You are on page 1of 21

Engineering Structures 266 (2022) 114422

Contents lists available at ScienceDirect

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

A new empirical formulation for the out-of-plane resistance of masonry


infills in reinforced concrete frames
Fabio Di Trapani *, Alessandro Vizzino , Giovanni Tomaselli , Antonio Pio Sberna ,
Gabriele Bertagnoli
Politecnico di Torino, Dipartimento di Ingegneria Strutturale Edile e Geotecnica (DISEG), Corso Duca degli Abruzzi 24, 10129 Turin, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: The paper presents a new empirical expression estimating the out-of-plane (OOP) resistance of infilled reinforced
Out-of-plane concrete frames subject to horizontal forces. The new model is calibrated through an optimization process based
FEM on a hybrid dataset, including experimental data from real tests and from numerical simulations obtained from a
Masonry
refined FE micro-model realized in Abaqus®. The new expression considers the effect of vertical loads and also
Infilled Frames
Reinforced concrete
introduces a conversion factor to uniformize point-load and uniform load out-of-plane tests. The final expression
Data-driven is also flexible with respect to available data on the infill material properties, as it is specialized in two versions,
Empirical one providing the conventional compressive strength of the units, the other providing the conventional
compressive strength of masonry as possible inputs. Results presented in the paper will show a noticeable ac­
curacy of the proposed model in estimating the ultimate out-of-plane load of a masonry infill wall, with respect to
available models. The formula provides the reference undamaged OOP resistance and can be easily combined
with available strength-reduction functions to consider the effect of prior in-plane loading.

1. Introduction converged in claiming that, following a seismic event, infill panels are
weakened due to combined IP and OOP damage, which leads to different
Evaluation of out-of-plane (OOP) resistance of infilled frames is an potential outcomes, ranging from the loss of functionality of the infill to
issue of primary importance in the assessment of seismic risk of frame its complete collapse. Because of this, a number of analytical relation­
structures. In fact, even if infills are not intended as structural elements, ships were proposed to evaluate the out-of-plane resistance (Dawe and
they strongly interact with primary structures undergoing in-plane (IP) Seah 1989 [1], Angel et al. 1994 [2], Bashandy et al. 1995 [5], Abrams
damage due to the seismically induced interstorey drift, and then et al. 1996 [6], Flanagan and Bennett 1999 [3]). A simplified relation­
becoming weaker against out-of-plane forces. OOP failure of infills is a ship, summarizing the outcomes by [2,6] was also included in FEMA 356
quite dangerous event, as it affects safety of people in the proximity area 2000 [7] technical guidelines. However, these formulations were based
of a building hit by an earthquake. Simple, and at the same time reliable, on limited experimental evidence, in fact more recent comparative
verification methods are therefore necessary to perform infill walls out- studies (e.g. Asteris et al. 2017 [8], Liberatore et al. 2020 [9]) high­
of-plane resistance safety checks both in the case of new and existing lighted significant dispersion of results by their application.
buildings. Experimental experience has been largely enriched in last ten years.
The investigation concerning the out-of-plane seismic performance Focusing on masonry infilled reinforced concrete frames the experi­
of masonry infills started several decades ago, however, in last ten years mental works by da Porto et al. 2013 [10], Hak et al. 2014 [11], Furtado
a huge concentration of experimental and numerical studies has been et al. 2016a [12], Akhoundi et al. 2016 [13], Ricci et al. 2018a [14],
recognized. First experimental tests (Dawe and Seah 1989 [1], Angel 2018b [15], De Risi et. al. 2019 [16], Di Domenico et. al 2021 [17],
1994 [2], Flanagan and Bennett 1999 [3], Calvi and Bolognini 2001 [4]) Agante et al. 2021[18] provide a clear image on how the research in­
have demonstrated that masonry infills can develop significant resis­ terest on the topic has increased in the last few years. Some of these
tance against OOP actions because of the arching mechanism when they studies provided pure out-of-plane tests while others included both in-
are adequately confined by a boundary frame. Secondly, these studies plane and out-of-plane testing of the infilled frame specimens. Recent

* Corresponding author at: Dipartimento di Ingegneria Strutturale, Edile e Geotecnica, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Turin, Italy.
E-mail address: fabio.ditrapani@polito.it (F. Di Trapani).

https://doi.org/10.1016/j.engstruct.2022.114422
Received 11 November 2021; Received in revised form 6 April 2022; Accepted 15 May 2022
Available online 13 July 2022
0141-0296/© 2022 Elsevier Ltd. All rights reserved.
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

comprehensive reviews collecting the experimental studies can be found associated with each parameter, and included in the proposed formula,
in [19,20]. has been carried out through a genetic algorithm based robust optimi­
Simultaneously, proposals for numerical modelling of the combined zation process involving all available data. A comparison of the pro­
in-plane/out-of-plane behaviour of infill walls have been provided. posed model with available literature expression is finally presented and
Among these Kadysiewski and Mosalam 2009 [21], Mosalam and Günay commented.
2015 [22], Furtado et al. 2016b [23], Ricci et al. 2018c [24], Mazza
2018 [25], Di Trapani et. al 2018 [26] proposed different ways to 2. Resume of available literature models for the evaluation of
integrate IP and OOP responses using simplified equivalent strut models. out-of-plane resistance of masonry infills
In particular, the arching action was well reproduced in [26] by using
fiber-section elements. Eight well-known literature models are here selected and briefly
Going back to the prediction of the out-of-plane resistance without IP recalled. The out-of-plane resistance of an infill wall is generally
damage, additional analytical expressions (e.g. Ricci et al. 2018c [24], measured by the maximum uniform lateral pressure causing the collapse
Liberatore et al. 2020 [9], Pradhan et al. 2021 [27]) based on the (qOOP), however, it is also reasonable to convert ultimate OOP pressures
enriched experimental dataset have been recently proposed. Within the into ultimate OOP forces by multiplying pressures by the net area of the
aforementioned studies, and other additional ones, OOP strength infill (Ai), intended as the product of the net infill length (l) and height
reduction functions have been also proposed as a function of the in-plane (h) (Ai = l × h).
interstorey drift [9,14,27,28]. One of the most popular expressions for the estimation of the ulti­
A further comparison of results of most of the above listed analytical mate OOP load capacity of an infilled frame is the one proposed by
models for the prediction of the OOP resistance of masonry infill without Angel, 1994 [2] and Abrams et al., 1996 [6]. These studies, evaluated
prior in-plane damage confirms that although starting from similar the out-of-plane resistance of masonry infills including the arching ac­
theoretical considerations, they are often conflicting [9]. In particular, tion, the degree of previous in-plane damage and the effect of frame
the observation of data (e.g. in [9]) gives the feeling that some of them stiffness according to the following formula:
are more reliable in some cases and less in others. This recognized dif­ ′
2fm
ficulty in achieving a general validity by the above listed analytical qOOP = ⋅λ⋅R1 ⋅R2 (1)
models, which are described in detail in the following section can be (h/t)
interpreted considering three major aspects:
where fm is the compressive strength of the masonry expressed in MPa

• available analytical expressions are relatively too simple to capture and t is the thickness of the infill. The lengths are expressed in mm.
the potential resistance variability associated with the large possible Coefficients R1, R2 and λ, are used to consider the effect of in-plane
combination of arrangements of masonry with boundary frames, damage (Eq. (2)), the frame members flexural stiffness (Eq. (3)) and
especially in terms of relative strength and stiffness; the infill slenderness (Eq. (4)) respectively, so that:
• experimental background is continuously updating, therefore, older ( ( ){
h
( )[
h
( )]})Δ/2Δcr
h
formulas reflect the limited experimental state-of-the-art at the time; R1 = 1.08+
t
− 0.015+
t
− 0.00049+0.000013
t
• on the contrary, some experimental datasets used to calibrate the
(2)
expressions are too wide, as they include data from infilled rein­
forced concrete frames, steel frames and confined masonry, also R2 = 0.357 + 2.49 × 10− 14
EI⩽1.0 (3)
considering different ways of application of horizontal and vertical
loads. λ = 0.154e− 0.0985(h/t)
(4)

In addition, inaccurate estimations of the OOP resistance of an in which Δ is the current in-plane relative displacement, Δcr is the same
infilled frame without prior in-plane damage influence the reliability of displacement associated with the formation of the first crack in the
further estimations of the residual OOP plane strength considering the panel, EI is the flexural rigidity of the smallest element composing the
in-plane damage. concrete frame. It observed that in the case of no in-plane damage
In this framework, the definition of a robust generalized relationship (Δ=0), coefficient R1 is equal to 1.
providing a reliable estimation of the undamaged OOP resistance of In FEMA 356 [7] a simplified and more practice version of Eq. (1) is
masonry infill is still needed. provided. In the code it is stated that if arching action has to be
As a consequence of the aforementioned considerations, this paper considered, the lower bound OOP strength of an infill panel, can be
proposes a new empirical formulation for the estimation of the out-of- determined as follows:
plane resistance of infilled RC frames without prior in-plane damage.
The formulation of the new predictive model is based on a hybrid

0.7fm
qOOP = ⋅λ2 (5)
dataset collecting data from real experimental tests and numerical tests (h/t)
from a refined FE model realized in ABAQUS [29]. After a proper
experimental calibration and validation, the refined numerical model is where λ2 values are provided in a table as a function of the slenderness of
used to generate additional reference numerical tests, increasing the the infill (h/t). An interpolation of λ2 vs. (h/t) data of can be summarized
robustness of the reference dataset. The model also allowed to investi­ by λ2 = 1.38(h/t)− 1.41 [8].
gate and highlight the role of additional relevant aspects, such as the Another relationship proposed by Dawe and Seah, 1989 [1],
effect of vertical loads and the modality of application of the lateral load included both flexural and torsional stiffness of the frame members ac­
during the experimental tests (e.g. 4-point load or uniform load), which cording to the following formula to be used for four-side restrained
have not been considered in previous research. Based on this, the new infills:
proposed relationship introduces two additional specific parameters,
(6)
′ 0.75 2
qOOP = 4.5(fm ) t (α/l2.5 + β/h2.5 )
taking into account of the effect of vertical loads and introducing a
correction factor to uniformized results by OOP tests carried out by
where
point-load devices or airbags. The advantage introduced by the pro­
posed empirical formulation derive from the major number of parame­
ters that it was possible to simultaneously consider within the out-of-
plane capacity model. The calibration of the empirical coefficients

2
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

1 3. Refined numerical FE model: definition, calibration and


α = (Ec Ic h2 + Gc Jc th)0.25 ⩽50
h validation
(7)
1
β = (Ec Ib l2 + Gc Jb tl)0.25 ⩽50
l 3.1. FE model definition and calibration
In previous equations, qOOP is expressed in kPa, the lengths are
The refined FE micro-model was realized with the Abaqus [29]
expressed in mm, Ec and Gc are the frame Young’s modulus and shear
software platform. The reference experimental test used for the model
modulus respectively, Ic and Ib are the moments of inertia of beams and
definition and calibration is specimen OOP_4E/80 by Ricci et al. 2018a
columns and Jc and Jb are the torsional constants of beams and columns,
[14], consisting of a hollow clay masonry infilled RC frame. The infill is
respectively. Coefficients α and β assume more relevance in the case of
restrained at the four sides and has dimensions 2350 × 1830 mm and
the more flexible steel frames. Later, a further expression including also
thickness 80 mm (Fig. 1). The out-of-plane load was applied by imposing
the two-way effect of the arching mechanism has been proposed by
an out-of-plane displacement with an actuator equipped with four point-
Bashandy et al. 1995 [5] in the following form:
load devices pushing on four points placed at 1/3 of the infill main di­
( ) ( )
8Myv 8Myh xyv l agonals (Fig. 1). The frame is restrained at the top by a steel device to
qOOP = (l − h) + 8Myv ln(2) + ln l (8)
h h xyh l − h/2 avoid OOP rotations. No vertical loads were applied during the test.
Mechanical properties of specimen materials are shown in Table 1.
where the terms Myh, Myv, xyh and xyv have the following expressions: Masonry blocks constituting the infill are modelled individually as
′ ′ well as frame and reinforcement elements. Mortar joints between blocks
85fm 85fm
Myv = (t − xyv )2 ; Myh = (t − xyh )2 (9) and between blocks and columns are modelled by frictional interface
4 4
elements. Some views of the model assembly are shown in Fig. 2.
tfm

tfm
′ The concrete damaged plasticity (CDP) material model is used to
xyv = [ ]
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ; xyv = [
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
] (10) model the behaviour of concrete frame members and masonry blocks.
The units depicted in Fig. 1 are phenomenologically modelled as solid
h
√̅̅̅̅̅̅̅̅̅̅̅̅̅ ̅ l
√̅̅̅̅̅̅̅̅̅̅̅̅̅
1000Em 1 − 2 2
1000Em 1− 2 2
2 (h/2) +t 2 (l/2) +t
isotropic brick elements. To take into account the effect of the ortho­
In the Eurocode 6 [30], a simpler one-way arching action-based tropic behaviour due to the directionality of the holes, a reference
formula is also proposed: conventional resistance (̃f ) is defined and assigned to the units, so that:
b

fm
′ √̅̅̅̅̅̅̅̅̅̅̅
qOOP = (11) ̃f b = fbh ⋅fbv (16)
(h/t)2
Symbols appearing in Eq. (11) have the same meaning of those where fbh and fbv are the experimental mean gross compressive strengths
of the units along the horizontal and vertical directions. This approach
previously described. In this formula the out-of-plane resistance has
inverse proportionality with the square of the infill slenderness ratio. A allows making the model computationally sustainable without losing in
accuracy in estimating the overall response [31], although local
modified version of the predictive model in Eq. (11) has been proposed
by Ricci et al. 2018c [24], who corrected the expression by the intro­ cracking patterns could partially differ from the real ones. Moreover,
this allows considering the specimen as independent from the shape of
duction of empirical coefficients and using a wider dataset. The final
expression was: the units and suitable for simulating masonries made by different ty­
pologies of units. Given that, no experimental evaluations about elastic
t1.59 moduli of the units were available, a conventional elastic modulus to be
qOOP = 1.95⋅fm0.35 ⋅ 2.96 (12)
h assigned to the bricks was estimated as a function of the previously
defined conventional strength (̃f ), in analogy to what suggested in
b
where fm is expressed in MPa and the lengths are expressed in m.

Eurocode 6 [32] for the estimation of the elastic modulus of masonry:


More recently, Liberatore et al. [9] provided a further update
considering also the influence of the aspect ratio (h/l) on the infill with ̃ b = 1000⋅̃f b
E (17)
the following expression:
( ) With reference to Fig. 3, the CDP model provides the following
0.26fm0.9 h general responses in tension and compression:
qOOP = 1.23
⋅ (13)
(h/t) l
σc = (1 − dc )E0 (εc − εplc )
(18)
The authors suggest using Eq. (13) for masonry infills having a σt = (1 − dt )E0 (εt − εplt )
compressive strength not larger than 15 MPa. Another recent expression
considering the aspect ratio of the infill wall, has been proposed by where σc and σt are the current stresses in compression and tension, E0 is
Pradhan et al. 2021 [27] as follows: the initial elastic modulus, εc and εt are the current total strains in
1.1 compression and tension, εpl pl
c and εc are the plastic strains in compres­
800̃f m
qOOP = (14) sion and tension, dc and dt are the damage parameters in tension and
(h/t)1.9 (h/l)1.25 compression and evaluated as:
The major novelty of this model regards the use of the conventional ⎧
(if εc ⩽εcc )

(if εt ⩽εcr )
⎨0 ⎨0
strength of the infill wall (f̃m ) as defined by Di Trapani et al. [31–32], dc = σc ; dt = σt (19)
⎩1 − (if εc > εcc ) ⎩1 − (if εt > εcr )
that is: σ cc σt0
√̅̅̅̅̅̅̅̅̅̅̅̅̅
̃f m = fmh ⋅fmv (15) where σcc and σ t0 are the peak stresses in compression and tension.
The constitutive law used to define the behaviour of the masonry
where fmh and fmv are the compressive strengths of the masonry along the
blocks in compression is the one by Kent-Scott-Park [33] (parabolic with
two orthogonal directions.
linear softening) (Fig. 4a). The model proposed by Hsu & Mo 2010 [34]
was used to describe the behaviour of the blocks in tension (Fig. 4b).
Considering the material properties listed in Table 1 and according to
Eqs. (16)–(17), the conventional compressive strength and elastic

3
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 1. Design details of specimen OOP/4E/80 [14].

angle of 10◦ was adopted as suggested by Van der Pluijm et al. 2000
Table 1 [36]. Plastic parameters regulating the eccentricity (ε), biaxial resis­
Material mechanical properties of specimen 80/OOP/4E [14].
tance domain fb0/fco, and viscosity were assumed as suggested in [29].
Mechanical properties Mean value Steel reinforcement was modelled using 1D truss elements whose me­
(MPa)
chanical response is simulated by a simple elasto-plastic with strain
Concrete compressive strength (fcm) 36.00 hardening material model. Steel rebars were modelled as embedded
Steel rebars’ yielding stress (fym) 552 elements within the concrete, so that relative sliding between steel bars
Masonry compressive strength (parallel to the holes) (fmh) 2.45
Masonry compressive strength (perpendicular to the holes) (fmv) 1.81
and concrete could not occur. Mortar joints behaviour was modelled
Masonry elastic modulus (parallel to the holes) (Emh) 1255 using elasto-plastic interfaces with friction and cohesion using the
Masonry elastic modulus (perpendicular to the holes) (Emv) 1090 surface-based cohesive model implemented in the Abaqus platform. The
Masonry tensile strength (fmt) 0.23 initial elastic relationship between the relative joint displacements
Masonry shear modulus (Gm) 315
(normal and tangential) is expressed by:
Bricks’ compressive strength (parallel to holes) (fbh) 5.00*
Bricks’ compressive strength (perpendicular to holes) (fbv) 2.00*
⎛ ⎞ ⎛ ⎞⎛ ⎞
tn knn 0 0 δn
Mortar compressive strength (fmm) 8.30 ⎝ ts ⎠ = ⎝ 0 kss 0 ⎠⎝ δs ⎠ (20)
*Lower bound declared by the producer. tt 0 0 ktt δt

where tn, ts and tt are the normal and tangential interface stresses, knn, kss,
moduli of the units are ̃f b = 3.16 MPa and E ̃b = 3160 MPa. As regards
and ktt are the interface normal and tangential stiffnesses, δn, δs and δt
conventional tensile strength this was estimated as a function of the
√̅̅̅̅ are the normal and tangential relative interface displacements. A view of
compressive strength as ̃f bt = 0.33 ̃f b ([35]) resulting in 0.58 MPa. the interface damage mechanisms is shown in Fig. 5.
The elastic and plastic parameters used for concrete and masonry are The interface stiffness values have been initially obtained as:
reported in Table 2. As regards the angle of dilatancy for the concrete a
value of 37◦ was assumed as suggested in [29], while for masonry an

Fig. 2. FE model assembly: a) concrete frame; b) reinforcement; c) infilled frame.

4
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 3. General working principle of concrete damaged plasticity model: a) in compression; b) in tension.

Fig. 4. Constitutive CDP laws adopted for masonry blocks: a) in compression; b) in tension.

Table 2
Material elasticity and plasticity parameters assumed for concrete and masonry blocks.
Material Mass density Elasticity parameters Plasticity parameters

kN/m3 Young’s modulus (MPa) Poisson’s ratio ν Dilatation Angle ψ (◦ ) Eccentricity ε fb0/fco Kc Viscosity

Concrete 25 32,308 0.3 37 0.1 1.16 0.667 0.0003


Masonry units 11 3160 0.2 10 0.1 1.16 0.667 0.0003

Fig. 5. Normal and tangential damage interface damage mechanisms.

Table 3
Interface mechanical properties.
Contact behaviour Cohesive behaviour

Tangential Normal Interface stiffness Damage initiation strengths Damage evolution fracture energy
(N/mm3) (MPa) (N/mm)

Friction coefficient (-) Normal Tangential Tangential Normal Tangential Tangential


μ Hard contact Knn Kss Ktt fmm,t fmt GF
0.6 200 88 88 0.83 0.23 0.02075

5
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

̃b
E Gm
knn = ; kss = ktt = (21)
tj tj

where tj is the thickness of the joint (10 mm). The final optimal values
were then calibrated based on the experimental response and are re­
ported in Table 3. The normal behaviour of the interfaces is modelled as
hard contact in compression, while the tensile response is limited by the
tensile strength of the mortar fmm,t, here estimated as 10% of the mortar
compressive strength fmm. The tangential inelastic response is regulated
by the Mohr-Coulomb criterion as:
τ = c0 + μσ n (22)

where c0 is the cohesion, here assumed as the shear strength of the


masonry in absence of normal stresses (fmt), μis the friction coefficient
and σn the normal stress acting on the interface. The interface damage
evolution is regulated by the fracture energy that is here evaluated as GF
= 0.00025 fmm,t (MN/m) [36]. A summary of the interface mechanical
Fig. 7. Comparison between numerical simulations and experimental response
parameters and the adopted values is reported in Table 3. Concrete
of the test OOP/4E, t = 80 mm by Ricci et al. 2018a [14].
members and masonry units were modelled by solid 3D elements with 8
nodes (C3D8R, with reduced integration and hourglass control). The
mesh size was defined as 50 × 50 × 40 mm (where 40 mm is the The analysis also allowed an investigation in terms of stress distri­
dimension along the thickness). This was choice was taken after bution and damage. Figs. 9–10 show the maximum and minimum and
considering a 80 × 80 × 80 mm mesh providing less accurate results, as principal stresses on the windward and leeward sides of the specimen in
shown in the subsequent section. A more refined mesh was not consid­ correspondence of the peak load. The analysis of the stress field confirms
ered due to the significantly highier computational effort required. A that the 2-way bending response of the infill. The observation of the
dynamic implicit analysis with quasi-static analysis was performed to stress values also reveals compressive stress concentration in corre­
simulate the OOP test by imposing horizontal displacements at the four spondence of the four points of application of the horizontai loads
loading plates. Base and top constraints were introduced to simulate the (Fig. 9a), as well at the sides of the infill on the leeward face (Fig. 9b). At
actual test restraint conditions. The model assembly with the applied the same time the major tensile fields are revealed orthogonally to the
displacements and the imposed constraints is shown in Fig. 6a. The mesh diagonals of the infill.
refinement of the elements is represented in Fig. 6b. The arching response is highlighted in Fig. 11, which shows a median
vertical section of the specimen with the compressive principal stresses
(coulure fields) map at different stages of the analysis. In can be clearly
3.2. FE model test results observed the transition of the compression fields from the windward
midspan sections to the leeward sides of the infills in contact with the
Results of the numerical simulations of the OOP test with the so foundation beam.
defined FE model are shown in Fig. 7 and compared with the experi­ The damage maps in Fig. 12, show the damage patterns in
mental response. From the comparisons, it can be observed that the compression and tension at the end of the simulation measured by the
model with a more refined mesh (50 × 50 × 40 mm) is able to reproduce damage indexes dc and dt above defined. As regards the windward side
more accurately the experimental behaviour especially in terms of peak (Fig. 12a), it can be observed that the damage in compression is prop­
resistance, and is assumed as the reference one. Moderate discrepancies, agated from the regions of application of the lateral load to the corners
related to the above-described modelling assumptions, are observed for of the frame. In Fig. 12b the tensile damage on the leeward side of the
what concerns the initial stiffness and the post peak behaviour. The infill is shown. It develops along the two diagonal failure lines in
deformed shape of the model in correspondence of the peak load is also proximity of the point of application of the lateral load.
shown in Fig. 8. The latter also demonstrates the kinematic consistency It is noteworthy comparing the tensile damage in Fig. 12b with the
between the real test and the numerical simulation. experimental damage pattern reported in Fig. 12c, which qualitatively

Fig. 6. Definition of the FE model: a) Scheme of the model assembly, loads and constraints; b) Mesh of the model,

6
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 8. Deformed shape of the infilled frame micro-model in correspondence of the peak-load.

Fig. 9. Minimum principal stress fields (compression) at the peak load (MPa): a) windward side; b) leeward side.

Fig. 10. Maximum principal stress fields (tension) at the peak load (MPa): a) windward side; b) leeward side.

confirms the orientation of the diagonal failure lines in tension. quite similar. However, masonry compressive strength determined
experimentally along the two directions was significantly higher with
3.3. FE model general validation respect to the other specimens. It should be observed that the units’
compressive strengths indicated in Table 4, refer to a lower bound
In order to generalize the validity of the previously described declared by the producer, therefore it is reasonable assuming that their
modelling approach, two validation tests were carried out against two actual strength is higher. In order to take into account this, a strength
further experimental tests. In detail, the tests were those on specimen increment factor was determined by considering the ratio between the
OOP/4E/120 mm by Ricci et al. 2018b [15] and, specimen OOP by De conventional resistances of the masonry evaluated as in Eq. (15):
Risi et al. 2019 [16]. The first test was the same as the one previously ̃f
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
4.63 × 2.37 3.31
analysed except for the thickness of the infill which is 120 mm instead of αs = m, OOP = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ = = 1.55 (23)
̃f m, OOP/4E 2.45 × 1.85 2.13
80 mm. Material properties and geometry of the specimen are again
those in Table 1, therefore material elasticity, plasticity and interface In this way, to model the units of specimen OOP [16] it is assumed
properties remained the same as those reported in Table 2 and Table 3. ̃f
b,OOP = αs f b,OOP/4E = 1.55 × 3.16 = 4.90MPa, and according to Eq. (17)
̃
Specimen OOP by De Risi et al. [16] has an arrangement that is similar to
the previous ones. The frame reinforcement is the same as specimen Eb = 4900 MPa. Stress–strain responses of the units in tension and
̃
OOP/4E/80, while the infill geometry is square (1830 mm × 1830 mm) compression have been recalibrated as a function of these values.Inter­
with a thickness of 80 mm. Material properties are reported in Table 4. face mechanical properties have been re-evaluated according to the
Masonry units were the same as those used for specimens 4E/OOP/80 criteria described in the previous section. Resulting values are reported
and 4E/OOP/120 and also the compressive strength of the mortar was in Table 5.Experimental/numerical comparisons of the OOP responses

7
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 11. Compressive stress field in the median section of the specimen: a) at the first cracking; b) at the peak load; c) at the end of the test. Values in MPa.

Fig. 12. Damage pattern on the infill at the end of the test: a) FE model damage in compression (dc) on the windward side; b) FE model damage (dt) in tension on the
leeward side; c) experimental damage on the leeward side [14].

fact that masonry units of the real specimen are hollow and then
Table 4
potentially subjected to localized crushed due to stress-concentrations
Material mechanical properties of specimen OOP/4E [14].
causing the instability of the clay layers. However, this slight inconsis­
Mechanical properties Mean value tency did not affect the prediction of the ultimate load which is the
(MPa)
major aim of the simulations. The deformed shapes of the FE models are
Concrete compressive strength (fcm) 42.90 also reported in Fig. 14.
Steel rebars’ yielding stress (fym) 524.5
Masonry compressive strength (parallel to the holes) (fmh) 4.63
Masonry compressive strength (perpendicular to the holes) (fmv) 2.37
4. Parametric investigation and collection of numerical tests
Masonry elastic modulus (parallel to the holes) (Emh) 3452
Masonry elastic modulus (perpendicular to the holes) (Emv) 1891 Additional numerical models were generated starting from the
Masonry tensile strength (fmt) 0.29 reference ones above defined. The additional models were defined by
Masonry shear modulus (Gm) 1622
varying single geometrical and mechanical parameters with respect to
Bricks’ compressive strength (parallel to holes) (f*bh) 5.00*
Bricks’ compressive strength (perpendicular to holes) (f*bv) 2.00* the original values. In this way the influence of each variation to the
Mortar compressive strength (fmm) 9.44 overall out-of-plane resistance is analysed. Moreover, the additional
numerical tests analysed, and the respective results in terms of out-of-
*Lower bound declared by the producer.
plane resistance, are used to enrich the dataset of experimental tests
already available to perform the empirical data processing described in
of the specimens are shown in Fig. 13. The latter confirm a good
the subsequent section.
agreement of the FE model in predicting the peak load and to follow the
Parameters varied with respect to the original configuration are:
post-peak behaviour, although the numerical response seems charac­
terized by a more ductile behaviour. This slight inconsistency with the
• the infill slenderness (h/t).
more brittle trend of the experimental response can be justified by the

Table 5
Interface mechanical properties for specimen OOP by De Risi et al. [16].
Contact behaviour Cohesive behaviour

Tangential Normal Interface stiffness Damage initiation strengths Damage evolution fracture energy
(N/mm3) (MPa) (N/mm)

Friction coefficient (-) Hard contact Normal Tangential Tangential Normal Tangential Tangential
μ Knn Kss Ktt fmm,t fmt GF
0.6 300 132 132 0.94 0.29 0.0179

8
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 13. Comparison of numerical and experimental OOP responses of: a) specimen OOP/4E/t=120 mm by Ricci et al. 2018b [15]; b) specimen OOP/t=80 mm by
De Risi et al. 2019 [16].

Fig. 14. Deformed shapes of the FE models: a) specimen OOP/4E/t=120 mm by Ricci et al. 2018b [15]; b) specimen OOP/t= 80mm by De Risi et al. 2019 [16].

• the units conventional resistance (̃f b ). different specimens, characterized by the aspect ratios (l/h) 1 and 1.28.
• the entity of the vertical loads acting on the beam (q) and on the The slenderness was varied between 9.15 and 22.88, to cover a suffi­
columns (Qc). ciently wide range. The OOP force displacement curves in Fig. 15
• the mode of application of the lateral load (4-point or uniform). confirm the major influence of the slenderness ratio to the OOP resis­
tance. Maximum OOP forces values collected in Fig. 16a highlight the
Table 6 summarizes the numerical tests carried out and the param­ strong inverse proportionality.
eters each time varied. The effect of the conventional units’ resistance variation was eval­
The variation of the slenderness ratio (h/t) was performed on two uated on three different models, namely Ricci et al. [14] (t=80 mm),

Table 6
Reference specimens and generated numerical specimens’ details (varied parameters are in bold).
t h l h/t ̃f q Q = ql v Load Type
b
(mm) (mm) (mm) (-) (MPa) (kN/m) (kN) (%) (–)

Ricci et al. (2018a) [14] 80 1830 2350 22.88 3.16 0 0 0 4-point


FEM-R2018a-T1 200 1830 2350 9.15 3.16 0 0 0 4-point
FEM-R2018a-R1 80 1830 2350 22.88 1.58 0 0 0 4-point
FEM-R2018a-R2 80 1830 2350 22.88 6.32 0 0 0 4-point
FEM-R2018a-Q1 80 1830 2350 22.88 3.16 10 23.5 0.56 4-point
FEM-R2018a-Q2 80 1830 2350 22.88 3.16 20 47.0 1.11 4-point
FEM-R2018a-Q3 80 1830 2350 22.88 3.16 30 70.5 1.67 4-point
FEM-R2018a-L1 80 1830 2350 22.88 3.16 0 0 0 Uniform

Ricci et al. (2018b) [15] 120 1830 2350 15.25 3.16 0 0 0 4-point
FEM-R2018b-R1 120 1830 2350 15.25 1.58 0 0 0 4-point
FEM-R2018b-R2 120 1830 2350 15.25 6.32 0 0 0 4-point
FEM-R2018b-L1 120 1830 2350 15.25 3.16 0 0 0 Uniform

De Risi et al. (2019) [16] 80 1830 1830 22.88 4.90 0 0 0 4-point


FEM-DR2019-T1 120 1830 1830 15.25 4.90 0 0 0 4-point
FEM-DR2019-T2 200 1830 1830 9.15 4.90 0 0 0 4-point
FEM-DR2019-R1 80 1830 1830 22.88 2.45 0 0 0 4-point
FEM-DR2019-R2 80 1830 1830 22.88 9.80 0 0 0 4-point
FEM-DR2019-T1 80 1830 1830 22.88 4.90 0 0 0 Uniform

9
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 15. Effect of infill slenderness variation on the OOP response in specimens with: a) l/h = 1.28 (Ricci et al. [14–15]): b) l/h = 1.0 (De Risi et al. [16]).

Fig. 16. Maximum varying geometrical and mechanical parameters: a) effect of slenderness variation; b) effect conventional unit strength variation; c) effect of
vertical load by its application.

Ricci et al. [15] (t=120 mm) and De Risi et al. [16] (t=80 mm). The it can be observed from Fig. 16b, this behaviour seems to be charac­
conventional strength was varied in the range 0.5–2.0̃f b,ref , where ̃f b,ref is terized by a limit, beyond which further increases of the unit strength do
the conventional resistance evaluated by the reference experimental not significantly affect the OOP resistance. This result can be justified by
tests. Results show that an increment of the unit strength results in an the fact that a masonry strength increment is effective to increase the
increase of the out-of-plane resistance of the infilled frame (Fig. 17). As OOP resistance until the stability of the arching mechanism is granted.
The latter is related to the slenderness of the wall, which has the major

Fig. 17. Effect of unit’s compressive strength variation on the OOP response of: a) Ricci et al. [14]; b) Ricci et al. [15]; c) De Risi et al. [16].

10
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

influence on the overall response. the infill wall. The load transfer is instead effective in the case of
The influence of vertical loads acting was investigated using spec­ application of the vertical load on the beam, as this is governed by the
imen OOP/4E/80 by Ricci et al. [14] as reference. The analyses sepa­ flexural stiffness of the beam itself. However, it is reasonable expecting
rately evaluated the influence of distributed vertical loads acting on the that beams with high flexural stiffness could reduce the influence of the
beams (e.g. representing a load form the slab) and concentrated vertical distributed vertical loads on the OOP resistance.
loads acting on the columns (e.g. representing loads from the upper A final investigation has been carried out simulating the out-of-plane
stories). The distributed load (q) was varied in the range 0–30 kN/m. In tests of the previously considered specimens (OOP/4E/80 [14], OOP/
order to assess also the effect of the vertical loads acting on the columns, 4E/120 [15], OOP/120 [16]) by applying a uniform load pressure
a parallel investigation was carried out by applying the same resulting instead of the original 4-point loading. This investigation aims to check
vertical force (Qb=ql) as a half per column and removing the load from the consistency between the outcomes by the two main modalities of
the beam. The two modes of application of vertical loads are depicted in application of the horizontal load in OOP experimental testing of ma­
Fig. 18. sonry infills, namely by a 4-point device or by an airbag. The two
The rate of vertical load is summarized by the general dimensionless different load modalities are introduced in the reference FE models as
parameter (ν) defined as: depicted in Fig. 20.
Results of the tests are shown in Fig. 21 for the different specimens.
Q
ν= (24) The latter clearly highlight a significant dependence of the out-of-plane
Ac,tot ⋅fcm + l⋅t⋅fmv
resistance of the infills on the mode of application of the horizontal load.
The latter represents the ratio of axial force with respect to the total In particular, results show a doubling of resistance in the case of uniform
vertical bearing capacity of the infill-frame system. In Eq. (24) Q is the loading. This can be justified by the fact that a uniform load distribution
total vertical load acting on the system, that is: involves the whole infill in the resisting mechanism. Damage localiza­
tion due to point-loading badly conditions the overall resistance. This
Q = Qc + Qb (25)
effect is significantly reduced in the case of a uniform pressure and6
because of this, the OOP capacity results larger.
where Qc is the sum of the axial forces acting on the columns and Qb is
The outcomes of this investigation bring to a major consideration.
the resulting force by the distributed vertical load (q) acting of the beam
Namely that it is not correct, as it has been done in previous studies,
(Qb=ql). The term Ac,tot is the sum of the cross-sectional areas of the
comparing and collecting together experimental test data from the two
columns. For specimens having identical columns as the current ones Ac,
different modalities of application the horizontal loads. A proper con­
tot=2bchc, where bc and hc are the base and the height of the columns
version factor should be used to make the data comparable. The large
cross-section respectively. Parametric studies here considered were
dispersion recognized by applying the existing empirical formulas to
carried out by applying Qc or Qb loads to get the same values of ν, namely
determine the OOP capacity is probably mainly due to this inconsis­
0.56%, 1.11% and 1.67% (Table 6). Results of the tests showed that
tency, therefore, the following section will provide a new empirical
vertical loads applied on the beam (Fig. 19a) have a noticeable effect in
formula taking into account both the modes of application of horizonal
increasing the OOP capacity. A linear increment of the OOP resistance is
and vertical loads as well as other relevant parameters.
observed as a function of the extent of the vertical load (Fig. 16c). This
trend is justified by the pre-stressing action exerted on the infill by the
compression load, which makes the arching mechanism more effective. 5. Empirical formulation of the out-of-plane resistance capacity
On the contrary, the variation of vertical loads applied on the columns model
have shown to not influence the OOP peak load of the infill wall
(Fig. 19b). This is also clearly observable from the constant trend of the 5.1. Dataset definition
OOP capacity (as depicted in Fig. 16c) in comparison with the one
previously observed case. The negligible effect of the vertical loads The formulation of the new empirical out-of-plane capacity model is
applied on the columns to the OOP resistance of the infills is related to carried out defining a robust database of properly selected tests. The
the major axial stiffness of the columns. In fact, axial deformations are new database is a hybrid database, namely it is composed of 23 exper­
quite low, and this limits the effective load transfer from the column to imental tests [2,4,12–18,37–42] augmented by the 15 numerical simu­
lations above presented. The composition of the dataset provided to

Fig. 18. Modes of application of vertical load: a) on the beam; b) on the columns.

11
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 19. Effect of dimensionless axial force on the OOP response of: a) Vertical load applied on the beam; b) vertical load applied on the columns.

Fig. 20. Modes of application of horizontal load: a) 4-point loading; b) uniform loading.

Fig. 21. Effect of the modality of application of the lateral load: a) Ricci et al. [14]; b) Ricci et al. [15]; c) De Risi et al. [16].

maintain the number of numerical tests not larger than 40% of the total concrete frames infilled with any masonry typology and presenting a
38 tests. In this way, experimental and numerical test are balanced with complete, or almost complete, experimental characterization of the
slight prevalence of the experimental ones. The selection of the experi­ masonry units and prisms mechanical properties. General details of
mental tests has been carried out among the available tests on reinforced selected experimental tests are reported in Table 7 as well as the

12
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Table 7
Hybrid dataset: Specimens and test details and geometric data.
Reference Specimen Masonry units Horizontal load type Vertical load type l h t bc hc h/t l/h
(mm) (mm) (mm) (mm) (mm) (–) (–)

Calvi & Bolognini [4] 10 Clay hollow 4-point on columns 4200 2750 135 300 300 20.4 1.53
Inf_01 Clay hollow Airbag on columns 4200 2300 150 300 300 15.3 1.83
Furtado et al. [12]
Inf_02 Clay hollow Airbag on columns 4200 2300 150 300 300 15.3 1.83
Akhoundi et al. [13] SIF-B Clay hollow Airbag none 2415 1635 80 160 160 20.4 1.48
Furtado et al. [37] M4 Clay hollow Airbag on columns 4200 2300 150 300 300 15.3 1.83
Ricci et al. [14] OOP/4E/80 Clay hollow 4-point none 2350 1830 80 200 270 22.9 1.28
Ricci et al. [15] OOP/4E/120 Clay hollow 4-point none 2350 1830 120 200 270 15.3 1.28
De Risi et al. [16] OOP/120 Clay hollow 4-point none 1830 1830 80 200 270 22.9 1.00
Koutas & Bournas [38] S_CON Clay solid 4-point none 1700 1250 65 200 140 19.2 1.36
Nasiri & Liu [39] IFNG Concr. hollow Airbag none 1350 980 90 180 180 10.9 1.38
Angel [2] 1 Clay solid Airbag none 2438.4 1625.6 47.6 304 304 34.2 1.50
Sepasdar [40] IF-ND Concr. hollow Airbag none 1350 980 90 180 180 10.9 1.38
Di Domenico et al. [17] 120S-OOP Clay hollow 4-point none 1830 1830 120 200 270 15.3 1.00
Agante et al. [18] LWC_Ref Concr. hollow Airbag none 4800 3300 315 300 300 10.5 1.45
Varela-Rivera et al. [41] S1 Concr. hollow Airbag none 3700 2700 150 150 200 18.0 1.37

W1 Concr. hollow Airbag on the beam 3770 2760 113 120 250 24.4 1.37
W2 Clay hollow Airbag on the beam 3770 2760 115 120 250 24.0 1.37
W3 Clay hollow Airbag on the beam 3770 2760 120 120 250 23.0 1.37
W4 Clay solid Airbag on the beam 3770 2760 114 120 250 24.2 1.37
Moreno-Herrera [42]
W5 Concr. hollow Airbag on the beam 2950 2760 113 120 250 24.4 1.07
W6 Clay hollow Airbag on the beam 2950 2760 115 120 250 24.0 1.07
W7 Clay hollow Airbag on the beam 2950 2760 120 120 250 23.0 1.07
W8 Clay solid Airbag on the beam 2950 2760 114 120 250 24.2 1.07
FEM-R2018a-L1 Clay hollow Uniform none 2350 1830 80 200 270 22.9 1.28
FEM-DR2019-T1 Clay hollow 4-point none 1830 1830 120 200 270 15.3 1.00
FEM-R2018a-T1 Clay hollow 4-point none 2350 1830 200 200 270 9.2 1.28
FEM-DR2019-T2 Clay hollow 4-point none 1830 1830 200 200 270 9.2 1.00
FEM-R2018a-R1 Clay hollow 4-point none 2350 1830 80 200 270 22.9 1.28
FEM-R2018a-R2 Clay hollow 4-point none 2350 1830 80 200 270 22.9 1.28
FEM-DR2019-R1 Clay hollow 4-point none 2350 1830 80 200 270 22.9 1.28
FEM - Analyses FEM-DR2019-R2 Clay hollow 4-point none 2350 1830 80 200 270 22.9 1.28
FEM-R2018b-R1 Clay hollow 4-point none 2350 1830 120 200 270 15.3 1.28
FEM-R2018b-R2 Clay hollow 4-point none 2350 1830 120 200 270 15.3 1.28
FEM-R2018a-Q1 Clay hollow 4-point on the beam 2350 1830 80 200 270 22.9 1.28
FEM-R2018a-Q2 Clay hollow 4-point on the beam 2350 1830 80 200 270 22.9 1.28
FEM-R2018a-Q3 Clay hollow 4-point on the beam 2350 1830 80 200 270 22.9 1.28
FEM-R2018b-L1 Clay hollow Uniform none 2350 1830 120 200 270 15.3 1.28
FEM-DR2019-T1 Clay hollow Uniform none 1830 1830 80 200 270 22.9 1.00

geometrical characteristics of specimens. In Table 7 the loading mo­ 5.2. Formulation of the empirical capacity model for the OOP resistance
dality (4-point or uniform) is also specified. 4-point loading tests provide
positioning the loading points at 1/3 and 2/3 of the length of each di­ Tables 7 and 8 associate geometrical and mechanical data of the
agonal of the infill. Uniform loading tests are carried out by an airbag specimens with the outcomes of the tests in terms of maximum OOP
(real specimens) or a uniform pressure (FE models). Experimental me­ resistance (FOOP). Data processing was performed to derive an empirical
chanical properties of specimens are listed in Table 8. For some of the analytical relationship between the out-of-plane resistance and the most
selected tests, experimental test data of the masonry (e.g. the influencing geometric and mechanical features of a generic infilled RC
compressive strength of masonry prims or the compressive strength of frame.
the units) were available only in one direction. In these cases, the cor­ In consideration of past experimental and numerical studies and of
responding values in the orthogonal direction were extrapolated based the previously presented parametric analysis, the search for a new
on the void percentage ratios along the two directions or proportionally empirical formulation considered the following major parameters:
to already known horizontal/vertical strength ratios (e.g. in the cases in aspect ratio of the infill (l/h), area of the infill (l⋅h), slenderness of the
which complete data in the two directions were known only for the infill (h/t), conventional resistance of the units (̃f ), conventional
b
units). This approximation may be considered strong, but it is necessary
resistance of the masonry (̃f m ), dimensionless vertical load (ν) and mode
to avoid discarding too many specimens from the dataset. Experimental
of application of the horizontal load (α). The area of the infill is a
test data are augmented by the results of the 15 numerical tests previ­
parameter introduced to consider the scale factor affecting the response
ously carried out with the refined FE models. The addition of further
of small specimens with respect to the largest ones. In fact, the dataset
numerical tests allows the dataset being more robust, especially with
included specimens ranging from as scale factor of 1:3 (1.22 m2) to 1:1
respect to some parameters having a minor representativeness (e.g. the
(11.55 m2). As regards the coefficient α, the latter has the scope of
mode of application of the horizontal and vertical loads). Details of the
uniformizing OOP test results obtained by 4-point load tests and airbag
numerical specimens are also reported in Tables 7 and 8.
tests (uniform loading). The proposed predictive relationship is formu­
In Table 8, the resulting vertical loads on the beams or on the col­
lated in terms of maximum OOP force. In consideration of the above
umns are also reported together with the respective dimensionless
mentioned parameters, the proposed shape for the formula is the
values defined as in Eq. (21). The maximum out-of-plane forces (FOOP)
following:
associated with the single tests are finally also included in Table 8.
( )a2 ( )a3 ( )a4
l⋅h l l a5
FOOP = a1 ⋅α⋅ ⋅ ⋅ ⋅̃f ⋅(1 + νeff )a6 (26)
100 h t

13
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Table 8
Hybrid dataset: Specimens and test mechanical properties and maximum OOP loads.
Reference fcm Ecm Emv Emh fmv fmh ̃f fbv fbh ̃f fmm Qc Qb = v FOOP
m b
ql
(MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (kN) (kN) (%) (kN)

Calvi & Bolognini [4] 29.3 30378.9 1873.0 991.0 1.10 1.11 1.10 15.40 2.80 6.57 5.54 800 – 13.6 33.7
26.8 24700.0 1417.6 3116.3 0.53 0.67 0.60 4.10 5.93 4.93 16.55 600 – 11.6 76.0
Furtado et al. [12]
26.8 24700.0 1417.6 3116.3 0.53 0.67 0.60 4.10 5.93 4.93 5.66 600 – 11.6 69.0
Akhoundi et al. [13] 25.0 29962.0 – – 1.00 2.97a 1.72 2.75 8.17 4.74 – – – – 39.7
Furtado et al. [37] 21.4 24300.0 941.9 – 1.10 1.59a 1.32 4.10 5.93 4.93 8.76 540 – 11.9 46.4
Ricci et al. [14] 36.0 32308.2 1090.0 1255.0 1.81 2.45 2.11 2.00 5.00 3.16 8.3 – – – 22.0
Ricci et al. [15] 42.2 33885.7 1455.0 1262.0 1.65 2.12 1.87 2.00 5.00 3.16 8.3 – – – 41.9
De Risi et al. [16] 42.9 34053.4 1891.0 3452.0 2.37 4.63 3.31 2.00 5.00 4.74 9.44 – – – 29.1
Koutas & Bournas [38] 22.1 27908.7 2500.0 2500.0 9.70 9.70b 9.70 21.20 21.20 21.20 11.5 – – – 29.0
Nasiri & Liu [39] 43.8 27800.0 3000.0 – 17.10 24.93a 20.65 25.00 36.45 30.19 21.3 – – – 140.0
Angel et al. [2] 55.4 24821.0 7846.2 7846.2 11.51 11.51b 11.51 23.90 23.90 23.90 – – – – 33.9
Sepasdar [40] 38.5 16911.0 7990.0 – 9.40 13.70b 11.35 12.80 18.66b 15.45 20.4 – – – 87.7
Di Domenico et al. [17] 42.9 34053.4 – – 1.29 1.64 1.45 2.00 5.00 3.16 9.44 – – – 33.3
Agante et al. [18] 22.9 24300.0 3251.0 – 1.82 1.43b 1.61 2.60 2.04b 2.31 6.58 – – – 229.6
Varela-Rivera et al. [41] 14.8 9614.0 4379.0 – 2.84 2.13b 2.46 5.47 4.10b 4.74 2.89 – – – 87.8
22.8 28171.1 3720.0 – 3.72 2.79b 3.22 6.58 4.94b 5.70 3.43 – 37.7 1.28 91.7
25.9 29276.1 6480.0 – 6.48 4.76b 5.55 16.33 11.99b 13.99 3.43 – 37.7 0.86 109.2
24.4 28750.1 6170.0 – 6.17 4.10b 5.03 18.90 12.57b 15.41 3.43 – 37.7 0.89 115.1
25.9 29279.5 4150.0 – 4.15 4.15b 4.15 10.64 10.64b 10.64 3.43 – 37.7 1.13 76.3
Moreno-Herrera [42]
23.9 28575.7 3720.0 – 3.72 2.79b 3.22 6.58 4.94b 5.70 3.43 – 29.5 1.10 109.4
25.7 29187.7 6480.0 – 6.48 4.76b 5.55 16.33 11.99b 13.99 3.43 – 29.5 0.79 143.4
25.2 29015.9 6170.0 – 6.17 4.10b 5.03 18.90 12.57b 15.41 3.43 – 29.5 0.80 147.0
25.0 28974.3 4150.0 – 4.15 4.15b 4.15 10.64 10.64b 10.64 3.43 – 29.5 1.02 115.9

FEM-R2018a-L1 36.0 32308.2 1090.0 1255.0 1.81 2.45 2.11 2.00 5.00 3.16 – – – – 45.4
FEM-DR2019-T1 42.9 34053.4 1891.0 3452.0 2.37 4.63 3.31 2.00 5.00 4.74 – – – – 57.0
FEM-R2018a-T1 36.0 32308.2 1090.0 1255.0 1.81 2.45 2.11 2.00 5.00 3.16 – – – – 101.4
FEM-DR2019-T2 42.9 34053.4 1891.0 3452.0 2.37 4.63 3.31 2.00 5.00 4.74 – – – – 133.7
FEM-R2018a-R1 36.0 32308.2 545.0c 627.5c 0.91c 1.23c 1.05 1.00 2.50 1.58 – – – – 16.3
FEM-R2018a-R2 36.0 32308.2 2180.0c 2510.0c 3.62c 4.90c 4.21 4.00 10.00 6.32 – – – – 29.6
FEM-DR2019-R1 42.9 34053.4 945.5c 1726.0c 1.19c 2.32c 1.66 1.00 2.50 2.37 – – – – 21.5
FEM-DR2019-R2 42.9 34053.4 3782.0c 6904.0c 4.74c 9.26c 6.63 4.00 10.00 9.48 – – – – 39.0
FEM-R2018b-R1 42.2 33885.7 727.5c 631.0c 0.83c 1.06c 0.94 1.00 2.50 1.58 – – – – 32.1
FEM-R2018b-R2 42.2 33885.7 2910.0c 2524.0c 3.30c 4.24c 3.74 4.00 10.00 6.32 – – – – 58.2
FEM-R2018a-Q1 36.0 32308.2 1090.0 1255.0 1.81 2.45 2.11 2.00 5.00 3.16 – – 23.5 0.56 23.3
FEM-R2018a-Q2 36.0 32308.2 1090.0 1255.0 1.81 2.45 2.11 2.00 5.00 3.16 – – 47 1.11 24.7
FEM-R2018a-Q3 36.0 32308.2 1090.0 1255.0 1.81 2.45 2.11 2.00 5.00 3.16 – – 70.5 1.67 26.0
FEM-R2018b-L1 42.2 33885.7 1455.0 1262.0 1.65 2.12 1.87 2.00 5.00 3.16 – – – – 76.4
FEM-DR2019-T1 42.9 34053.4 1891.0 3452.0 2.37 4.63 3.31 2.00 5.00 4.74 – – – – 58.2
a
Evaluated as proportional to the strength ratio fbh/fbv.
b
Evaluated as inversely proportional to the void ratios along the directions of the units.
c
Evaluated as proportional to the conventional units compressive strength.

where αtake the values: coefficients a1 , a2 , a3 , a4 , a5 , a6 , a7 , a8 , a9 , a10 provide the weight of


⎧ each parameter to the OOP resistance. The latter are calibrated by means
⎨1
⎪ 4 − point load of an optimization protocol minimizing the dispersion of predicted re­
α=
( )a8
l (27) sults with respect to experimental ones. In detail the objective function

⎩ a7 ⋅
h
uniform load (OF) to be minimized considers the sum of the squared differences be­
pred
tween results predicted by the model (FOOP,i (X)) according to Eq. (26)
and ν eff is the effective dimensionless axial load, defined as: exp
and the respective experimental values (FOOP,i ) as follows:
a9 ⋅Qb + a10 ⋅Qc
(28) ∑
n
νeff = ( )2
Ac,tot ⋅fcm + l⋅t⋅fmv OF(X) = pred
FOOP,i exp
(X) − FOOP,i (30)
i=1
In Eqs. (26)–(28) input dimensions are expressed in millimeters,
input forces are expressed in Newtons, while the output OOP force is in where n is the number of specimens and X the design vector gathering all
kN. The term ̃f indicates a conventional strength, that can be the con­ the parameters to be calibrated as:
ventional resistance of the units (̃f b ) or conventional resistance of the X = [a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 ]T (31)
masonry (̃f ), namely:
m
Because of the possible choice on the conventional strength (̃f b or ̃f m )
〈̃ the optimization of the coefficients is provided twice, once in the case of
fb
̃f = or (29) using ̃f and once in the case of using ̃f .
b m
̃f m The unconstrained optimization problem is solved by using a genetic
algorithm to minimize the objective function in Eq. (30). The genetic
The possibility to select ̃f b or ̃f m is introduced to allow using the algorithm optimization histories converging to the optimal solutions are
model even if only one of the two parameters is known. The ten constant shown in Fig. 22 for the cases in which it was assumed ̃f = ̃f or ̃f = ̃f . b m

14
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 22. Objective function minimization by the genetic algorithm for the calibration of the constant coefficients: a) Eqs. (22–24) with ̃f = ̃f b ; b) Eqs. (22–24) with
̃f = ̃f .
m

Table 9
Optimal coefficients for Eqs. (26)–(28) in the cases of adopting ̃f b or ̃f m .
̃f a1 a2 a3 a4 a5 a6 a7 a9 a9 a10

̃f 9.77 0.57 − 1.23 − 1.71 0.31 18.33 2.27 − 0.72 0.0 1.0
b
̃f 3.82 0.62 − 1.13 − 1.51 0.36 17.99 2.05 − 0.51 0.0 1.0
m

The finally obtained coefficients, to use in Eq. (26)–(28) are reported in In the case of adopting ̃f m as strength parameter the predictive
Table 9. It is noteworthy observing that coefficient a9, modulating the equation is instead:
effect of the axial load on the columns took the value 0 in both the cases, ( )0.62 ( )− 1.13 ( )− 1.51
while a10 was 1. This means that the axial force on the columns is not w⋅l l h 0.36
FOOP = 3.83⋅α⋅ ⋅ ⋅ ⋅̃f m ⋅(1 + νeff )17.99 (34)
influencing the response, and therefore, the dimensionless axial force ν 100 h t
eff coincides with the one associated only with the loads on the beam.
The predictive equations for the OOP resistance are specialized as with:
described in the following. ⎧
⎨1
⎪ 4 - point load
In the case of adopting ̃f as strength parameter one obtains:
b α= ( )− 0.51 (35)
⎪ l
( )0.57 ( )− 1.23 ( )− 1.71 ⎩ 2.05⋅ uniform load
l⋅h l h 0.31 h
FOOP = 9.77⋅α⋅ ⋅ ⋅ ⋅̃f b ⋅(1 + νeff )18.33 (32)
100 h t
In both cases the effective dimensionless vertical load (νeff) is evalu­
ated considering only vertical loads acting on the beams (νeff = νb) so that:
with:
⎧ Qb
νeff = νb = (36)
⎨1
⎪ 4 - point load
Ac,tot fcm + l⋅t⋅fmv
α= ( )− 0.72 (33)
l

⎩ 2.27⋅ uniform load The performances of the proposed model are shown in Fig. 23 both in
h
the cases of using ̃f b or ̃f m as strength parameters. It can be observed that

Fig. 23. Proposed model performances: a) with ̃f b (Eq. (28)); b) with ̃f m (Eq. (30)).

15
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

in both cases the model provided a quite good predictive capacity. In ultimate pressure (qOOP), the latter was multiplied by the area of the
addition, in Fig. 23 the bisector lines (representing the exact corre­ infill (l × h) to be converted in an ultimate load. The reference tests
spondence between experimental values and numerical predictions) are values and the predicted OOP resistances by the different models are
compared with the regression lines associated with the distribution of shown in Table 10, together with the respective predicted/experimental
data, showing that the two line almost completely overlapped. Statisti­ (P/E) ratios. In the same table, averages (μ), standard deviations (σ ) and
cal data and comparison of the proposed model with the literature coefficients of variations (COV) of P/E ratios are also reported. Results
available ones are provided in the subsequent section. are also graphically represented in Fig. 25. The comparison of the model
Besides the predictive performance, one of the major advances of the performances highlights that, on average, the predictive capacity of the
model is the possibility to use it even if the infill mechanical properties formulas proposed during the time has improved. It can be observed that
are partially known, namely if only the strengths of the units or the early proposed formulations (e.g. Dawe and Seah [1], Angel [2], FEM
strengths of the masonry along the two directions are known. 356 [7], which are based on a limited experimental background have
A further comment should be done on the validity of the proposed low predictive reliability against all the tests of the database. The
model with respect to the real lateral load conditions. In fact, the actual average P/E ratios are in a range 0.22–0.56, denoting significant un­
distribution of out-of-plane inertial forces follows the modal shape of the derestimation. More recent formulations (Ricci et al. [24], Liberatore
infill, while the model is calibrated on the experimental tests carried out et al. [9], Pradhan et al. [27]) improved the average predictive capacity.
by applying point loads by an actuator or uniform loads by an airbag. Average P/E ratios range between 0.73 and 1.29, however they are
Based on the parametric study previously shown, it is reasonable characterized by a significant dispersion of results, as denoted by the
thinking that these two conditions, considered with the coefficient α, are large standard deviations (0.34–0.53) and coefficient of variations
boundary conditions. Therefore, the most likely OOP resistance (41–46%). It is also noteworthy observing that Eurocode 6 [30] formula,
considering the real distribution of the inertial forces can be obtained by despite being not properly recent, performed similarly to the most recent
averaging the two outcomes from the α coefficients in Eqs. (33) and (35). formulations, with μ=0.75, σ =0.36 and COV=48%.
This results in: For what concerns the formulation proposed in the paper, the latter
⎧ ( )− 0.72 definitively showed a significant improvement of the predictive capac­

⎪ l ity. The two versions of the formula had similar performances with

⎪ 0.5 + 1.135⋅
⎨ if ̃f b is used
h
α= (37) μ=1.01, σ =0.16 and COV=16% for Eq. (28) (based on ̃f ) and μ=1.03,
b
⎪ ( )− 0.51



⎩ 0.5 + 1.025⋅
l
if ̃f m is used =0.15 and COV=14% for Eq. (30) (based on ̃f m ). The significantly
h improved predictive capacity of the proposed model with respect to the
A sample of the application of Eq. (37) average α coefficients with previous ones, can be justified by different aspects. First, the dataset
Eqs. (32) and (34) is shown in Fig. 24, together with the boundary limits. used for the calibration was restricted only to infilled RC frame speci­
mens. This reduced the heterogeneity of the used specimens. Moreover,
6. Predictive performance of the proposed model with respect to the tests were selected among those having a complete experimental
the existing ones characterization of masonry and units. Second, the model considers
more parameters than the other formulas. In particular, parameters
A comparative analysis of the proposed empirical relationship with related to the modality of application of vertical and horizontal loads
respect to the predictive models available in the literature is finally have shown to play a relevant role. The new model can uniformize re­
carried out. Experimental and numerical test data by the dataset are sults by tests carried out using a 4-point loading device or an airbag and
used for the comparison. Compared models are those by Dawe and Seah point load on the columns or uniform load on the beams. Finally, the
[1], Angel [2], Bashandi et al. [5], FEMA 356 [7], Eurocode 6 [30], Ricci dataset has been enriched with 15 numerical specimens. This allowed
et al. [24], Liberatore et al. [9], Pradhan et al. [27] and the proposed one simulating and investigating additional geometrical and mechanical
configuration of the specimens, making the database more robust.
in the cases of adopting ̃f or ̃f as conventional strengths (Eqs. (32) and
b m A final comparison of the models is proposed in Fig. 26, where the
(34)). Reference OOP resistance values, both experimental and numer­
frequency distributions of each model are depicted. The latter are fitted
ical, are here generally named as experimental for the sake of brevity.
by a lognormal distribution. This kind of representation of data ex­
Comparisons are carried out in terms of ultimate out-of-plane load
presses more clearly the tendencies above described. A more concise
(FOOP). For the models providing the ultimate OOP resistance as an
comparison is illustrated in Fig. 27, where the fitting lognormal PDFs of

Fig. 24. Proposed model boundary and average OOP values: a) with ̃f b ; b) with ̃f b . It is assumed l = 1000 mm, h = 1000 mm,̃f b = 1 MPa,̃f m = 1 MPa,νeff = 0.

16
F. Di Trapani et al.
Table 10
Reference (Exp.) and predicted (Pred.) maximum OOP strengths (FOOP) by the different models and predicted/experimental (P/R) ratios.
Dawe & Seah (1989) Angel (1994) Bashandy. (1995) FEMA 356 Eurocode 6 Ricci (2018a) Liberatore (2020) Pradhan (2021) Proposed Eq. (28) Proposed Eq. (30)

Test Exp. Pred. P/E Pred. P/E Pred. P/E Pred. P/E Pred. P/E Pred. P/E Pred. P/E Pred. P/E Pred. P/E Pred. P/E
(kN) (kN) (-) (kN) (-) (kN) (-) (kN) (-) (kN) (-) (kN) (-) (kN) (-) (kN) (-) (kN) (-) (kN) (-)

Calvi & Bolognini [4] 33.7 21.6 0.64 22.4 0.66 36.6 1.08 8.6 0.26 30.6 0.91 48.3 1.43 52.6 1.56 19.8 0.59 44.8 1.33 34.4 1.02
76.0 17.5 0.23 17.6 0.23 30.8 0.40 6.9 0.09 21.8 0.29 62.8 0.83 27.1 0.36 11.5 0.15 71.0 0.93 46.8 0.62
Furtado et al. [12]
69.0 17.5 0.25 17.6 0.25 30.8 0.45 6.9 0.10 21.8 0.32 62.8 0.91 27.1 0.39 11.5 0.17 71.0 1.03 46.8 0.68
Akhoundi et al. [13] 39.7 6.3 0.16 3.2 0.08 8.8 0.22 2.7 0.07 9.5 0.24 32.4 0.82 17.0 0.43 6.3 0.16 39.1 0.98 36.2 0.91
Furtado et al. [37] 46.4 30.1 0.65 36.1 0.78 49.6 1.07 14.3 0.31 45.2 0.97 81.1 1.75 52.2 1.12 22.6 0.49 71.0 1.53 62.3 1.34
Ricci et al. [14] 22.0 14.4 0.66 6.8 0.31 6.7 0.30 4.0 0.18 14.9 0.68 31.1 1.41 31.6 1.44 14.9 0.68 20.6 0.94 24.0 1.09
Ricci et al. [15] 41.9 30.7 0.73 20.2 0.48 39.1 0.93 9.6 0.23 30.5 0.73 57.4 1.37 47.9 1.14 28.3 0.68 41.3 0.99 42.5 1.01
De Risi et al. [16] 29.1 19.7 0.67 7.1 0.24 16.1 0.55 4.1 0.14 15.2 0.52 26.6 0.91 40.3 1.38 26.1 0.90 27.6 0.95 32.1 1.10
Koutas & Bournas [38] 29.0 29.0 1.00 19.3 0.67 8.2 0.28 16.0 0.55 55.7 1.92 61.4 2.12 82.7 2.85 51.2 1.77 31.5 1.09 32.7 1.13
Nasiri & Liu [39] 140.0 126.9 0.91 91.4 0.65 117.3 0.84 69.2 0.49 190.8 1.36 160.8 1.15 170.5 1.22 172.5 1.23 126.3 0.90 130.0 0.93
Angel et al. [2] 33.9 25.2 0.74 11.3 0.33 11.4 0.34 8.9 0.26 39.1 1.15 34.1 1.00 80.5 2.38 34.3 1.01 26.2 0.77 32.1 0.95
Sepasdar [40] 87.7 71.5 0.82 47.4 0.54 151.6 1.73 38.1 0.43 104.9 1.20 130.4 1.49 99.5 1.13 89.3 1.02 102.4 1.17 104.9 1.20
Di Domenico et al. [17] 33.3 28.1 0.84 12.3 0.37 34.1 1.02 5.9 0.18 18.6 0.56 41.0 1.23 38.4 1.15 22.8 0.69 48.7 1.46 44.2 1.33
Agante et al. [41] 229.6 137.4 0.60 231.2 1.01 409.1 1.78 96.9 0.42 262.7 1.14 177.1 0.77 269.9 1.18 176.7 0.77 221.6 0.97 234.2 1.02
Varela-Rivera et al. [41] 87.8 26.1 0.30 31.4 0.36 115.6 1.32 25.9 0.29 87.6 1.00 72.7 0.83 138.6 1.58 70.0 0.80 95.1 1.08 99.7 1.14
91.7 25.8 0.28 20.5 0.22 47.2 0.51 16.9 0.18 64.9 0.71 49.7 0.54 126.8 1.38 55.2 0.60 77.7 0.85 89.7 0.98
109.2 41.0 0.38 38.3 0.35 88.3 0.81 30.7 0.28 117.1 1.07 62.1 0.57 213.6 1.96 105.0 0.96 98.4 0.90 104.0 0.95
115.1 42.8 0.37 41.8 0.36 99.1 0.86 32.4 0.28 121.4 1.05 65.3 0.57 215.3 1.87 107.9 0.94 109.5 0.95 107.5 0.93
76.3 28.8 0.38 23.8 0.31 54.6 0.72 19.3 0.25 73.7 0.97 52.4 0.69 141.5 1.86 63.3 0.83 93.3 1.22 96.9 1.27
Moreno-Herrera [42]
17

109.4 28.0 0.26 16.1 0.15 50.8 0.46 13.2 0.12 50.8 0.46 38.9 0.36 126.8 1.16 58.7 0.54 105.7 0.97 111.7 1.02
143.4 44.2 0.31 30.0 0.21 94.6 0.66 24.0 0.17 91.6 0.64 48.6 0.34 213.6 1.49 111.7 0.78 136.3 0.95 131.8 0.92
147.0 46.4 0.32 32.8 0.22 105.3 0.72 25.4 0.17 95.0 0.65 51.1 0.35 215.3 1.46 114.7 0.78 151.3 1.03 135.9 0.92
115.9 31.1 0.27 18.6 0.16 58.6 0.51 15.1 0.13 57.6 0.50 41.0 0.35 141.5 1.22 67.3 0.58 128.4 1.11 122.1 1.05

45.4 14.4 0.32 6.8 0.15 6.7 0.15 4.0 0.09 14.9 0.33 31.1 0.68 31.6 0.70 14.9 0.33 39.1 0.86 43.4 0.96
57.0 44.4 0.78 22.7 0.40 58.4 1.03 10.8 0.19 34.1 0.60 50.7 0.89 66.3 1.16 56.5 0.99 55.3 0.97 59.3 1.04
101.4 90.9 0.90 66.1 0.65 140.9 1.39 36.2 0.36 93.0 0.92 133.5 1.32 97.6 0.96 85.1 0.84 99.0 0.98 95.9 0.95
133.7 124.1 0.93 68.9 0.52 199.8 1.49 36.9 0.28 94.8 0.71 114.3 0.85 124.3 0.93 149.1 1.12 132.5 0.99 128.3 0.96
16.3 8.6 0.53 3.4 0.21 3.4 0.21 2.0 0.12 7.4 0.46 24.4 1.50 16.9 1.04 7.0 0.43 16.6 1.02 18.7 1.15
29.6 24.3 0.82 13.7 0.46 13.4 0.45 8.0 0.27 29.8 1.01 39.6 1.34 59.0 1.99 32.0 1.08 25.6 0.87 30.8 1.04
21.5 10.6 0.49 4.6 0.21 6.7 0.31 2.6 0.12 9.7 0.45 26.8 1.25 21.6 1.00 11.5 0.53 18.9 0.88 22.0 1.03
FEM - Analyses 39.0 30.1 0.77 18.3 0.47 26.8 0.69 10.4 0.27 39.0 1.00 43.6 1.12 75.2 1.93 52.6 1.35 29.1 0.75 36.2 0.93
32.1 18.3 0.57 10.1 0.32 19.6 0.61 4.8 0.15 15.3 0.48 45.0 1.40 25.7 0.80 13.2 0.41 33.2 1.04 33.2 1.03
58.2 51.7 0.89 40.4 0.69 78.3 1.35 19.3 0.33 61.0 1.05 73.1 1.26 89.4 1.54 60.7 1.04 51.3 0.88 54.5 0.94
23.3 14.4 0.62 6.8 0.29 6.7 0.29 4.0 0.17 14.9 0.64 31.1 1.33 31.6 1.36 14.9 0.64 22.8 0.98 26.5 1.14
24.7 14.4 0.58 6.8 0.28 6.7 0.27 4.0 0.16 14.9 0.60 31.1 1.26 31.6 1.28 14.9 0.60 25.3 1.02 29.3 1.19

Engineering Structures 266 (2022) 114422


26.0 14.4 0.55 6.8 0.26 6.7 0.26 4.0 0.15 14.9 0.57 31.1 1.19 31.6 1.21 14.9 0.57 27.9 1.07 32.4 1.24
76.4 30.7 0.40 20.2 0.26 39.1 0.51 9.6 0.13 30.5 0.40 57.4 0.75 47.9 0.63 28.3 0.37 78.3 1.03 76.8 1.01
58.2 19.7 0.34 7.1 0.12 16.1 0.28 4.1 0.07 15.2 0.26 26.6 0.46 40.3 0.69 26.1 0.45 62.8 1.08 65.9 1.13

μ 0.56 μ 0.38 μ 0.71 μ 0.22 μ 0.75 μ 1.01 μ 1.29 μ 0.73 μ 1.01 μ 1.03
σ 0.24 σ 0.21 σ 0.44 σ 0.12 σ 0.36 σ 0.42 σ 0.53 σ 0.34 σ 0.16 σ 0.15
COV 0.43 COV 0.55 COV 0.63 COV 0.53 COV 0.48 COV 0.42 COV 0.41 COV 0.46 COV 0.16 COV 0.14
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 25. Comparison of the predictive capacity of the considered models.

18
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

Fig. 26. Predicted/Experimental OOP strength ratios probabilistic distributions.

Fig. 27. Comparison of Predicted/Experimental OOP strength ratios PDF by the different models.

19
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

the different models are overlapped. with or simulated tests on the specimens.

7. Conclusions CRediT authorship contribution statement

Assessment of out-of-plane resistance of masonry infilled frames is Fabio Di Trapani: Conceptualization, Methodology, Writing – re­
quite important to perform safety verifications of infills against view & editing, Supervision. Alessandro Vizzino: Formal analysis,
earthquake-induced OOP forces. However, available literature models, Writing – review & editing. Giovanni Tomaselli: Formal analysis,
as well as technical-code models for the prediction of the out-of-plane Writing – review & editing. Antonio Pio Sberna: Formal analysis,
resistance provide conflicting results, being in general too conserva­ Writing – review & editing. Gabriele Bertagnoli: Writing – review &
tive or, on the contrary, overestimating the capacity. The paper critically editing, Supervision.
analysed the reasons of these inconsistencies, with special reference to
infilled reinforced concrete frame, concluding that:
Older literature models have been calibrated based on too limited Declaration of Competing Interest
experimental data;
More recent ones have been calibrated on too heterogenous datasets, The authors declare that they have no known competing financial
including also confined masonry and steel surrounding frames; interests or personal relationships that could have appeared to influence
None of the available models has been calibrated considering a the work reported in this paper.
proper homogenization of the tests, which have been conducted by
adopting different modalities to apply vertical loads (e.g. on the columns Acknowledgements
or on the beams) and horizontal loads (e.g. point loads of uniform loads).
The paper proposed a new empirical formulation for the determi­ This paper was supported by DPC-ReLuis 2019–2021, WP10, Subtask
nation of the out-of-plane resistance of masonry infills in RC frames 10.1.2 Non-structural masonry (infills and partitions).
without prior in-plane damage. An hybrid database, composed of 23
experimental tests and 15 numerical simulations by a refined FE micro- References
model was specifically defined. FE models allowed making the dataset
[1] Dawe JL, Seah CK. Out-of-plane resistance of concrete masonry infilled panels. Can
more robust and investigating on the influence of some relevant pa­
J Civ Eng 1989;16(6):854–64.
rameters. Results from the parametric study allowed concluding that: [2] Angel RE. Behavior of reinforced concrete frames with masonry infill walls.
University of Illinois at Urbana-Champaign; 1994. PhD Thesis, Illinois.
[3] Flanagan RD, Bennett RM. Bidirectional behavior of structural clay tile infilled
– Vertical loads acting on the columns do not influence the OOP
frames. ASCE J Struct Eng 1999;125(3):236–44.
resistance, which is instead influenced by the loads acting on the top [4] Calvi GM, Bolognini D. Seismic response of reinforced concrete frames infilled with
beam; weakly reinforced masonry panels. J Earthquake Eng 2001;5(2):153–85.
– The modality of application of the horizontal load significantly in­ [5] Bashandy T, Rubiano NR, Klingner RE. Evaluation and analytical verification of
infilled frame test data. P.M. Ferguson Structural Engineering Laboratory, Report
fluences the OOP resistance. In particular, uniform loading tests No.95-1, Department of Civil Engineering University of Texas at Austin, Austin, Tx;
provided double resistance with respect to 4-point loading. 1995.
[6] Abrams DP, Angel R, Uzarski J. Out-of-plane strength of unreinforced masonry
infill panels. Earthquake Spectra 1996;12(4):825–44.
Postprocessing of data and optimization of empirical coefficients [7] FEMA 356. Prestandard and commentary for the seismic rehabilitation of
allowed defining a new relationship for the direct estimation of the OOP buildings, Washington DC; 2000.
resistance of a generic infilled frame. The proposed model showed to [8] Asteris PG, Cavaleri L, Di Trapani F, Tsaris AK. Numerical modelling of out-of-lane
response of inflled frames: state of the art and future challenges for the equivalent
perform better the available ones and significantly reducing the srut macromodels. Eng Struct 2017;32:110–22.
dispersion. The reasons of its better capability in estimating experi­ [9] Liberatore L, AlShawa O, Marson C, Pasca M, Sorrentino L. Out-of-plane capacity
mental results is justified by the following major considerations: equations for masonry infill walls accounting for openings and boundary
conditions. Eng Struct 2020;207:110198. https://doi.org/10.1016/j.
engstruct.2020.110198.
– The model takes into account more relevant parameters than avail­ [10] Da Porto F, Guidi G, Benetta MD, Verlato N. Combined in-plane/out-of-plane
able formulas. experimental behaviour of reinforced and strengthened infill masonry walls. In:
The masonry society, 12th Canadian masonry symposium; 2013.
– The introduction of a conversion factor (α) allows uniformizing
[11] Hak S, Morandi P, Magenes G. Out-of-plane experimental response of strong
experimental tests carried out with point-load or uniform-load masonry infills. 2nd European conference on earthquake engineering and
devices; seismology; 2014.
– The introduction of the effective dimensionless axial force (ν eff) al­ [12] Furtado A, Rodrigues H, Arêde A, Varum H. Experimental evaluation of out-of-
plane capacity of masonry infill walls. Eng Struct 2016;111:48–63.
lows considering the increase of the arching effect as a function of the [13] Akhoundi F, Vasconcelos G, Lourenço P. Experimental out-of-plane behavior of
stress-state on the infill due to the vertical loads on the beam; brick masonry infilled frames. Int J Arch Heritage 2018;18:1–7.
[14] Ricci P, Di Domenico M, Verderame GM. Experimental assessment of the in-plane/
– The model makes use of the conventional strengths ̃f and ̃f , which
b m out-of-plane interaction in unreinforced masonry infill walls. Eng Struct 2018;173:
consider the orthotropic behaviour of the masonry and of the units. 960–78.
[15] Ricci P, Di Domenico M, Verderame GM. Experimental investigation of the
influence of slenderness ratio and of the in-plane/out-of-plane interaction on the
The possible choice between ̃f b and ̃f m as conventional strength is out-of-plane strength of URM infill walls. Const Build Mater 2018;191:507–22.
also a major advantage, that is useful in practice when only one of the [16] De Risi MT, Di Domenico M, Ricci P, Verderame GM, Manfredi G. Experimental
investigation on the influence of the aspect ratio on the in-plane/out-of-plane
two strengths can be determined by the available information. It should
interaction for masonry infills in RC frames. Eng Struct 2019;189:523–40.
be finally said that the actual trend of the out-of-plane inertial forces is [17] Di Domenico M, De Risi MT, Ricci P, Verderame GM, Manfredi G. Empirical
different from point loading and uniform loading, but the latter two can prediction of the in-plane/out-of-plane interaction effects in clay brick
be considered as limit boundary conditions. Because of this, the load- unreinforced masonry infill walls. Eng Struct 2021;227:111438. https://doi.org/
10.1016/j.engstruct.2020.111438.
related coefficient α should be conveniently averaged as suggested in [18] Agante M, Furtado A, Rodrigues H, Arêde A, Fernandes P, Varum H. Experimental
Eq. (37). Further research should address the actual dynamic response characterization of the out-of-plane behaviour of masonry infill walls made of

20
F. Di Trapani et al. Engineering Structures 266 (2022) 114422

lightweight concrete blocks. Eng Struct 2021;244:112755. https://doi.org/ [30] Eurocode 6. Design of structures for earthquake resistance, Part 1: General rules,
10.1016/j.engstruct.2021.112755. seismic actions and rules for buildings Masonry Structures, European Committee
[19] Pradhan B, Zizzo M, Sarhosis V, Cavaleri L. Out-of-plane behaviour of unreinforced for Standardization; Brussels, Belgium; 2005.
masonry infill walls: review of the experimental studies and analysis of the [31] Di Trapani F, Bertagnoli G, Ferrotto MF, Gino D. Empirical equations for the direct
influencing parameters. Structures 2021;33:4387–406. definition of stress-strain laws for fiber-section based macromodeling of infilled
[20] Furtado A, Rodrigues H, Arêde A, Varum H. Out-of-plane behavior of masonry frames. J Eng Mech 2018;144(11):04018101. https://doi.org/10.1061/(ASCE)
infilled RC frames based on the experimental tests available: a systematic review. EM.1943-7889.0001532.
Const Build Mater 2018;168:831–48. [32] Di Trapani F. A novel data-driven force–displacement macro-model for nonlinear
[21] Kadysiewski S, Mosalam KM. Modeling of unreinforced masonry infill walls analysis of infilled frames: development, validation and reliability comparison.
considering in-plane and out-of-plane interaction. PEER 2008/102, Univ. of Bull Earthq Eng 2021;19(14):6157–86.
California, Berkeley, CA; 2009. [33] Kent DC, Park R. Flexural members with confined concrete. J Struct Div 1971;97
[22] Mosalam KM, Günay S. Progressive collapse analysis of RC frames with URM infill (7):1969–90.
walls considering in-plane/out-of-plane interaction. Earthquake Spectra 2015;31 [34] Hsu TTC, Mo YL. Unified theory of concrete structures. John Wiley&Sons; 2010.
(2):921–43. [35] Van der Pluijm R, Rutten H, Ceelen M. Shear behaviour of bed joints. In: 12th
[23] Furtado A, Rodrigues H, Arêde A, Varum H. Simplifed macro-model for infill international brick/block masonry conference; 2000. p. 1849–62.
masonry walls considering the out-of-plane behaviour. Earthq Eng Struct Dyn [36] Vos E. Influence of loading rate and radial pressure on bond in reinforced concrete:
2016;45:507–24. a numerical and experimental approach. Delft: Delft University Press; 1983.
[24] Ricci P, Di Domenico M, Verderame GM. Empirical-based out-of-plane URM infill [37] Furtado A, Rodrigues H, Arêde A, Varum H. Effect of the panel width support and
wall model accounting for the interaction with in-plane demand. Earthq Eng Struct columns axial load on the infill masonry walls out-of-plane behavior. J Earthq Eng
Dyn 2018;47(3):802–27. 2020;24(4):653–81.
[25] Mazza F. In-plane–out-of-plane non-linear model of masonry infills in the seismic [38] Koutas LN, Bournas DA. Out-of-plane strengthening of masonry-infilled RC frames
analysis of r.c.-framed buildings. Earthq Eng Struct Dyn 2018;48(4):432–53. with textile-reinforced mortar jackets. J Compos Constr 2019;23(1):04018079.
[26] Di Trapani F, Shing PB, Cavaleri L. Macro-element model for in-plane and out-of- https://doi.org/10.1061/(ASCE)CC.1943-5614.0000911.
plane responses of masonry infills in frame structures. J Struct Eng (ASCE) 2018; [39] Nasiri E, Liu Y. Effect of prior in-plane damage on the out-of-plane performance of
144(2):04017198. https://doi.org/10.1061/(ASCE)ST.1943-541X.0001926. concrete masonry infills. Eng Struct 2020;222:111–49.
[27] Pradhan B, Sarhosis V, Ferrotto MF, Penava D, Cavaleri L. Prediction equations for [40] Sepasdar R. Experimental investigation on the out-of-plane behaviour of concrete
out-of-plane capacity of unreinforced masonry infill walls based on a masonry infilled frames. Dalhousie University; 2017. Master’s thesis.
macroelement model parametric analysis. J Eng Mech 2021;147(11):04021096. [41] Varela-Rivera J, Polanco-May M, Fernandez-Baqueiro L, Moreno EI. Confined
https://doi.org/10.1061/(ASCE)EM.1943-7889.0001998. masonry walls subjected to combined axial loads and out-of-plane uniform
[28] Cavaleri L, Zizzo M, Asteris PG. Residual out-of-plane capacity of inflls damaged by pressures. Can J Civ Eng 2012;39(4):439–47.
in-plane cyclic loads. Eng Struct 2020;209:109957. [42] Moreno-Herrera J, Varela-Rivera J, Fernandez-Baqueiro L. Out-of-plane design
[29] ABAQUS. ABAQUS theory and user manuals, version 6.14; 2014. procedure for confined masonry walls. J Struct Eng 2016;142(2):04015126.
https://doi.org/10.1061/(ASCE)ST.1943-541X.0001391.

21

You might also like