), or of the interface reduced shear modulus GBB on the left of the vault is observed. The above justifies the fact that for
(Fig. 17d) yield only negligible variations of the post-peak behaviour for kCSn = 1500 N/mm3 the vault deformations increase at an approxi
higher displacements. In particular, lowering kBBn corresponds to a mately constant load.
significantly reduced stiffness of the vault and a considerably lower load On the other hand, the interface tangential stiffness kCSt might in
carrying capacity, whereas increasing kBBn entails a small increase in the fluence more the first part of the response curve than the post-peak
stiffness and the maximum load. This means that an error in excess in the behaviour (see Fig. 20a). Specifically, a very low value of kCSt doesn’t
evaluation of kBBn brings almost negligible effects, but too low kBBn introduce substantial variations in the response, whereas an error in
values may lead to non-representative results. excess in the evaluation of kCSt leads to a stiffer response (the global
The interface tensile strength fmjt could significantly influence the stiffness of the reinforced vault raises from 747 N/mm to 1437 N/mm),
capacity of sustaining the load after the peak, see Fig. 18c. Indeed, although with no noticeable difference in the peak load and the post-
whereas by lowering fmjt concerning the nominal value of 0.02 MPa peak phase. This behaviour is likely related to the fact that kCSt = 600
small variations in the response (only in the immediately post-peak N/mm3, 5 times the nominal value, reaches the same value of the normal
phase) are observed, whereas an error in excess for fmjt could yield a stiffness kCSn.
non-reasonable high post-peak load carrying capacity. Moreover, even large variations of the interface tensile strength fit
For explaining this large influence of fmjt, notice that by assuming a 5 (Fig. 20c) result in a small variation of the peak load. In the post-peak
times higher value of fmjt = 0.1 MPa, the crack stresses configuration at phase, too low values of fit affect the response but in a limited and
the collapse in Fig. 18b reveals that the reinforcement does not scarcely predictable way, whereas too high values of fit might yield to a
completely debond from the substrate. On the contrary, debonding sustained post-peak phase, due to the hindering of the debonding of the
happens for the nominal value of fmjt (see Fig. 14b) and also for the 5 composite at the left abutment, and consequently to the induced delay in
times lower value of fmjt = 0.004 MPa (Fig. 18c). the collapse of the structure.
Errors in the determination of the interface tensile fracture energy
7.2. CS interface Gift yield variations only in the post-peak behaviour (Fig. 20d), inducing
a less or more fragile failure of the CS interface at the left abutment.
The efficiency of the reinforcement in improving the performance of Finally, no appreciable changes are observed by varying 5 times the
the structure in terms of strength and ductility strongly depends on the interface reduced shear modulus GCS (Fig. 20b), since the reference
bonding between the reinforcement FRCM composite and the substrate. value is very low. It is necessary to introduce substantially higher values
This aspect is here investigated by parametric analyses by varying the of GCS (yellow curve) to notice some variation of the response curve in
Fig. 18. Influence of variations of BB interface tensile strength fmjt (a). Distribution of the local stresses in the direction normal to the crack Sknn at the collapse for
fmjt = 0.1 MPa (b) and fmjt = 0.004 MPa (c).
15
Fig. 19. a) Influence of variations of CS interface normal stiffness kCSn b) Relative displacements in the direction normal to the interface cracks DUNy at the collapse.
Fig. 20. Influence of variations of CS interface tangent stiffness kCSt (a), reduced shear modulus GCS (b), tensile strength fit (c) and tensile fracture energy Gift (d).
16
the post-peak phase. by a stress plateau after the appearance of the first cracks, followed by
the exponential softening governed by the exponent c and by the
7.3. FRCM composite deformation at the end of the stress plateau εcu. Possibly, the latter can
be assumed equal to the deformation at the end of the elastic range; in
Here, the results obtained from the sensitivity analyses on the pa this case, no stress plateau is obtained. In particular, Fig. 22b shows the
rameters ruling the mechanical behaviour of the FRCM composite are curves corresponding to the JSCE tension model (6) for different values
presented. The reference values of these parameters are collected in of εcu and fixed exponent c; these curves tend to approach zero stress
Table 6. only for very high values of strain, providing for the residual tensile
The load–displacement curves obtained by assuming the Young’s strength characterizing FRCM composites. The effects of variations in
modulus of the composite matrix Ec equal to 32500 MPa and 1300 MPa, the exponent c on the constitutive response of FRCM in tension are
respectively, are compared in Fig. 21a to the curve obtained for nominal outlined in Fig. 22d for fixed εcu = 0.00032. The parameter c affects the
value Ec = 6500 MPa. Recall that the latter is obtained by averaging the area under the response curve; in particular, higher values of c reduce
results of uniaxial tensile tests performed, referred to the first phase of this area and consequently the tensile fracture energy, leading to a more
the tests, when both the matrix and the textile are subjected to the load. fragile behaviour.
It is easily seen that Ec significantly affects the stiffness of the reinforced Fig. 22a shows the numerical response curves obtained for different
structure and, secondarily, the peak load and the post-peak behaviour. values of εcu. In this case, no lower values than the nominal one (εcu =
In particular, for Ec = 1300 MPa the global stiffness is 40 % reduced and fcc/Ec = 0.00032) are considered, to avoid inconsistency with the value
the peak load decreases from 2204 N to 1995 N, whereas for Ec = 32500 of the Young’s modulus Ec and/or of the compressive strength fcc.
MPa the global stiffness is 30 % increased, but the peak load undergoes a Changes in εcu result in appreciable variations in the response for high
small increment, from 2204 N to 2224 N. Furthermore, it is possible to deformations; in particular, for εcu = 0.0016, 5 times the nominal value,
observe that higher FRCM Young’s modulus yields a more sustained a much more sustained post-peak behaviour is obtained.
response after the peak, whereas lower values of Ec result in a steeper The effects of variations of the exponent c on the global behaviour of
drop of the load after the maximum. the reinforced vault are shown in Fig. 22c. For c equal to 0.2 (light blue
Fig. 21b shows the response curves related to different values of the curve), a value usually considered for reinforced concrete, after the peak
FRCM matrix tensile strength fct. This parameter significantly affects the the load remains almost constant, while for c = 0.8 (green curve) the
numerical results. In particular, for fct = 10.4 MPa (5 times the nominal load–displacement curve is only marginally different from that corre
value), after reaching a load of 2314 N, the composite partially detaches sponding to the nominal value c = 0.4. Anyway, it is worth noting that
from the left abutment causing the load to decrease up to 2080 N. After the curve obtained for c = 0.8 better approximates the experimental
that, the load starts increasing again and no cracks occurred in the curve in the post-peak phase.
matrix. On the contrary, for fct = 0.4 MPa (20 % of the nominal value), As for variations of the compressive strength of the cementitious
the structure reaches a much lower maximum load of 1830 N, and over matrix fcc (see Fig. 23a), for fcc = 37.4 MPa, five times higher than the
600 N the global stiffness is reduced due to the opening of the first hinge nominal value, the maximum load remains the same, but the global
under the load application point. The collapse is due to the development stiffness is almost doubled, while assuming fcc = 1.496 MPa (20 % of the
of cracks all over the matrix and to the partial debonding of the com nominal value), the peak load and the stiffness are slightly lowered.
posite from the left abutment. For the sake of completeness, it has to be Assuming too low values of the compressive fracture energy Gcc does not
noted that for fct = 10.4 MPa, the deformation at the end of the elastic seem to hold a significant influence on the mechanical behaviour of the
range εcu considered in the JSCE model is 0.0016, greater than that reinforced vault. In fact, if Gcc is decreased five times, the
considered in the calculations; therefore, the effects of tensile strength load–displacement curve (green curve in Fig. 23b) is practically super
variations on the structural response of the reinforced vault could not be imposed to that obtained for the nominal value Gcc = 11.97 N/mm. On
distinguished from that of the deformation εcu. the other hand, if Gcc is increased five times, the obtained peak load is
The JSCE model employed for the matrix in traction is characterized higher than the experimental value, and convergence problems occurred
Fig. 21. Influence of variations of FRCM matrix Young’s modulus Ec (a) and FRCM matrix tensile strength fct (b).
17
Fig. 22. Influence of variations of plateau end strain εcu. (a) and of the exponent c (c); JSCE model for different values of εcu (b); JSCE model for different values of
the power parameter c for a fixed value of εcu = 0.00032.
Fig. 23. Influence of variations of the matrix compressive strength fcc and compressive fracture energy Gcc.
18
(light blue curve in Fig. 23b). Fig. 27c shows that also changes in the brick tensile fracture energy
Finally, parametric analyses are performed on the elastic properties Gbft could lead to evident variations in the response curve. Indeed, an
and the tensile strength of the fibers. It is worth recalling that the increase or decrease of the load carrying capacity of the structure of
average tensile stress–strain response experimentally determined in about 30 % when Gbft is respectively increased or decreased five times is
[23] and corresponding to the average curve in Fig. 11 (in blue) is obtained. Thus, it emerges that numerical results are strongly affected
implemented for the bars, with the following reference values: Young’s by the tensile strength and the tensile fracture energy of the substrate
modulus Ef = 78.9 GPa and tensile strength fft = 529.16 MPa. material. On the contrary, the compressive fracture energy Gbc does not
Fig. 24a shows the effect of variations of the Young’s modulus Ef of influence at all the mechanical behaviour of the strengthened structure
the fibers, resulting in relatively small changes in the peak load and the since the three curves obtained for the different considered values of Gbc
post-peak behaviour. Variations of the tensile strength of the fibers fft are completely superimposed; thus, for the sake of brevity, these
(see Fig. 24b) yields even smaller variations of load carrying capacity load–displacement curves have not been reported.
and of the softening branch of the load–displacement curve. Furthermore, it is interesting to observe that an increase in the
brick’s compressive strength fbc affects the global stiffness of the vault. In
fact, the blue curve in Fig. 27b, corresponding to fbc = 53.5 MPa (five
7.4. Masonry bricks times the reference value fbc = 10.7 MPa), is characterized by a higher
stiffness than the reference curve. The latter is very close to the curve
For reinforced masonry arches and vaults, the mechanical properties obtained for the lower value of fbc. No noticeable changes in the peak
defining the substrate inelastic behaviour in tension (tensile strength load or the post-peak behaviour are observed by changing fbc.
and tensile fracture energy) and in compression (compressive strength Finally, Fig. 27d describes the effect of variations of bricks Young’s
and compressive fracture energy) can play an important role in modulus Eb. It is seen that this parameter greatly influences the overall
describing the load capacity of the whole structure. Indeed, the rein stiffness, the peak load and also the post-peak behaviour. Thus, for the
forcement prevents hinge openings at some of the block-joint interfaces numerical model to be representative, the Young’s modulus of the bricks
and the rotation of the blocks; this yields much higher tensile and has to be carefully characterized [53].
compressive stresses in the blocks with respect to what happens for
unreinforced structures. This is illustrated in Fig. 25, where the
maximum principal stresses are plotted for the unreinforced and rein 7.5. Concluding remarks
forced vaults. In particular, the figure displays the detail of the part of
the vault where the hinge labelled H2 opens; the same colour scale for The performed sensitivity analyses suggest that stiffness parameters
stresses is employed (from 0 to 0.8 MPa). In the case of the reinforced (interface stiffness and elastic moduli) can be even more influential than
vault, results for fbt = 0.8 MPa (the nominal value) and for fbt = 4 MPa strength parameters. This is likely because large variations in the stiff
are reported (see Fig. 25b and c). It is evident that the reinforcement ness of one of the components of the structural model could radically
system prevents the opening of joints, and this leads to the development vary the stress transfer arrangements, thus activating or completely
of higher tensile stresses in the bricks. deactivating some collapse mechanisms. On the other hand, large vari
As a consequence, by varying fbt while all the other mechanical pa ations in strength parameters are influent only if the related failure is
rameters being fixed, very different load–displacement curves are ob active, and to the extent to which that failure contributes to the overall
tained, see Fig. 27a. In particular, when the bricks tensile strength is behaviour and to which that parameter has been varied. The highly non-
increased five times (fbt = 4 MPa) the load–displacement curve does not linear behaviour of reinforced masonry arches and vaults and the pos
show any softening phase, and the collapse is obtained for a much higher sibility of several different stress transfer and internal failure mecha
load, due to the cracking of the matrix, no longer able to transfer the nisms render rather unpredictable the effects of variations of more than
stresses from the substrate to the fiber mesh (see also Fig. 26). If the one mechanical parameter, with possible unrealistic numerically simu
reference fbt is reduced five times, a sudden decrease of the stiffness is lated results.
noted after a load value of 1088 N, with a slight change in the peak load The above considerations, however, should not obscure the most
and no appreciable variations in the post-peak phase. important result of Section 7, namely that the proposed numerical
Fig. 24. Influence of variations of the fiber mesh Young’s modulus Ef and the tensile strength fft.
19
Fig. 25. Maximum principal stresses in the bricks of the unreinforced vault (a) and reinforced vault (b/c) corresponding to hinge H2 position (d/e).
Fig. 26. Local stresses in the direction normal to the crack at the collapse for fbt = 4 MPa.
modelling strategy is capable of giving representative results, satisfac pronounced influence on the numerical results. However, the most
torily close to the actual structural response, provided that mechanical influential parameters can be accurately determined by standard me
parameters are reasonably estimated. In particular, the model is quite chanical tests. In particular, this is true for the Young’s modulus of the
robust concerning small errors in the determination of mechanical pa matrix, the composite tensile strength, the composite tensile strain at the
rameters, apart from some of them that can be easily and suitably end of the stress plateau, and the FRCM compressive strength. For what
evaluated by standard mechanical tests. Moreover, the sensitivity ana concerns the exponent c, also influent on the post-peak response, the
lyses indicate those parameters that can be determined by a rough es value suggested by the literature allows for reasonable results.
timate, and those needing to be carefully determined for the accuracy of Finally, the numerically predicted response could be markedly
the simulated behaviour. influenced by large variations of masonry bricks mechanical parameters,
In particular, for the BB interface, a correct estimation of the normal with special reference to the Young’s modulus Eb, the tensile strength fbt
stiffness is required. Since this parameter is difficult to be experimen and the tensile fracture energy Gbft. However, Eb and fbt can be directly
tally characterized, it is possible to determine it by eq. (7). Thus, an determined employing ordinary mechanical tests, whereas the deter
accurate experimental determination of the Young’s moduli of the mination of the tensile fracture energy can be based on the literature.
bricks, the mortar, and the average thickness of masonry joints is Also the masonry bricks compressive strength, affecting the initial
needed. For the tensile strength of the BB interface, only a large over stiffness, can be easily evaluated by standard mechanical tests.
estimating error could secondarily affect the results in the post-peak
phase. Other BB interface parameters scarcely influence the overall 8. Pushover curves
response of the reinforced vault.
A correct determination of the parameters characterizing the CS As discussed in Section 6, the numerical model presented in Sections
interface appears to be a little more influential on the overall response 3-5 gives results quite similar to the experimental ones reported in [23]
reconstruction. In particular, a large overestimation of the normal in terms of load–displacement curve, peak load and displacement of the
stiffness or the tangent stiffness could lead to unrealistic stiffness and application point of the load in correspondence of the peak load. This is
post-peak behaviour. These stiffnesses are very difficult to identify by obtained by using mechanical parameters experimentally determined or
experimental tests; thus, the suggestion is to use eq. (7) and carefully evaluated according to the literature, without any calibration based on
evaluating by experimental tests the Young’s moduli of the bricks and the (unknown in advance for practical applications) experimental
the reinforcing mortar. The tensile strength of the CS interface can be response.
experimentally determined, but large errors influence practically only Thus, this numerical model can be considered representative of the
the post-peak behaviour. The same occurs for the CS tensile fracture actual structural behaviour and can be used to perform other kinds of
energy. structural analyses. In particular, here a pushover analysis aimed at
Some mechanical parameters of the FRCM composite show a determining the maximum seismic capacity in terms of horizontal
20
Fig. 27. Influence of variations of bricks tensile strength fbt (a), compressive strength fbc (b), tensile fracture energy Gbft (c), and Young’s modulus Eb (d).
ground acceleration for the examined vaults is performed, studying the (UV) is compared to that evaluated by applying the kinematic theorem
influence of the FRCM strengthening system [54]. of Limit Analysis. In particular, the latter provided a collapse horizontal
The pushover analysis is carried out by applying first the self-weight load multiplier λk = 1.03, which can be considered consistent with the
and then a system of horizontal forces proportional to the self-weight collapse horizontal load multiplier λ = 1.031 numerically evaluated
monotonically increased from zero in steps of suitable amplitude. One through the pushover analysis.
node at the centre of the keystone is assumed as the control point for The pushover analysis performed for the reinforced vault (RV) yields
determining the capacity curve. a horizontal loads multiplier at the collapse λ = 4.11, about 4 times
In particular, Fig. 28 shows the capacity curves for the unstrength higher than that obtained for the unstrengthened vault. As the capacity
ened (UV) and the reinforced vault (RV) in terms of the base shear force curves in Fig. 28 show, the reinforcement strongly reduces the lateral
normalized to the self-weight of the structure λ versus displacement of displacements of the structure, while increasing the load carrying ca
the selected control point. Notice that λ can be interpreted as the base pacity under horizontal loads. On the other hand, the unreinforced vault
horizontal acceleration in g. Moreover, the deformed configuration of is capable of quite larger horizontal displacements under an almost
the reinforced vault at collapse is shown and the interface relative dis constant load.
placements DUNy are plotted. In Fig. 29 the stresses developing in the reinforcement bar during the
Since for the examined load conditions no experimental results are pushover analysis are plotted and related to the position where each
available, for the validation of the numerical results the maximum mesh element of the bar is positioned along the span; the plotted curves
horizontal load numerically determined for the unstrengthened vault correspond to different values of λ = 1.03, which is the collapse
21
Fig. 28. Capacity curves for the strengthened and unstrengthened vault under horizontal loads; interface relative displacement DUNy at collapse.
Fig. 29. Reinforcement cross section stresses vs position of the reinforcement bar mesh element; cracks normal stresses Sknn at collapse.
multiplier of the unreinforced vault, 2.00, when the first hinge opens, that is in those parts of the vaults where the bond at the composite-
2.80, when the first cracks occur, 4.11 and 3.60 which correspond to the substrate interface weakened. At collapse, the stresses in the textile
maximum value and the last recorded value, respectively. In the same reach their peak and, contextually, the reinforcement detached almost
picture the crack pattern at collapse (λ = 3.60) has been reported and completely from the left abutment, causing the failure of the vault.
scaled to approximately fit the span length.
Starting from a value of λ = 2.00, at almost a quarter of the span, one 9. Conclusions
of the bricks started to slightly separate from the contiguous blocks, as
shown in Fig. 27 in the part of the vault marked with a red circle, and the The present paper proposes a new modelling strategy for masonry
interface relative displacements at the intrados grew from 0.11 mm up curved structures reinforced with FRCM composites, suitable for its use
to 0.35 mm, when the maximum value of λ (4.11) was reached. by practitioners in advanced commercial codes, like the finite element
At λ = 2.8 the first cracks occurred at the left abutment, spreading code DIANA FEA here used.
through the block towards the reinforcement. As the load was The modelling strategy here proposed combines the advantages of
increasing, the relative displacements at the CS interface started to grow both macro- and micro-modelling approaches. In particular, in the frame
both normally to the interfaces and, with less intensity, tangent to the of a macro-modelling approach, the reinforcement is described as a
interfaces. From the curves plotted in Fig. 29 it can be observed that in continuum whose mechanical properties are the ones of the whole
some parts of the composite, the stresses developing in the reinforce composite itself; a smeared cracking constitutive model is considered to
ment bar are very low, while the stresses increase, as the load grow, near reproduce the occurrence of cracks in the matrix in the areas where the
the abutment and in proximity to the part where the first hinge occurred, reinforcement prevents the opening of the hinges in the structure. The
22
presence of the fibers net is represented by means of bars embedded in Protection) and from the Italian Ministry of University and Research
the continuum without interfaces ruling the contact between the two. (MUR) in the framework of Project PRIN2020 #20209F3A37 is grate
All the other interactions between materials are described by a suitable fully acknowledged.
interface model, as it is typical of micro-models. The above approach
brings the advantage of reducing the number of mechanical parameters References
to be determined, along with the number of interfaces elements involved
in the model, with benefit in terms of computational costs. Each of the [1] Wang C, Sarhosis V, Nikitas N. Strengthening/Retrofitting Techniques on
Unreinforced Masonry Structure/Element Subjected to Seismic Loads: A Literature
analyses performed required less than1 h to be performed on a standard Review. The Open Construction and Building Technology Journal 2018;12(1):
notebook. 251–68.
For the validation of the proposed approach, a case study taken from [2] de Felice G, De Santis S, Garmendia L, Ghiassi B, Larrinaga P, Lourenço PB, et al.
Mortar-based systems for externally bonded strengthening of masonry. Materials
the recent literature has been considered. Particular attention has been and Structures/Materiaux et Constructions 2014;47(12):2021–37.
devoted to a suitable choice of the mechanical parameters that could not [3] I. Mascolo, A. Fortunato, C. Olivieri, A. Gesualdo, Seismic retrofitting techniques
be determined by experimental tests, and have to be evaluated indirectly for existing masonry buildings, COMPDYN Proceedings. 2021-June (2021).
https://doi.org/10.7712/120121.8514.19628.
by employing literature formulations. [4] Caggegi C, Carozzi FG, De Santis S, Fabbrocino F, Focacci F, Hojdys Ł, et al.
The comparison between the results obtained from the numerical Experimental analysis on tensile and bond properties of PBO and aramid fabric
model and the available experimental response showed the accuracy of reinforced cementitious matrix for strengthening masonry structures. Compos B
Eng 2017;127:175–95. https://doi.org/10.1016/j.compositesb.2017.05.048.
the proposed approach in reproducing the load carrying capacity of the
[5] Filippou CA, Chrysostomou CZ. Analytical model for textile reinforced mortar
examined curved structure also in the post-peak branch of the under monotonic loading. Constr Build Mater 2020;258:120178. https://doi.org/
load–displacement curve. In addition, the mechanisms ruling the 10.1016/J.CONBUILDMAT.2020.120178.
collapse have been correctly represented, especially with reference to [6] Grande E, Milani G, Imbimbo M. Theoretical model for the study of the tensile
behavior of FRCM reinforcements. Constr Build Mater 2020;236:117617. https://
the cracking of the composite and the stress transfer mechanisms. It can doi.org/10.1016/J.CONBUILDMAT.2019.117617.
be inferred that the effects of the curvature on the complex phenomena [7] Donnini J, Corinaldesi V, Nanni A. Mechanical properties of FRCM using carbon
involved in the structural behaviour of the FRCM reinforced vault has fabrics with different coating treatments. Compos B Eng 2016;88:220–8. https://
doi.org/10.1016/j.compositesb.2015.11.012.
been correctly represented. Notice that if at first glance it might emerge [8] Nerilli F, Marfia S, Sacco E. Micromechanical modeling of the constitutive response
a small inconsistency with the initial stiffness, the latter is likely due to of FRCM composites. Constr Build Mater 2020;236:117539.
the structure settling in the reference experiment, and therefore it is [9] Razavizadeh A, Ghiassi B, Oliveira DV. Bond behavior of SRG-strengthened
masonry units: Testing and numerical modeling. Constr Build Mater 2014;64:
possible to say that the proposed model is capable of recovering the 387–97. https://doi.org/10.1016/j.conbuildmat.2014.04.070.
more relevant aspects of the observed structural behavior. Moreover, it [10] Grande E, Imbimbo M, Sacco E. Investigation on the bond behavior of clay bricks
should be remarked that the material parameters have not been selected reinforced with SRP and SRG strengthening systems, Materials and Structures/
Materiaux et. Constructions 2015;48:3755–70. https://doi.org/10.1617/s11527-
fitting the results of the experimental curve, but based on available data 014-0437-x.
of the materials. [11] Allahvirdizadeh R, Oliveira DV, Silva RA. Numerical modeling of the seismic out-
The large sensitivity analysis shows the influence of variations of the of-plane response of a plain and TRM-strengthened rammed earth subassembly.
Eng Struct 2019;193:43–56.
mechanical parameters on the predicted overall response of the rein
[12] D’Ambra C, Lignola GP, Prota A, Fabbrocino F, Sacco E. FRCM strengthening of
forced structure. It emerges that for some parameters even large inac clay brick walls for out of plane loads. Compos B Eng 2019;174:107050. https://
curacies in their determination marginally affect the result, whereas doi.org/10.1016/j.compositesb.2019.107050.
other parameters need to be accurately identified to generate repre [13] Scacco J, Ghiassi B, Milani G, Lourenço PB. A fast modeling approach for numerical
analysis of unreinforced and FRCM reinforced masonry walls under out-of-plane
sentative results. These considerations might be very useful in practical loading. Compos B Eng 2020;180:107553. https://doi.org/10.1016/j.
applications to guide choices on the experimental analyses to be per compositesb.2019.107553.
formed and the criteria to be followed for accurately reproducing the [14] Wang X, Ghiassi B, Oliveira DV, Lam CC. Modelling the nonlinear behaviour of
masonry walls strengthened with textile reinforced mortars. Eng Struct 2017;134:
actual structural behaviour. 11–24.
Once the proposed numerical model has been validated with respect [15] Saidi M, Gabor A. Adaptation of the strain measurement in textile reinforced
to the experimental data, the influence of the FRCM reinforcement on cementitious matrix composites by distributed optical fibre and 2D digital image
correlation. Strain 2020;56:e12335.
the seismic capacity of the structure has been investigated by performing [16] Bertolesi E, Fagone M, Rotunno T, Grande E, Milani G. Experimental
pushover analyses both on the unstrengthened and strengthened vault. characterization of the textile-to-mortar bond through distributed optical sensors.
The obtained results suggest that the proposed modelling approach Constr Build Mater 2022;326:126640. https://doi.org/10.1016/J.
CONBUILDMAT.2022.126640.
could represent a useful compromise between the accuracy of the results [17] Cescatti E, da Porto F, Modena C. In-Situ Destructive Testing of Ancient
and the feasibility for use in practical applications. This is very inter Strengthened Masonry Vaults, International Journal of. Architectural Heritage
esting for the still open research field of the mechanics of FRCM rein 2018;12:350–61. https://doi.org/10.1080/15583058.2017.1323243.
[18] Borri A, Castori G, Corradi M. Intrados strengthening of brick masonry arches with
forced masonry arches and vaults. A further appealing aspect is that the
composite materials. Compos B Eng 2011;42:1164–72. https://doi.org/10.1016/j.
proposed modelling strategy requires for the mechanical parameters of compositesb.2011.03.005.
the masonry and the reinforcement a knowledge level readily achievable [19] Garmendia L, Marcos I, Garbin E, Valluzzi MR. Strengthening of masonry arches
with standard experimental approaches. with Textile-Reinforced Mortar: experimental behaviour and analytical
approaches, Materials and Structures/Materiaux et. Constructions 2014;47:
2067–80. https://doi.org/10.1617/s11527-014-0339-y.
Declaration of Competing Interest [20] Castellano A, Fraddosio A, Scacco J, Milani G, Piccioni MD. Dynamic Response of
FRCM Reinforced Masonry Arches. Key Eng Mater 2019;817:285–92. https://doi.
org/10.4028/www.scientific.net/kem.817.285.
The authors declare that they have no known competing financial [21] Boem I, Gattesco N. Cyclic behavior of masonry barrel vaults strengthened through
interests or personal relationships that could have appeared to influence Composite Reinforced Mortar, considering the role of the connection with the
the work reported in this paper. abutments. Eng Struct 2021;228:111518.
[22] Varró R, Bögöly G, Görög P. Laboratory and numerical analysis of failure of stone
masonry arches with and without reinforcement. Eng Fail Anal 2021;123:105272.
Data availability https://doi.org/10.1016/J.ENGFAILANAL.2021.105272.
[23] Carozzi FG, Poggi C, Bertolesi E, Milani G. Ancient masonry arches and vaults
strengthened with TRM, SRG and FRP composites: Experimental evaluation.
Data will be made available on request. Compos Struct 2018;187:466–80. https://doi.org/10.1016/j.
compstruct.2017.12.075.
Acknowledgment and Compliance with Ethical Standards [24] Malena M, de Felice G. Debonding of composites on a curved masonry substrate:
Experimental results and analytical formulation. Compos Struct 2014;112:
194–206. https://doi.org/10.1016/j.compstruct.2014.02.004.
Funding: Financial support from ReLUIS (Italian Department of Civil
23
[25] Alecci V, Misseri G, Rovero L, Stipo G, De Stefano M, Feo L, et al. Experimental [39] J.G. Rots, Computational modeling of concrete fracture, 1988. http://resolver.
investigation on masonry arches strengthened with PBO-FRCM composite. Compos tudelft.nl/uuid:06985d0d-1230-4a08-924a-2553a171f08f.
B Eng 2016;100:228–39. [40] H.P. Backes, Tensile strength of masonry, in: Proceedings of the 7th International
[26] de Santis S, Roscini F, de Felice G. Full-scale tests on masonry vaults strengthened Brick Maonry Conference, 1985: pp. 779–790.
with Steel Reinforced Grout. Compos B Eng 2018;141:20–36. https://doi.org/ [41] D.A. Hordijk, Local approach to fatigue of concrete, 1991.
10.1016/j.compositesb.2017.12.023. [42] Japan Society of Civil Engineers (JSCE), STANDARD SPECIFICATIONS FOR
[27] Borri A, Casadei P, Castori G, Hammond J. Strengthening of Brick Masonry Arches CONCRETE STRUCTURES - 2007 “Materials and Construction,” 2007.
with Externally Bonded Steel Reinforced Composites. J Compos Constr 2009;13: [43] Bejarano-Urrego L, Verstrynge E, Giardina G, Van Balen K. Crack growth in
468–75. https://doi.org/10.1061/(asce)cc.1943-5614.0000030. masonry: Numerical analysis and sensitivity study for discrete and smeared crack
[28] Pantò B, Malena M, de Felice G. A macro-modelling approach for arches modelling. Eng Struct 2018;165:471–85. https://doi.org/10.1016/j.
strengthened with externally bonded inorganic matrix composites. Structures engstruct.2018.03.030.
2021;33:4299–312. https://doi.org/10.1016/J.ISTRUC.2021.07.015. [44] Macorini L, Izzuddin BA. A non-linear interface element for 3D mesoscale analysis
[29] Bertolesi E, Carozzi FG, Milani G, Poggi C. Masonry arches retrofitted with steel of brick-masonry structures. Int J Numer Methods Eng 2011;85:1584–608. https://
reinforced grout materials: In-situ experimental tests and advanced FE simulations. doi.org/10.1002/nme.3046.
AIP Conf Proc 2017;1906. https://doi.org/10.1063/1.5012361. [45] Lourenço PB. Analysis of masonry structures with interface elements. Comput
[30] E. Ricci, D. v. Oliveira, E. Sacco, B. Ghiassi, Modelling of masonry arches Mech 1994;34.
strengthened at extrados with FRCM, in: Proceedings of the International Masonry [46] Gesualdo A, Brandonisio G, De-Luca A, Iannuzzo A, Montanino A, Olivieri C. Limit
Society Conferences, International Masonry Society, Milan, 2018: pp. 2150–2162. analysis of cloister vaults: The case study of Palazzo Caracciolo di Avellino. J Mech
[31] Zampieri P. Horizontal capacity of single-span masonry bridges with intrados Mater Struct 2019. https://doi.org/10.2140/jomms.2019.14.739.
FRCM strengthening. Compos Struct 2020;244:112238. [47] Bertolesi E, Milani G, Carozzi FG, Poggi C. Ancient masonry arches and vaults
[32] E. Bertolesi, G. Milani, J. Adam, P. Calderón, 3D advanced numerical modelling of strengthened with TRM, SRG and FRP composites: Numerical analyses. Compos
a catalan-layered masonry vault unreinforced and reinforced with glass-TRM Struct 2018;187:385–402. https://doi.org/10.1016/j.compstruct.2017.12.021.
materials and subjected to vertical support movements, COMPDYN Proceedings. [48] P.B. Lourenço, A user/programmer guide for the micro-modelling of masonry
2021-June (2021). https://doi.org/10.7712/120121.8495.19253. structures, Delft, 1996.
[33] Pantò B, Malena M, de Felice G. Non-Linear Modeling of Masonry Arches [49] STANDARD SPECIFICATIONS FOR CONCRETE STRUCTURES − 2007 “Design”
Strengthened with FRCM. Key Eng Mater 2017;747:93–100. https://doi.org/ Japan Society of Civil Engineers, (n.d.).
10.4028/www.scientific.net/kem.747.93. [50] Ceb-Fip C, Code M. Comite Euro-International Du Beton. Paris 1990;1991:87–109.
[34] de Santis S, Carozzi FG, de Felice G, Poggi C. Test methods for Textile Reinforced [51] Grande E, Milani G, Sacco E. Modelling and analysis of FRP-strengthened masonry
Mortar systems. Compos B Eng 2017;127:121–32. https://doi.org/10.1016/j. panels. Eng Struct 2008;30:1842–60. https://doi.org/10.1016/j.
compositesb.2017.03.016. engstruct.2007.12.007.
[35] Focacci F, D’Antino T, Carloni C. Tensile Testing of FRCM Coupons for Material [52] M. Angelillo, P.B. Lourenço, G. Milani, Masonry behaviour and modelling, in: CISM
Characterization: Discussion of Critical Aspects. J Compos Constr 2022;26: International Centre for Mechanical Sciences, Courses and Lectures, 2014: pp.
04022039. https://doi.org/10.1061/(ASCE)CC.1943-5614.0001223. 1–26. https://doi.org/10.1007/978-3-7091-1774-3_1.
[36] Lourenço PB. Computational strategies for masonry structures. Delft University of [53] Quinteros-Mayne R, de Arteaga I, Goñi-Lasheras R, Villarino A, Villarino JI. The
Technology; 1996. Doctoral Dissertation,. influence of the elastic modulus on the finite element structural analysis of
[37] Selby RG, Vecchio FJ. Three-dimensional Constitutive Relations for Reinforced masonry arches. Constr Build Mater 2019. https://doi.org/10.1016/j.
Concrete. Tech Rep 1993;93–02. conbuildmat.2019.06.013.
[38] Feenstra PH. Computational aspects of biaxial stress in plain and reinforced [54] N. Ademović, M. Hadzima-Nyarko, G. Pavić, Modern Strengthening Methods for
concrete. Delft University of Technology; 1993. URM, in: Sustainability and Automation in Smart Constructions, 2021: pp.
357–362. https://doi.org/10.1007/978-3-030-35533-3_43.
24

1 s2.0 S0141029622011920 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0141029622011920 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0141029622011920 Main

Uploaded by

Copyright:

Available Formats

Engineering Structures 274 (2023) 115116

Contents lists available at ScienceDirect

An effective numerical modelling strategy for FRCM strengthened curved

1. Introduction composites are known as Fabric Reinforced Cementitious Matrix

Fig. 6. Modelling of the reinforced vault.

C(0, − 1164.98) 60 300 2555 461 2025

5.3. FRCM composite 5.4. Masonry bricks

experimental tests on the materials or determined according to literature

6.1. Reinforced vault (RV)

The numerical behaviour of the reinforced vault is practically linear

stiffness of the simulated response with respect to the experimental

You might also like