Engineering Structures 246 (2021) 113046

Contents lists available at ScienceDirect

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

On the collapse of the masonry Medici tower: An integrated

discrete-analytical approach
Micaela Mercuri a , Madura Pathirage b , Amedeo Gregori a ,∗, Gianluca Cusatis b
a
Department of Civil, Construction-Architecture and Environmental Engineering, L’Aquila, Italy
b
Department of Civil and Environmental Engineering, Northwestern University, Evanston, IL, USA

ARTICLE INFO ABSTRACT

Keywords: Masonry towers are characterized by a high susceptibility to seismic actions. For this task different approaches
Earthquake exist and they are selected depending on the desired level of accuracy of the analysis. The identification of the
Unreinforced masonry tower correct collapse configuration is however complex and necessitates thorough on-site surveys. Construction
Discrete model
codes usually require the study of local and global collapse mechanisms based on simplified kinematic
Fracture mechanics
analysis. More elaborated approaches such as nonlinear finite element methods have been used to simulate the
Kinematic analysis
Seismic forces
response of masonry towers. Although successful in many applications, these methods are limited in accurately
capturing crack distributions and fracture mechanisms. In this work, an integrated discrete-analytical approach
is proposed. First, the Lattice Discrete Particle Model (LDPM), which simulates masonry at the stone level and
has a superior capability in capturing fracturing processes, is adopted to simulate masonry towers subjected
to seismic excitation. The numerical model is used to predict the actual collapse mechanism. Next, the final
fractured configuration is used in the kinematic analysis for the calculation of the ultimate condition. The
proposed method is used to analyze the collapse of the Medici tower that collapsed during the 2009 L’Aquila
earthquake. The simulations are able to predict the induced damage and the crack contours, which are used
then to identify six different failure configurations. The subsequent kinematic analyses take into account the
relative position of openings and fracture locations. The results show that the collapse of the Medici tower is
well replicated by LDPM and the corresponding kinematic analyses demonstrate the efficiency of the proposed
hybrid approach applied to this case study. The paper also points out that different load configurations, more
specifically the direction of the seismic action, result in certain cases in more diffused damage and a clear
failure pattern cannot be identified for kinematic analyses. In these cases, it appears fundamental to rely
mainly on comprehensive numerical models, such as LDPM, to study the fracturing process from the cracks
trigger to the ultimate complex collapse mechanism.

1. Introduction response [9–11]. In addition, for the same slenderness, it was ana-
lytically demonstrated that if one considers two geometrically similar
Masonry towers represent one of the most relevant architectural blocks, the larger of the two is more stable than the smaller one [12–
category composing the world cultural heritage [1–3]. Particularly, 14]. The constituent material, namely masonry, has a heterogeneous
numerous Italian medieval towns are connoted by the presence of civic nature [15]: several studies underlined that the proprieties of mortar
towers soaring in the urban skyline [4–6]. Unfortunately, unreinforced and stone aggregate [16–19], as well as variations in the quality of the
masonry tall buildings are characterized by high susceptibility to seis- construction work [20,21], are important aspects affecting the seismic
mic actions [7,8]. Fig. 1 shows three examples of towers destroyed behavior of towers. Other features, able to ensure a satisfactory global
by severe earthquakes occurred in the last decades. The main reasons
behavior, are the effectiveness of the connections among the external
of the seismic vulnerability of these structures can be traced back
walls of the tower [22–24] and between the vertical and horizontal
to their geometric and material features. In fact, the slenderness of
components [25,26], as well as the presence of suitable elements such
the tower, i.e. the predominance of one dimension over the other
two, is one of the main characteristics that determines the overall as ties or buttresses [27] that counteract horizontal thrusts.

∗ Correspondence to: Department of Civil, Construction-Architecture and Environmental Engineering, University of L’Aquila, Via Giovanni Gronchi 18, Zona
Industriale di Pile, 67100 L’Aquila, Italy.
E-mail address: [email protected] (A. Gregori).

https://doi.org/10.1016/j.engstruct.2021.113046
Received 2 March 2021; Received in revised form 30 June 2021; Accepted 13 August 2021
Available online 4 September 2021
0141-0296/© 2021 Elsevier Ltd. All rights reserved.
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 1. (a) Bhimsen tower in Nepal, UNESCO patrimony; (b) Clock tower in Finale Emilia, Emilia Romagna, Italy; (c) Tower in Novi, Modena, Italy.

Fig. 2. Recurrent fracturing patterns observed after seismic events in unreinforced masonry towers [28].

The study of unreinforced masonry towers appears to be fundamen- modifications over the course of the years, and (ii) the full character-
tal to ensure their long-term preservation. Nowadays, a wide literature ization of the structure both in terms of geometry and material [52].
is available dealing with the analysis of these types of structures [29– With this perspective, it is possible to apply the procedure described by
32]. The international debate on Cultural Heritage preservation pointed Giuffré in 1993 [53], who illustrated a methodology for the assessment
out the importance of performing effective and extensive experimen- of the vulnerability of unreinforced masonry structures. According to
tal campaigns to assess the seismic vulnerability of masonry slender this work, the collapse mechanisms of the structure are identified a
structures, such as towers [33–38]. Several in-situ techniques exist priori, by considering the involved portions of the building as a number
and have been applied to historical buildings. Among others, it is of rigid blocks connected by unilateral hinges or sliding joints, in order
worth mentioning non-destructive testing (NDT) methods, using for to obtain a kinematic chain. Each rigid macro-elements are assumed
example georadar [34] or thermographic analysis [38] to assess dam- to have unlimited compressive strength and their reciprocal interfaces
age, or slightly destructive methodologies, such as flat-jack tests, for to be characterized by the absence of tensile strength. Thus, for each
the measurement of in-situ compressive stresses [37,39]. However, rigid block, the linear and non linear kinematic analyses should be
as the realization of experiments is often a costly activity, during applied and, therefore the mechanism most likely to occur of all the
the last decades an increasing attention has been paid on the de- possible local mechanisms need to be identified. Particularly, Fig. 2
velopment of numerical tools and analytical models. The choice of shows that the aforementioned procedure is suitable to be applied to
the most appropriate approach depends on the desired level of accu- slender structures, such as masonry towers [28,54,55].
racy and simplicity [40]. The most widely used numerical approach Although successful to some extent, this methodology shows three
is the Finite Element Method (FEM). Different simulations based on main limitations: (i) the choice of the collapse mechanisms is affected
1D models [41,42], 3D models [43], or based on plates and shells by a certain level of uncertainty and it mostly relies on the experi-
theory [44] have been proposed to describe masonry at different levels ence of the analyst, (ii) both linear and non-linear kinematic analyses
of resolution. FEM allows to perform different typologies of analysis, have to be applied to each rigid block, thus making this procedure
either static [45–47] or dynamic [48], and provides the adoption of tedious to be performed, (iii) the identification of simplified typology
a variety of constitutive equations describing the material behavior, of collapse mechanism is inaccurate for complex geometries. In order
i.e. no tension [42] or elasto-plastic [49] approaches, or models with to overcome these drawbacks, it is necessary to directly simulate the
damage [50]. Although FEM has been widely used for regular masonry fracturing behavior of masonry towers. To this aim, several numer-
structures, it appears to be limited in simulating irregular masonries. In ical methods can be used, as they capture the mechanical behavior
fact, because of the heterogeneous nature of the material and the neces- of quasi-brittle materials with different degrees of accuracy [56,57].
sity of capturing complex crack distributions and fracture mechanisms, The so-called Lattice Discrete Particle Model (LDPM) is here adopted
the use of a dedicate modeling tool is necessary. to model the masonry fracturing behavior at the meso-scale. LDPM
For the assessment of seismic vulnerability, the Italian construction uniquely characterizes the localization of crack pattern that triggers the
code [51] prescribes to perform analysis of local and global collapse collapse mechanism. In this way, the pre-definition of multiple collapse
mechanisms. The identification of the collapse mechanisms is not a triv- mechanisms can be avoided and the kinematic analysis can be directly
ial task and it requires a preliminary thorough study of the unreinforced applied to the numerically calibrated fractured structural configuration.
masonry structure. The knowledge of the construction can be achieved Moreover, since LDPM captures the entire damage evolution phenom-
through the performance of a complete survey that allows (i) the ena, starting from cracks localization, propagation and up to the overall
understanding of the different historical phases affecting the building collapse, LDPM can be used as an alternative tool to the kinematic

2
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 3. (a) Location of the Medici tower in the landscape and within the core of Santo Stefano di Sessanio; (b) architectonic configuration of the tower before the 2009 L’Aquila
earthquake; (c) reproduction of the tower original profile through scaffolding and other prevention measures placed after the collapse in 2009.

analysis [58]. In this study, the LDPM is used complementary to the 3. Main features of the April 2009 seismic event
kinematic analysis to describe the fracturing behavior of the Medici
tower subjected to seismic excitation. The main characteristics of the On Monday, April the 6th, 2009, at 03:32:39 a.m. local time a
fracture, occurred during the 2009 L’Aquila earthquake, is identified devastating earthquake struck L’Aquila city and surrounding villages
from the numerical results and, subsequently, the individuated cracked in the Abruzzo region of central Italy. The magnitude of the event
configuration is used to perform the linear kinematic analysis. Finally, was estimated to be 𝑀𝐿 5.8 (Richter magnitude scale), and 𝑀𝑊 6.2
a comparison between the kinematic analysis and the lattice discrete (moment magnitude scale), according to the Italian National Institute
modeling is carried out, underlying advantages and drawbacks of the of Geophysics and Volcanology (INGV) [60]. The main shock was
proposed integrated approach. followed by a long-lasting seismic sequence, including more than 30
aftershocks with magnitude 3.5 < 𝑀𝐿 < 5.0. The earthquake has been
2. Historical evolution of the Medici tower interpreted as the result of a normal fault movement on a NW–SE
oriented structure, about 15-km-long, which is part of the 800-km-long
The Medici tower is located in Santo Stefano di Sessanio (L’Aquila, segmented normal fault system running all along the Apennines moun-
Italy), a urban aggregate belonging to the medieval period. The tower tain chain [61–63]. Specifically, the entire area affected by the seismic
is believed to be one of the most representative architecture of central activity covers an ellipse-shaped region parallel to the Apennines moun-
Italy, because of its spectacular location in the landscape, on top of a tain belt, with principal axis about 15-km-long and secondary axis
hill, and within the core of the medieval town, as shown in Fig. 3a. 2-5-km-long. Several studies showed that the 2009 L’Aquila earthquake
Before the collapse, the tower was the result of several modifications caused an up-dip slip movement with a rectangular rupture area of
occurred along the centuries. It is fundamental observing that the
approximately 17 × 14 m2 , at a depth ranging between 11.8 km and 0.6
knowledge of the historical phases affecting the building transforma-
km. The rupture plane is characterized by a strike of 147◦ , a dip of 43◦
tions is a fundamental prerequisite for any restoration intervention,
and a slip of 88◦ [64]. The main shaking was recorded by 55 stations
including the choice of the most appropriate modeling strategy (see
of the National Accelerometric Network, fourteen being in the Abruzzo
Section 4.2). Originally, during the 12th century, the cylindrical ma-
region. The nearest stations are AQG, AQV, AQK and AQA, located at
sonry tower was built without crowning on top. The hollow cylindrical
4.3 km, 4.8 km, 5.6 km and 5.8 km from the epicenter, respectively,
cross section had an internal diameter of 3.86 m and a thickness
on B- or C-type of soil (for the classification of soil types the reader
of 1.50 m. In addition to the tower entrance, two small windows
is referred to [65]). Table 1 reports the values of the Peak Ground
characterized the body of the tower: one was located at a height of
Acceleration (PGA) for the two horizontal and orthogonal directions
about 3.80 m from the bottom of the tower and the other consisted of
(PGA𝐸𝑊 and PGA𝑁𝑆 ) and for the vertical direction (PGA𝑍 ).
a narrow window placed near the top edge of the cylinder. During the
period of the Angioins’ domination, a crenelated masonry parapet was
added to the upper portion of the tower, as shown in Fig. 3a and b. 4. Numerical analysis
The battlement is characterized by 10 merlons and 30 machicolations
and its function was to serve as real defense presidium for the village. The mechanical behavior of the Medici tower subjected to the 2009
In this way, the total height of the tower reached about 20 m. During L’Aquila earthquake is numerically analyzed in this section. For this
the Second World War, the tower function changed, as it was used as purpose, the Lattice Discrete Particle Model (LDPM) is here adopted.
antiaircraft station and, for this scope, a concrete slab was built on
top of the tower. To exit the tower and access the crenelated parapet, 4.1. The lattice discrete particle model
a small window was opened at this floor. In April 2009, the tower
did not resist the strong L’Aquila earthquake and collapsed almost LDPM was proposed originally by Cusatis and coworkers to simulate
completely. At present, just a stump of the bottom part of the original the behavior of concrete at a meso-scale level. This peculiarity has
body remains, characterized by a variable height along its perimeter. been achieved by modeling the interaction between coarse aggregate
The stump presents its shortest side in correspondence of the bottom pieces [66,67]. Afterwards, LDPM has been successfully adopted to
window, at about 3.80 m height, and the tallest side in correspondence capture the behavior of several quasi-brittle materials [68–72], and
of the top window, at about 13 m height. After the 2009 L’Aquila to simulate complex multi-physical phenomena such as aging and
earthquake, shoring, scaffolding and other prevention measures were ASR [73–75]. Particularly, LDPM was recently used to simulate irreg-
placed around the ruins of the tower in order to prevent the falling ular masonry [58,76,77]. Key results including details about model
of additional masonry pieces and to reproduce the original profile of calibration and validation are presented in Appendix.
the tower, as shown in Fig. 3c. For a more detailed description of the LDPM allows the characterization of irregular masonry as two-
Medici tower, the reader is referred to Gregori and coworkers [59]. phase material, i.e. stone-aggregate and mortar (Fig. 4a). The potential

3
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Table 1
Station of the Accelerometric National Network located in proximity of the epicenter.
Station Record Latitude Longitude Altitude Soil Epic. dist. PGA𝐸𝑊 PGA𝑁𝑆 PGA𝑍
[Id] [Id] [◦ ] [◦ ] [m] [–] [km] [g] [g] [g]
AQG FA030 42.37 13.34 721 B 4.3 0.42 0.43 0.22
AQV GX066 42.38 13.34 692 C 4.8 0.67 0.56 0.51
AQK AM043 42.34 13.40 726 B 5.6 0.34 0.34 0.35
AQA CU104 42.38 13.34 693 C 5.8 0.39 0.45 0.38

Fig. 4. (a) Microstructure of irregular stone masonry; (b) aggregate distribution in an irregular masonry specimen; (c) two adjacent LDPM polyhedral cells enclosing the associated
stone-aggregate pieces; tetrahedron portion associated with a stone-aggregate and a triangular LDPM facet.

failure is assumed to occur at the aggregate–mortar interface or within Similarly, one can define the traction vector as 𝐭𝑘 = [𝑡𝑁 𝑡𝑀 𝑡𝐿 ]𝑇 ,
the mortar layer [40], which is consistent with typical experimental where 𝑡𝑁 is the normal component, 𝑡𝑀 and 𝑡𝐿 are the shear com-
observations on irregular masonry. ponents. For the sake of readability, the subscript 𝑘 that designates
The geometrical meso-structure of masonry is obtained through the the facet is dropped in the following equations. In order to describe
following steps: (i) stone-aggregate pieces are assumed to be spheri- the behavior of the material, one needs to introduce the constitutive
cal particles; they are randomly placed within the specimen volume equations imposed at the centroid of each facet. The elastic behavior is
through a trial and error procedure, from the largest to the smallest defined through linear relations between the normal and shear stresses,
size. These particles or stones follow a particle size distribution function and the corresponding strains as 𝑡𝑁 = 𝐸𝑁 𝑒𝑁 , 𝑡𝑀 = 𝐸𝑇 𝑒𝑀 and 𝑡𝐿 =
which is defined from a set of mix-design parameters: cement-mortar 𝐸𝑇 𝑒𝐿 , 𝐸𝑁 = 𝐸0 and 𝐸𝑇 = 𝛼𝐸0 , 𝐸0 ≈ 𝐸∕(1 − 2𝜈) and 𝛼 ≈ (1 − 4𝜈)∕(1 +
content 𝑐, water-to-mortar ratio 𝑤∕𝑐, maximum and minimum stone- 𝜈) are the effective normal modulus and the shear–normal coupling
aggregate size 𝑑𝑎 and 𝑑0 , respectively, and Fuller coefficient 𝑛𝑓 . Fig. 4b
parameter, respectively, and 𝐸 is the macroscopic Young’s modulus and
shows an example of particle placement inside the volume of a masonry
𝜈 is the macroscopic Poisson’s ratio of the masonry.
semicircular arch; (ii) zero-radius particles are randomly placed on
In order to describe the inelastic behavior, one needs to distinguish
the external surface of the sample for the application of the boundary
three sets of mechanisms.
conditions; (iii) a Delaunay tetrahedralization procedure connects the
centers of the spherical particles (or nodes), defining a lattice system; The first mechanism is the fracturing and cohesive behavior under
(iv) a three-dimensional domain tessellation is then performed, result- tension and tension/shear occurring for 𝑒𝑁 > 0. One can define the
1
ing in a system of polyhedral cells, each of which encloses a particle effective strain as 𝑒 = (𝑒2𝑁 + 𝛼(𝑒2𝑀 + 𝑒2𝐿 )) 2 , and the effective stress as
1
(Fig. 4c). The polyhedral cells form a network of triangular facets 𝑡 = (𝑡2𝑁 + (𝑡2𝑀 + 𝑡2𝐿 )∕𝛼) and write the relationship between stresses
2
that are assumed to be the potential material failure locations. The and strains through damage-type constitutive equations as 𝑡𝑁 = 𝑡𝑒𝑁 ∕𝑒,
stones and the surrounding mortar are thus represented by irregular 𝑡𝑀 = 𝛼𝑡𝑒𝑀 ∕𝑒 and 𝑡𝐿 = 𝛼𝑡𝑒𝐿 ∕𝑒 [79,80].
meso-cells that well idealize the real textures and shapes of typical The effective stress 𝑡 is defined incrementally as 𝑡̇ = 𝐸𝑁 𝑒̇ and
masonry stones, and are intrinsically designed to produce a statistically its magnitude is limited by a strain-dependent boundary 0 ⩽ 𝑡 ⩽
isotropic masonry material often observed in real structures. Three sets [ ]
𝜎𝑏𝑡 (𝑒, 𝜔) in which 𝜎𝑏𝑡 (𝑒, 𝜔) = 𝜎0 (𝜔) exp −𝐻0 (𝜔)⟨𝑒max − 𝑒0 (𝜔)⟩∕𝜎0 (𝜔) ,
of equations are written to complete the discrete model framework:
⟨𝑥⟩ = max(𝑥, 0), 𝜔 is a variable defining the degree of interaction√
definition of strains at each triangular facet, constitutive equations
between shear and normal loading defined as tan(𝜔) = (𝑒𝑁 )∕( 𝛼𝑒𝑇 ) =
which relate facet strain vector with facet stress vector, and particle √ 1
equilibrium equations. The constitutive equations describe a softening (𝑡𝑁 𝛼)∕(𝑡𝑇 ); 𝑒𝑇 is the total shear strain defined as 𝑒𝑇 = (𝑒2𝑀 + 𝑒2𝐿 ) 2 , and
1
behavior for pure tension and shear-tension and a plastic hardening 𝑡𝑇 is the total shear stress defined as 𝑡𝑇 = (𝑡2𝑀 + 𝑡2𝐿 ) 2 .
behavior for pure compression and shear-compression. The maximum effective strain is time dependent and is defined as
1
If 𝐱𝑖 and 𝐱𝑗 are the positions of nodes 𝑖 and 𝑗, adjacent to the facet 𝑒max (𝜏) = (𝑒2𝑁,max (𝜏) + 𝛼𝑒2𝑇 ,max (𝜏)) 2 , where 𝑒𝑁,max (𝜏) = max𝜏 ′ <𝜏 [𝑒𝑁 (𝜏 ′ )]
𝑘, the facet strains are defined as: and 𝑒𝑇 ,max (𝜏) = max𝜏 ′ <𝜏 [𝑒𝑇 (𝜏 ′ )]. The strength limit of the effective stress
[ 𝑇 ]𝑇
𝐧𝑘 [[𝐮𝑘 ]] 𝐦𝑇𝑘 [[𝐮𝑘 ]] 𝐥𝑇𝑘 [[𝐮𝑘 ]] that defines the transition between pure tension and pure shear is
𝐞𝑘 = [𝑒𝑁 𝑒𝑀 𝑒𝐿 ]𝑇 = (1) √
𝑙 𝑙 𝑙
− sin(𝜔) + sin2 (𝜔) + 4𝛼 cos2 (𝜔)∕𝑟2𝑠𝑡
where 𝑒𝑁 is the normal strain component, and 𝑒𝑀 , 𝑒𝐿 are the tangential 𝜎0 (𝜔) = 𝜎𝑡 (2)
2𝛼 cos2 (𝜔)∕𝑟2𝑠𝑡
strain components, [[𝐮𝑘 ]] = 𝐮𝑗 − 𝐮𝑖 is the displacement jump at the
centroid of the facet 𝑘, 𝑙 = ‖𝐱𝑗 − 𝐱𝑖 ‖2 is the distance between where 𝑟𝑠𝑡 = 𝜎𝑠 ∕𝜎𝑡 is the shear to tensile strength ratio, 𝜎𝑠 is the shear
the two nodes, 𝐧𝑘 = (𝐱𝑗 − 𝐱𝑖 )∕𝑙 and 𝐦𝑘 , 𝐥𝑘 are two unit vectors strength and 𝜎𝑡 is the tensile strength. The post-peak softening modulus
mutually orthogonal in the facet plane projected orthogonally to the is controlled by the effective softening modulus 𝐻0 (𝜔) = 𝐻𝑠 ∕𝛼 + (𝐻𝑡 −
line connecting the adjacent nodes Fig. 1g. It was demonstrated [78] 𝐻𝑠 ∕𝛼) (2𝜔∕𝜋)𝑛𝑡 , in which 𝐻𝑡 = 2𝐸0 ∕(𝑙𝑡 ∕𝑙 − 1), 𝐻𝑠 = 𝑟𝑠 𝐸0 and 𝑛𝑡 is
that this definition of strains is consistent with the classical definition the softening exponent; 𝑙𝑡 is the tensile characteristic length defined
of strains in continuum mechanics. as 𝑙𝑡 = 2𝐸0 𝐺𝑡 ∕𝜎𝑡2 , 𝐺𝑡 is the mesoscale fracture energy.

4
M. Mercuri et al. Engineering Structures 246 (2021) 113046

The second set of equations describes the mechanism behind pore Table 2
Kinematic analytical parameters for the computation of the structural capacity.
collapse and material compaction 𝑒𝑁 < 0 under high confining pres-
sures. The strain-hardening behavior in compression is simulated with Latitude Longitude 𝛼0 𝑒∗ 𝐹𝑐 𝑎∗0
[◦ ] [◦ ] [–] [–] [–] [m/s2 ]
the following strain-dependent boundary 𝑡̇ 𝑁 = 𝐸𝑁 𝑒̇ 𝑁 and −𝜎𝑏𝑐 (𝑒𝐷 , 𝑒𝑉 )
42.34 13.65 0.232 0.896 1.35 1.882
⩽ 𝑡𝑁 ⩽ 0, where 𝜎𝑏𝑐 = 𝜎𝑐0 + 𝐻𝑐 (−𝑒𝑉 − 𝑒𝑐0 ) if −𝑒𝑉 ⩽ 𝑒𝑐1 , otherwise 𝜎𝑏𝑐 =
𝜎𝑐1 exp[(−𝑒𝑉 −𝑒𝑐1 )𝐻𝑐 ∕𝜎𝑐1 ] and 𝐻𝑐 = 𝐻𝑐1 +(𝐻𝑐0 −𝐻𝑐1 )∕(1+𝜅𝑐2 (𝑟𝐷𝑉 −𝜅𝑐1 )),
𝜎𝑐1 = 𝜎𝑐0 +𝐻𝑐 (𝑒𝑐1 −𝑒𝑐0 ), 𝑒𝑐1 = 𝜅𝑐0 𝑒𝑐0 , 𝑒𝑐0 = 𝜎𝑐0 ∕𝐸0 , 𝑒𝑉 = (𝑉 −𝑉0 )∕𝑉0 is the
volumetric strain computed at the LDPM tetrahedral level, 𝑒𝐷 = 𝑒𝑁 −𝑒𝑉 , modeled individually as elastic tetrahedral finite elements. The slab
𝑟𝐷𝑉 = |𝑒𝐷 |∕(𝑒𝑉 0 − 𝑒𝑉 ), for 𝑒𝐷 ⩽ 0 and 𝑟𝐷𝑉 = |𝑒𝐷 |∕(𝑒𝑉 0 ), for 𝑒𝐷 > 0, built on top of the tower is also modeled using elastic finite elements.
𝑒𝑉 0 = 0.1𝑒𝑐0 , 𝜎𝑐0 is the meso-scale yielding compressive stress, 𝐻𝑐0 is the A penalty constraint algorithm is used to connect the finite elements
initial hardening modulus, and 𝜅𝑐0 , 𝜅𝑐1 , 𝜅𝑐2 are parameters governing (namely the 30 machicolations and the top slab) to the main LDPM
the triaxial behavior at very high confinement. core (i.e. the body tower). The nodes belonging to the bottom surface of
The third failure type considered in LDPM describes the frictional the tower are restrained by fixing all the rotations and the translations
behavior. In the presence of compressive stresses, the shear strength
perpendicular to the direction of the seismic action. For all the simu-
increases due to frictional effects. The frictional behavior is computed
lations, three different LDPM meshes corresponding to three random
using a nonlinear Mohr–Coulomb model in which the internal friction
stone distributions within the volume of the body tower were used.
coefficient varies from an initial value 𝜇0 to zero with the following
The presence of the self-weight was considered by preliminary applying
formulation:
gravity loads. In order to dissipate kinetic energy and converge to a
𝜎𝑏𝑠 (𝑡𝑁 ) = 𝜎𝑠 + 𝜇0 𝜎𝑁0 − 𝜇0 𝜎𝑁0 exp(𝑡𝑁 ∕𝜎𝑁0 ) (3) steady state equilibrated solution, the nodal velocities of the particles
were scaled by a factor slightly lower than one (dynamic relaxation).
where 𝜎𝑠 is the cohesion and 𝜎𝑁0 is the so-called transitional stress.
When a discrete crack forms, gravity loads in large structures such as
Finally, the governing equations are completed by writing the equi-
the studied tower might induce multi-axial stresses in structures made
librium equations of each LDPM cell:
∑ ∑ of quasi-brittle materials, in this case masonry towers. Experiments
𝐴𝑘 𝐭𝑘 + 𝑉𝐼 𝐛 = 𝟎, 𝐴𝑘 𝐜𝑘 × 𝐭𝑘 = 𝟎 (4) and numerical simulations showed recently that these states of stresses
𝑘∈𝐼 𝑘∈𝐼 affect the size of the fracture process zone ahead of the crack tip [81,82]
where 𝐼 is the set containing all the facets of a generic polyhedral cell and therefore the resulting fracturing behavior of the structure. LDPM
𝐼, 𝐴𝑘 is the area of the facet 𝑘, 𝐜𝑘 is the vector representing the distance is able to capture this phenomenon as its constitutive equations for
between the center of the facet 𝑘 and the center of the cell, 𝑉𝐼 is the friction and tensile softening behavior are defined in a vectorial form
cell volume and 𝐛 is the external body force applied to the cell. at each facet of different orientations.

4.2. Modeling strategy

4.2.2. Inputs for the seismic action
4.2.1. The meso-scale characterization The load is imposed as a velocity applied to a node list contained
The geometrical characterization of the LDPM mesostructure was in a 500 mm height portion located at the bottom of the cylinder. For
calibrated in Angiolilli and coworkers [76]. More specifically, the the benchmark case, the direction and the magnitude of the velocity
stone-to-mortar ratio 𝑎∕𝑐 = 3.4 corresponding to the ratio between the are consistent with the ones of the 2009 L’Aquila earthquake and
volume of stones and the volume of cement-mortar, and the water-to- they are obtained following the procedure described in Decanini and
mortar ratio 𝑤∕𝑐 = 0.5 were assumed based on the actual masonry com- coworkers [83]. More specifically, the AQV accelerometric station,
position. The cement-mortar content parameter 𝑐 = 427.5 kg/m3 was provides three histories of accelerations for the main shock of the
calculated such that the total mass density 𝜌 was equal to 1800.0 kg/m3 L’Aquila seismic sequence, that are related to the North–South direction
using the formula: 𝜌 = 𝑐 (1+𝑤∕𝑐+𝑎∕𝑐). The stones were assumed to have (a𝑁𝑆 (t)), the East–West direction (a𝐸𝑊 (t)), and the vertical direction
characteristic size within the range 𝑑0 = 33 mm and 𝑑𝑎 = 200 mm in (a𝑍 (t)), respectively [84]. In general, knowing the evolution in time of
order to simulate the coarse gravel. The Fuller coefficient 𝑛𝑓 = 0.5 was the ground acceleration a𝑖 (t) for the generic direction 𝑖, it is possible
assumed as default parameter since no specific sieve curve was assumed to obtain the history of velocity related to the same direction v𝑖 (t) by
for the preparation of the specimens in the experimental campaign. 𝑡
integrating in time the history of acceleration: 𝑣𝑖 (𝑡) = ∫𝑡 2 𝑎𝑖 (𝑡)𝑑𝑡. In
1
The identified model parameters are as follows: normal elastic modulus this study, by neglecting the vertical component of the seismic action,
𝐸0 = 1200 MPa, shear–normal coupling parameter 𝛼 = 0.065, tensile and by considering a significant observation time of 10 s, one gets
strength 𝜎𝑡 = 0.3 MPa, tensile fracture energy 𝐺𝑡 = 11 N/m, shear-to- the histories of velocities in the North–South direction v𝑁𝑆 (t) and in
tensile strength ratio 𝜎𝑠 /𝜎𝑡 = 1.25, softening exponent 𝑛𝑡 = 0.2, yielding
the Est-West direction v𝐸𝑊 (t), as shown in Fig. 6a and b. Hence, the
compressive stress 𝜎𝑐0 = 125 MPa, initial hardening modulus 𝐻𝑐0 /𝐸0
histories of velocities are projected along the parallel direction (Parallel
= 0.4, final hardening modulus 𝐻𝑐1 = 1, transitional strain ratio 𝑘𝑐0 =
Strike) and the orthogonal direction (Normal Strike) to the fault plane
1.75, deviatoric strain threshold ratio 𝑘𝑐1 = 1, deviatoric damage pa-
(see Fig. 6c and d). Finally, the magnitude of the velocity to be applied
rameter 𝑘𝑐2 = 5, initial friction coefficient 𝜇0 = 0.2, asymptotic friction
to the bottom of the cylinder was set to 590 mm/s: this value is the
coefficient 𝜇∞ = 0, transitional stress 𝜎𝑁0 = 42 MPa, densification ratio
sum of the two consecutive maximum Peaks Ground Velocities (PGV)
𝐸𝑑 /𝐸0 = 1 and 𝑟𝑠 = 0. The details of the calibration and validation of
the LDPM is given in Appendix. taken in absolute value, i.e. 330 mm/s and 260 mm/s, as shown in
In order to accurately simulate the seismic behavior of the tower, Fig. 6d. Once the benchmark direction and magnitude of the velocity
the choice of the most suitable modeling strategy accounted for the were set to 147◦ of Strike (see Section 3) and 590 mm/s, the other
several structural and functional modifications that affected the tower five cases were investigated, by varying either the magnitude or the
along the centuries. The sub-portions of the overall geometry are direction of the velocity. Particularly, fixing the velocity direction to
individuated to be representative of the real case of study (see Fig. 5a 147◦ with respect to the North–South axis, two magnitudes of velocity
and b): the original hollow cylindrical body of the tower and the bat- were considered, i.e. -330 mm/s and 260 mm/s, being the negative
tlement composed by 10 merlons are modeled together using a single and positive maximum PGV, respectively. Moreover, fixing the velocity
LDPM mesh (see Fig. 5b and c) in which the openings are included magnitude to 590 mm/s, three directions of velocity were considered,
to reproduce the presence of the windows. The 30 machicolations are i.e. 15◦ , 45◦ and 90◦ with respect to the North–South axis, respectively.

5
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 5. (a) Picture showing the Medici tower before the 2009 L’Aquila earthquake (USRC: Uff. Spec. Ric. Com. Cratere); LDPM mesh of the Medici tower showing: (b) the cell
outline on the external cylindrical surface, (c) the distribution of spherical aggregates within the volume.

Fig. 6. Histories of velocity along (a) the North–South direction, (b) the East–West direction, (c) the parallel direction of the fault plane (Strike Parallel), (d) the normal direction
of the fault plane (Strike Normal).

Table 3 actual failure shown in Fig. 7c. First, damage occurs at the top part
Seismic parameters for the computation of the structural demand.
where the concrete slab is modeled. Next, the main contour of the
𝑎𝑔,𝑆𝐿𝑉 𝐹0 𝑇𝑐∗ 𝑇1 𝑆𝑆 𝑆𝑇 𝑆 𝑞 𝑆𝑒,𝑆𝐿𝑉 (𝑇1 ) 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 𝑓𝑎,𝑆𝐿𝑉
mesoscale crack starts from the bottom opening, and propagates diag-
[g] [–] [s] [s] [–] [–] [–] [–] [g] [m/s2 ] [–]
onally reaching the top narrow window. This consideration suggests
0.256 2.365 0.344 0.404 1.16 1.40 1.62 1.75 4.07 2.33 0.809
a correspondence between the type of failure and the presence of
doors and windows, that interrupts the structural continuity of the
masonry cylinder. For simple and complex geometries, Mercuri and
4.3. Analysis of results and discussion coworkers [58] recently pointed out that the presence of openings
greatly affects the seismic out-of-plane behavior of one- or two-story
The results of the numerical simulations preformed on the bench- masonry walls, placed individually or collocated within the continuity
mark case of Fig. 7a and b show that LDPM has the predictive capability of the facade. They also demonstrated that the reductions of stiffness
of simulating the behavior of the tower under seismic excitation. In- and bearing capacity are greater if the amount of opening within the
deed, the simulated failures (Fig. 7a and b) seem to agree with the wall increases. These observations appear to be realistic also for the

Table 4
Geometrical parameters for the linear kinematic analysis.
Case [Id] Description fracture 𝑦𝐵𝑊 [m] 𝑦𝑇 𝑊 [m]
b-B From the bottom of the lower opening to the bottom of the higher window 3.80 13.00
m-B From the middle of the lower opening to the bottom of the higher window 4.90 13.00
h-B From the top of the lower opening to the bottom of the higher window 6.00 13.00
b-H From the bottom of the lower opening to the top of the higher window 3.80 13.50
m-H From the middle of the lower opening to the top of the higher window 4.90 13.50
h-H From the top of the lower opening to the top of the higher window 6.00 13.50

6
M. Mercuri et al. Engineering Structures 246 (2021) 113046

description of the seismic behavior of complex geometries, as for the To evaluate the structural capacity, the activating acceleration 𝑎∗0
tower analyzed here. Thus, it is of primary relevance to account for has to be computed. As a first step, the Virtual Work Theorem is applied
the openings while considering the most suitable modeling strategy. It by writing a work balance equation:
is also of paramount importance to observe diffused cracks at the top
of the tower, with fractures extending all around the merlons of the 𝑊𝐸,𝑆 + 𝑊𝐸,𝐺 = 𝑊𝐼 (7)
battlement (see Fig. 7a and b). This phenomenon is probably caused where 𝑊𝐸,𝑆 is the external work done by seismic forces, 𝑊𝐸,𝐺 is the
by the motion of the rigid concrete slab built on top of the tower external work done by gravity forces, and 𝑊𝐼 is the work done by
during the Second World War, when the tower was used as antiaircraft internal forces. In the case of rigid bodies, the internal work is set equal
station. The slab, modeled using an elastic finite element mesh, turned to zero, as no elastic/post-elastic deformation takes place. Developing
out to fall down almost untouched as consequence of the April 2009 the latter expression, the horizontal multiplier 𝛼0 that activates the
earthquake. A few preliminary simulations of the tower without the analyzed failure mode can be identified:
presence of the concrete slab were performed to analyze its effect on the ( 𝑛 )
failure of the tower. There was no damage predicted at the top of the ∑ ∑
𝑛+𝑚 ∑
𝑛
𝛼0 𝑚𝑏,𝑖 𝛿𝑥,𝑖 + 𝑚𝑓 ,𝑖 𝛿𝑥,𝑖 + 𝑔 𝑚𝑏,𝑖 𝛿𝑦,𝑖 = 0 (8)
tower, which might confirm that in the actual failure, the collapse of the 𝑖=1 𝑖=𝑛+1 𝑖=1
concrete slab occurred independently from the tower. In addition, the
where 𝑛 is the number of blocks of the kinematic chain, 𝑚 is the number
main fracture patterns were very similar to the ones seen in Fig. 7. The
of internal floors, 𝑚𝑏,𝑖 is the mass of the generic block, 𝑚𝑓 ,𝑖 is the mass of
weak points related to the position of the discontinuities seem to mainly
the generic floor, 𝛿𝑥,𝑖 is the virtual horizontal displacement of the point
govern the location and fracture patterns of the tower. These results
of application of the 𝑖th seismic force and 𝛿𝑦,𝑖 is the virtual vertical
emphasize that the most appropriate modeling choices depend upon
displacement of the 𝑖th weight-force.
a deep knowledge of the building historical phases, achieved through
Finally, the activating acceleration 𝑎∗0 can be computed through the
profound investigations.
following equation:
Moreover, it is worth mentioning that the simulated fracturing
∑
behavior of the tower mostly depends on the direction of the seismic 𝛼0 𝑛+𝑚
𝑖=1 𝑚𝑖 𝛼
𝑎∗0 = = ∗0 (9)
action rather than the magnitude of the velocity, as shown in Fig. 7c, 𝑀 ∗ 𝐹𝐶 𝑒 𝐹𝐶
d and e. The fractured configurations related to the two cases with
where 𝑚𝑖 is the mass of the generic block or floor, 𝐹𝐶 is the confidence
Strike fixed at 147◦ and PGV of 330 m/s and 260 m/s qualitatively
factor accounting for the knowledge of the building. A value of 𝐹𝐶 =
coincide with the benchmark case (characterized by a PGV of 590 mm/s
1.35 was assumed in this study since the compressive strength is infi-
and shown in Fig. 7a and b) and, for the sake of brevity, they are ∑
nite. The expression 𝑒∗ = 𝑀 ∗ ∕ 𝑛𝑖=1 𝑚𝑖 represents the participant mass
not reported in this paper. On the other hand, Fig. 7 also reports
ratio of the considered mechanism and 𝑀 ∗ stand for the participant
the three different failure mechanisms of the tower, by setting the
mass obtained as a function of the virtual displacements:
PGV to 590 mm/s and varying the direction of the seismic action.
(∑𝑛+𝑚 )2
In particular, Fig. 7c, d and e show that the fractures become more 𝑚𝑖 𝛿𝑥,𝑖
diffused and almost vertical as the seismic direction rotates with respect 𝑀 ∗ = ∑𝑖=1 𝑛+𝑚
(10)
2
𝑖=1 𝑚𝑖 𝛿𝑥,𝑖
to the North–South direction, of 90◦ , 45◦ and 15◦ respectively. The
differences in the configuration of the fracturing patterns can be due Table 2 shows the values of the kinematic parameters adopted for
to the variability in the relative position of the openings with respect the computation of the structural capacity.
to the direction of the seismic action. Overall, LDPM shows a wide Aiming to quantify the structural demand 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 , the fundamental
capability to capture the fracturing behavior up to the overall collapse period of the structure 𝑇1 is preliminary computed. With this purpose,
of the structure for simple and complex geometries subjected to a several empirical formulas have been proposed trying to estimate the
variety of loading conditions. fundamental period of masonry towers as a function of their height
𝐻 [65,85,86]. In particular, Ranieri and Fabbroncino [86] proposed
5. The linear kinematic analysis a correlation between 𝑇1 and 𝐻 and compared the formula against a
database of 30 Italian historical masonry towers, demonstrating that
In this section, the linear kinematic analysis is applied to the struc- their expression provides the most accurate prediction for this specific
tural configuration identified as benchmark case in Sections 4.2 and structural typology. Hence, their formula will be adopted in this study:
4.3, that reproduced the fracture pattern occurred during the 2009
L’Aquila earthquake. As already mentioned, the main contour is di- 𝑇1 = 0.013𝐻 1.138 (11)
agonal and it extends from the bottom opening to the top narrow
window. For the kinematic analysis performed in this section, the crack If the activated portion lies directly on the foundation, 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 is
is assumed to trigger from the bottom of the lower opening, at about evaluated as follows:
3.80 m from the ground, reaching the bottom of the top window, at 𝑎𝑔,𝑆𝐿𝑉 𝑆
about 13.00 m from the ground. According to [65], within the context 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 = (12)
𝑞
of the linear kinematic analysis, the Life Safety Limit State (SLV) check
where 𝑎𝑔,𝑆𝐿𝑉 is the reference PGA considered for a return period of 475
for the generic mechanism is carried out by comparing the activating
years, corresponding to the SLV Limit State, 𝑆 is a correction factor
acceleration 𝑎∗0 , that represents the capacity of the structure, with the
accounting for the soil type and topography conditions, and 𝑞 is the
maximum expected spectral acceleration for the structure in ultimate
reduction factor that describes the ability of the system to dissipate
conditions 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 , that represents the demand:
energy in the non-elastic phase. The values of the aforementioned
𝑎∗0 ≥ 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 (5) parameters are listed in Table 3 and they are computed according
to [65,87]. In particular, the expression adopted for the behavior factor
An alternative expression for the check of SLV Limit State consists is 𝑞 = 𝑞0 𝐾𝑅 [65], where 𝑞0 = 1.75𝛼𝑢 /𝛼1 for ordinary masonry, and
in the following inequality, stating that the acceleration factor 𝑓𝑎,𝑆𝐿𝑉 𝐾𝑅 = 1 being the construction regular in height.
is greater than one: If the activated portion is located some distance above the founda-
𝑎∗0 tion, an other expression has to be considered in addition to Eq. (12),
𝑓𝑎,𝑆𝐿𝑉 = ≥1 (6) accounting for the potential amplifications in acceleration due to the
𝑎𝑒𝑥𝑝,𝑆𝐿𝑉

7
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 7. Cracked configuration in the Medici tower, simulating the failure occurred during the 2009 L’Aquila earthquake (a) in the South-East facade; (b) in the North-West facade;
(c) Tower stump after the 2009 L’Aquila earthquake; meso-scale crack openings assuming a direction of the seismic action rotated with respect to the North–South direction of:
(d) 90◦ ; (e) 45◦ ; (f) 15◦ .

dynamic response of the structure. In this case, 𝑎𝑔,𝑆𝐿𝑉 is the maximum where 𝑆𝑒,𝑆𝐿𝑉 (𝑇1 ) is the amplitude of the elastic spectrum corre-
of the two following expressions: sponding to SLV Limit State, evaluated for the fundamental period of
the structure 𝑇1 , 𝜓(𝑧) = 𝑧∕𝐻 is the first vibration mode, approximated
⎧ 𝑎𝑔,𝑆𝐿𝑉 𝑆 with a linear distribution along the height and normalized to one in
⎪ 𝑞 correspondence of the building top, with 𝑧 being the height of the
𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 = max ⎨ (13)
𝑆 (𝑇 )𝜓(𝑧)𝛾 kinematic hinge related to the considered mechanism with respect to
⎪ 𝑒,𝑆𝐿𝑉 1
⎩ 𝑞
the foundation and 𝐻 the total height of the structure. In Eq. (13),
the expression 𝛾 = 3𝑁∕(2𝑁 + 1) represents the modal participation

8
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 8. Geometrical configurations for linear kinematic analysis preformed assuming different diagonal fractures in function of the openings position 𝑦𝑊 ; the reader is referred to
Table 4 for the detailed description of the cracked pattern.

Fig. 9. (a) Horizontal multiplier 𝛼0 as a function of the openings position 𝑦𝑊 ; (b) activating acceleration 𝑎∗0 as a function of the openings position 𝑦𝑊 .

Table 5 level of the floor connected by beam elements. Thus, the formula
Analytical parameters for the SLV limit state check.
well approximates structural configurations whose floor masses are
Case 𝛼0 𝑒∗ 𝑎∗0 𝑓𝑎,𝑆𝐿𝑉 SLV check
[Id] [–] [–] [–] [m/s2 ] [–]
significantly greater than the masses of the vertical structures. For
b-B 0.232 0.896 1.882 0.809 x
the Medici tower the vertical masonry structure is more massive than
m-B 0.250 0.888 2.049 0.881 x the floors and, therefore, a coefficient 𝑁 = 1 is assumed. Thus, the
h-B 0.272 0.880 2.243 0.963 x modal participation coefficient turns out to be 𝛾 = 1, which means
b-H 0.228 0.892 1.857 0.798 x
m-H 0.246 0.884 2.020 0.868 x assuming that the first mode is the only one that significantly affects the
h-H 0.267 0.877 2.209 0.970 x acceleration 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 . This assumption seems to be coherent with the
x = not satisfied. simulated type of mechanism (shown in Fig. 7a and b) and the actual
failure (see Fig. 7c). The values of the aforementioned parameters are
listed in Table 3 and they are computed according to [65,87]. Table 3
coefficient, where 𝑁 is the number of the levels of the structure. It reports also the value of the acceleration factor (𝑓𝑎,𝑆𝐿𝑉 = 0.809) for
is worth noting that this simplified formula is applicable to structures the considered mechanism. Being 𝑓𝑎,𝑆𝐿𝑉 less than 1, the SLV check is
that can be plausibly modeled with lumped masses located at the not verified and the collapse mechanism is activated.

9
M. Mercuri et al. Engineering Structures 246 (2021) 113046

6. Bridging LDPM with kinematic analysis fracturing and the collapse behavior of masonry structures with elabo-
rate more numerical models, such as LDPM, in addition or in alternative
The results of the numerical simulations performed on the bench- to simplified kinematic analysis. These advanced numerical models
mark case of Fig. 7a and b showed that the main crack initiates showed that the failure criteria and, more generally, the structural
from the bottom opening, and propagates diagonally reaching the top behavior depend on: (i) the presence of openings within walls, (ii)
narrow window. Nevertheless, because of the heterogeneous nature of the correct application of boundary conditions and (iii) the proper
the masonry, the mutual position between the fracture contour and consideration of the seismic direction. As a matter of fact, the correct
the openings is not clearly defined in the numerical results. On the choice of the boundary conditions is not trivial and it implies the
other hand, the kinematic analysis performed in Section 5 assumed the convergence of different disciplines related to structural engineering
diagonal crack to start from the bottom of the lower opening, at about and history/architecture, aiming to account for the building evolution.
3.80 m from the ground, reaching the bottom of the top window, at Also, the proper consideration of the seismic forces is a challenging
task and should be the result of an organic approach in which the
about 13.00 m from the ground (Case b-B in Fig. 8a). However, the
soil–structure interaction links the structural and geo-mechanical fields.
randomness affecting the fracturing pattern should be considered while
Consequently, the adoption of a specific modeling strategy, framed
bridging the LDPM results with the kinematic analysis. With this aim,
within the structural design calculation, should represent a step of
five additional cases (shown in Fig. 8 and Table 4) are analytically
a more general and interdisciplinary process, in which each level of
investigated by varying the reciprocal position between the diagonal
knowledge is intrinsically consistent and organically organized.
fracture and the location of the openings. Fig. 8 illustrates the cracked
configurations and Table 4 reports the geometrical parameters assumed 7. Summary and conclusions
for the linear kinematic analysis. The first set of three cases (cases b-B,
m-B and h-B, corresponding to Fig. 8a, b and c, respectively) provides Due to their slender geometry and the brittle heterogeneous na-
the diagonal crack to reach the bottom of the top window, at about ture of their constituent material, ancient masonry towers are highly
13.00 m from the ground, and to trigger from the bottom, the middle vulnerable to seismic actions. The mechanical response, the induced
and the top of the lower opening, at 3.80 m, 4.90 m and 6.0 m from damage, and the resulting collapse mechanism due to the action of
the ground, respectively. The second set of three cases (cases b-H, m-H the earthquake depend on the geometrical features and properties of
and h-H, corresponding to Fig. 8d, e and f, respectively) assumes the masonry. In order to calculate the ultimate conditions of these towers,
diagonal crack to reach the top of the highest window, at about 13.50 kinematic analyses are most often performed. Building codes usually
m from the ground, and to trigger from the bottom, the middle and top require the identification of collapse mechanisms, which in turns re-
of the lower opening, at 3.80 m, 4.90 m and 6.0 m from the ground, quire a preliminary thorough survey of the analyzed structure. In this
respectively. study, a comprehensive numerical model, namely the Lattice Discrete
The results of the kinematic analysis are reported in Table 5 and Particle Model, is adopted to model the masonry fracturing behavior
Fig. 9. For all the cases, the SLV check is not verified and the collapse and predict the most likely to occur collapse mechanism. The infor-
mechanisms are activated. In fact, the activating acceleration 𝑎∗0 is al- mation about crack distribution and location are then used to define
ways lower than the maximum expected spectral acceleration 𝑎𝑒𝑥𝑝,𝑆𝐿𝑉 , the fractured structural configuration in the kinematic analysis. This
as shown in Fig. 9b. Although the checks are not verified for all the paper incorporates this method to investigate the fracturing behavior
cases, each one of them shows a different value of acceleration factor of the Medici tower subjected to the seismic excitation due to the 2009
𝑓𝑎,𝑆𝐿𝑉 , ranging from 0.798 to 0.970. In addition, a trend can be found: L’Aquila earthquake. Crack contours are simulated and six different
the acceleration factor 𝑓𝑎,𝑆𝐿𝑉 increases as the position of the crack failure locations are assumed for the kinematic analysis. The obtained
in the vicinity of the bottom opening is assumed to be located at the results suggest the following conclusions:
bottom, the middle, or at the top (see Fig. 9c). It is worth noting that • The failure of the Medici tower under seismic excitation is numer-
the assumption made for the benchmark case, i.e. crack starting from ically well replicated by LDPM.
the bottom of the bottom window to the bottom of the top window, is • The failure pattern greatly depend on the direction of the seismic
in favor of safety as the acceleration factor is one of the lowest among action.
the six cases (𝑓𝑎,𝑆𝐿𝑉 = 0.809). The damage surveys following relevant • The kinematic analysis performed with six different configura-
seismic events brought to light recurring types of crack depending on tions underline the importance of identifying the correct collapse
the presence and the location of the openings within the masonry walls. mechanism to obtain more reliable SLV checks.
For walls included within the continuity of the facade, one of the
most recurrent observations of the damage consists in inclined cracks The proposed method was successful in analyzing the ultimate condi-
that propagate from the top of the bottom opening to the bottom of tion of the Medici tower. It appears however that in other cases, for
the top window and they are known in the field as the Saint Andrea instance if different directions of the seismic action are assumed, the
Cross [88,89]. For the tower analyzed in this study, the results shown in use of simplified kinematic analyses is only a crude approximation.
In fact, these situations call for advanced numerical models capable
Fig. 9c indicate that such an assumption does not lead to a calculation
of accurately capturing fracture propagation and crack distribution in
of the ultimate condition on the safest side. In fact, for the analyzed
order to assess the vulnerability of existing masonry structures.
geometry (Case h-B shown in Fig. 8c), the corresponding value of the
acceleration factor is very close to the threshold value 1.0 (𝑓𝑎,𝑆𝐿𝑉 =
CRediT authorship contribution statement
0.963), which questions the reliability of the SLV check.
The results demonstrate that the method proposed in this paper, Micaela Mercuri: Formal analysis, Investigation, Writing - origi-
i.e. identifying the main failure pattern from the lattice discrete model- nal draft. Madura Pathirage: Software, Writing - review & editing.
ing, and then performing and refining the simplified kinematic analysis, Amedeo Gregori: Conceptualization, Resources. Gianluca Cusatis:
allows to obtain realistic ultimate conditions for a given geometry. Conceptualization, Supervision, Writing - review & editing.
However, in other situations, it appears difficult or impossible to pro-
ceed this way, as the damage cannot be described by a single crack. Declaration of competing interest
For instance, Fig. 7c, d and e showed that the fracture becomes less
localized and more distributed as the seismic direction changes. Similar The authors declare that they have no known competing finan-
conclusions have been reached in the work of Mercuri and cowork- cial interests or personal relationships that could have appeared to
ers [58]: in the most general case, it is necessary to analyze the influence the work reported in this paper.

10
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 10. (a) LDPM cells for unreinforced stone masonry wall; (b) Contours of meso-scale crack opening at failure for unreinforced masonry specimen at 𝛾 = 0.005; (c) Comparison
between experiments and simulations computed for stone masonry specimens tested under diagonal compression: the R and NR results refer to the reinforced and unreinforced
cases, respectively.

Acknowledgments of experiments computed for both the reinforced R and unreinforced

NR cases, whereas the two marked curves represent the average of
This research was supported in part through the computational the numerical responses of both the R and NR cases performed for
resources provided for the Quest high performance computing facility three different random particle configurations. For more details on the
at Northwestern University which is jointly supported by the Office performed numerical simulations, and in particular on the method used
of the Provost, the Office for Research, and Northwestern University to account for the reinforcement, the reader is referred to [76].
Information Technology.
Out-of-plane behavior simulations
Funding
The LDPM was also recently used to understand the out-of-plane be-
This work of the first author was supported by the National Op-
havior of masonry walls. In this study, Mercuri and coworkers [58] first
erational Programme on Research and Innovation 2014–2020 (CCI
validated the model against a set of experimental data in which three
2014IT16M2OP005) and the European Social Fund.
masonry panels of different dimensions were loaded in out-of-plane
direction. Further analyses were conducted, for instance simulations
Appendix
of overturning walls with and without openings and more complex
mechanisms to reproduce the behavior of systems of panels included
within the continuity of facades. Here, the validation with experimental
Diagonal tests simulations
data is summarized. In the experiments [96], the three panels were
built on top of a masonry base with a larger thickness. A steel cable was
The first study in which the Lattice Discrete Particle Model was
fixed at 2/3 of the panel height at the center of the transverse section on
used to simulate irregular masonry was conducted by Angiolilli and
one end and attached to a restraint frame on the other end. The cable
coworkers [76]. In this study, diagonal compression tests were carried
was equipped with a load cell to measure the applied force 𝐹 along
out on six masonry samples. Two of them were plain masonry whereas
the others were reinforced with glass-fiber reinforced cementitious the cable. The displacement 𝑑 was measured at the location where the
matrix made of lime-based mortar. The masonry panels were subjected cable was fixed. For the testing setup, the test was carried out in three
to a compressive force applied through L-shaped steel plates along one phases. In Phase 1, the cable was pulled horizontally in order to form
of the panels’ diagonals. The tests were performed under force-control a cracked surface at the bottom of the specimen. In Phase 2, the cable
conditions and the load 𝑃 was measured. The applied shear stress 𝜏 was released to obtain a standing panel as in the initial configuration.
defined as 𝜏 = 0.707 𝑃 ∕𝐴𝑛 (ASTM E519-02 [90], where 𝐴𝑛 is the net Finally, in Phase 3, the cable was pulled horizontally and statically
area of the panel) was calculated. In addition, the shear strain 𝛾 was in displacement controlled conditions. The LDPM parameters were
monitored (following the standard [90], 𝛾 = 𝜀ℎ + |𝜀𝑣 |, where 𝜀ℎ and assumed based on the work described earlier on diagonal compression
𝜀𝑣 are the positive tensile and negative compressive strain measured performed on a similar stone masonry mix composition. The base of
along the two diagonals). The LDPM parameters were best fitted on the the specimen and the panel were modeled using one LDPM mesh.
𝜏-𝛾 experimental curves recorded during the diagonal compression tests Two sets of simulations were performed. The first strategy was to
and on the compressive strength of unreinforced samples in the range reproduce the same testing procedure as in the experiments, i.e. in
from 0.8 MPa to 1.2 MPa chosen on the basis of Standards [65,87,91] three phases. The second strategy, less computationally expensive, was
and experimental tests [92–95]. to apply the static load directly to the uncracked specimen, from the
Fig. 10 illustrates some numerical results obtained and the compar- beginning of the test, up to cracking and overturning of the panel. It
ison with experimental data. In particular, Fig. 10a shows the model was shown that replicating the experimental procedure and loading
used for the simulations. Two steel plates at the corners of the panels directly the sample from the initial uncracked configuration provides
and a horizontal steel plate placed at the specimen basement were mod- identical results. Figs. 11 and 12 show the load–displacement curves
eled as rigid triangular shell elements. A sliding with friction constraint and the contours of meso-scale crack opening, respectively, for the
algorithm was used for the contact between the masonry panel and the three panels. One can see that the numerical results well predict the
plates. In Fig. 10b one can observe the crack opening at failure for the experimental data. Fig. 11 also includes the results using non-linear
unreinforced specimen, corresponding to a value of the macroscopic kinematic analysis, and more specifically the rigid-block displacement-
shear strain 𝛾 equal to 0.005. The main cracks occurred along the based method (RBM), which also agrees with the numerical results. For
compressed diagonal as observed in the experiments. Fig. 10c points more details on the capabilities of the LDPM to capture the out-of-plane
out some quantitative results: the two areas represent the envelope mechanical behavior of masonry, the reader is referred to [58].

11
M. Mercuri et al. Engineering Structures 246 (2021) 113046

Fig. 11. Force versus displacement curves in: (a) panel 1; (b) panel 2; (c) panel 3.

Fig. 12. Contours of meso-scale crack opening in: (a) panel 1; (b) panel 2; (c) panel 3.: (a) panel 1; (b) panel 2; (c) panel 3.

References [14] Coccia S, Di Carlo F, Imperatore S. Force reduction factor for out-of-plane simple
mechanisms of masonry structures. Bull Earthq Eng 2017;15(3):1241–59.
[1] Mendes P, Baptista M, Agostinho L, Lagomarsino S, Costav J. Structural and [15] Vailati M, Monti G, Khazna M, Realfonzo R, De Iuliis M. Probabilistic seismic
dynamic analysis of N. Sra. do Carmo church, Lagos Portugal, In: Proceedings response analysis of existing masonry structures. 2016.
EURODYN2005, Structural Dynamics, 2005: p. 311–318. [16] Casolo S, Uva G. Nonlinear analysis of out-of-plane masonry façades: full
[2] Osmancikli G, Uçak Ş, Turan FN, Türker T, Bayraktar A. Investigation of dynamic versus pushover methods by rigid body and spring model. Earthquake
restoration effects on the dynamic characteristics of the hagia sophia bell-tower Eng Struct Dyn 2013;42(4):499–521.
by ambient vibration test. Constr Build Mater 2012;29:564–72. [17] Felice GD, Giannini R. Out-of-plane seismic resistance of masonry walls. J
[3] DelloRusso S, Juneja G, Gabby B, Dusenberry D. Monitoring and repair Earthquake Eng 2001;5(02):253–71.
of the milwaukee city hall masonry tower. J Perform Constr Facilities [18] de Felice G. Out-of-plane seismic capacity of masonry depending on wall section
2008;22(4):197–206. morphology. Int J Arch Heritage 2011;5(4–5):466–82.
[4] Pieraccini M, Dei D, Betti M, Bartoli G, Tucci G, Guardini N. Dynamic iden- [19] Giuffrè A. A mechanical model for statics and dynamics of historical masonry
tification of historic masonry towers through an expeditious and no-contact buildings. In: Protection of the Architectural Heritage Against Earthquakes.
approach: Application to the ‘‘Torre del Mangia’’ in siena (Italy). J Cult Heritage Springer; 1996, p. 71–152.
2014;15(3):275–82.
[20] Lourenço PB, Rots JG. Multisurface interface model for analysis of masonry
[5] Bartoli G, Betti M, Giordano S. In situ static and dynamic investigations on the structures. J Eng Mech 1997;123(7):660–8.
‘‘torre grossa’’ masonry tower. Eng Struct 2013;52:718–33.
[21] Giamundo V, Sarhosis V, Lignola G, Sheng Y, Manfredi G. Evaluation of different
[6] Russo G, Bergamo O, Damiani L, Lugato D. Experimental analysis of the ‘‘Saint
computational modelling strategies for the analysis of low strength masonry
Andrea’’ Masonry Bell Tower in Venice. A new method for the determination of
structures. Eng Struct 2014;73:160–9.
‘‘Tower Global Young’s modulus e’’. Eng Struct 2010;32(2):353–60.
[22] Costa AA, Arêde A, Costa A, Oliveira CS. In situ cyclic tests on existing
[7] Lagomarsino S. Damage assessment of churches after L’Aquila earthquake (2009).
stone masonry walls and strengthening solutions. Earthquake Eng Struct Dyn
Bull Earthq Eng 2012;10(1):73–92.
2011;40(4):449–71.
[8] Poiani M, Gazzani V, Clementi F, Milani G, Valente M, Lenci S. Iconic crumbling
of the clock tower in amatrice after 2016 central Italy seismic sequence: [23] Costa AA, Arêde A, Costa A, Oliveira CS. Out-of-plane behaviour of existing stone
advanced numerical insight. Proc Struct Integr 2018;11:314–21. masonry buildings: experimental evaluation. Bull Earthq Eng 2012;10(1):93–111.
[9] Lang K, Bachmann H. On the seismic vulnerability of existing unreinforced [24] Menegotto M, Monti G, Salvini S, Vailati M. Improvement of transverse connec-
masonry buildings. J Earthquake Eng 2003;7(03):407–26. tion of masonry walls through afrp bars. In: Advances in FRP Composites in Civil
[10] Morandi P, Magenes G. Second order effects in out-of-plane strength of urm walls Engineering. Springer; 2011, p. 947–50.
subjected to bending and compression. ROSE Rep 2006. [25] D’Ayala D, Speranza E. Definition of collapse mechanisms and seismic vulnera-
[11] Ferreira TM, Costa AA, Costa A. Analysis of the out-of-plane seismic behavior of bility of historic masonry buildings. Earthquake Spectra 2003;19(3):479–509.
unreinforced masonry: A literature review. Int J Arch Heritage 2015;9(8):949–72. [26] Valluzzi MR. On the vulnerability of historical masonry structures: analysis and
[12] Housner GW. The behavior of inverted pendulum structures during earthquakes. mitigation. Mater Struct 2007;40(7):723–43.
Bull Seismol Soc Am 1963;53(2):403–17. [27] Spacone E, Sepe V, Raka E. Safety assessment of masonry building aggregates
[13] Sorrentino L, Masiani R, Decanini LD. Overturning of rocking rigid bodies under in Poggio Picenze, following the L’Aquila 2009 earthquake. In: 15th World
transient ground motions. Struct Eng Mech 2006;22(3):293–310. Conference on Earthquake Engineering, Lisbon; 2012:.

12
M. Mercuri et al. Engineering Structures 246 (2021) 113046

[28] nazionale per la difesa dai terremoti G, Doglioni F, Moretti A, Petrini V. Le [60] di Geofisica e Vulcanologia IN. The l’aquila seismic sequence. 2009, http://www.
chiese e il terremoto: dalla vulnerabilità constatata nel terremoto del Friuli al ingv.it.
miglioramento antisismico nel restauro, verso una politica di prevenzione. Lint; [61] Ameri G, Augliera P, Bindi D, D’Alema E, Ladina C, Lovati S, Luzi L, Marzorati S,
1994. Massa M, Pacor F, et al. Strong-motion parameters of the Mw = 6.3 Abruzzo
[29] Ivorra S, Pallarés FJ. Dynamic investigations on a masonry bell tower. Eng Struct (Central Italy) earthquake. 2009.
2006;28(5):660–7. [62] Chioccarelli E, De Luca F, Iervolino I. Preliminary study of L’aquila earthquake
[30] Abruzzese D, Miccoli L, Yuan J. Mechanical behavior of leaning masonry huzhu ground motion records V5. 20. 2009, ReLuis Report, http://www.reluis.it.
pagoda. J Cult Heritage 2009;10(4):480–6. [63] Decanini L, Liberatore L, Mollaioli F, Monti G, Al Shawa O. Studio preliminare
[31] Valente M, Milani G. Seismic assessment of historical masonry towers by della domanda sismica elastica ed anelastica in termini di energia, spostamento
means of simplified approaches and standard FEM. Constr Build Mater e forze (Rel. 1.0). 2009, Web Report Available at http://www.reluis.it.
2016;108:74–104. [64] Chioccarelli E, Iervolino I. Direttività e azione sismica: discussione per l’evento
[32] Sarhosis V, Milani G, Formisano A, Fabbrocino F. Evaluation of different de L’Aquila. Convegno Anidis 2009.
approaches for the estimation of the seismic vulnerability of masonry towers. [65] Code-NTC18 I. Norme tecniche per le costruzioni in zone sismiche. Ministerial
Bull Earthq Eng 2018;16(3):1511–45. Decree DM 2018.
[33] Committee I, et al. Recommendations for the analysis, conservation and [66] Cusatis G, Pelessone D, Mencarelli A. Lattice discrete particle model (LDPM) for
structural restoration of architectural heritage. 2005, See www.icomos.org. failure behavior of concrete. I: Theory. Cem Concr Compos 2011;33(9):881–90.
[67] Cusatis G, Mencarelli A, Pelessone D, Baylot J. Lattice discrete particle model
[34] Binda L, Zanzi L, Lualdi M, Condoleo P. The use of georadar to assess damage
(LDPM) for failure behavior of concrete. II: Calibration and validation. Cem
to a masonry Bell Tower in Cremona, Italy. Ndt E Int 2005;38(3):171–9.
Concr Compos 2011;33(9):891–905.
[35] Lourenço PB. Assessment, diagnosis and strengthening of Outeiro Church,
[68] Li W, Rezakhani R, Jin C, Zhou X, Cusatis G. A multiscale framework for the
Portugal. Constr Build Mater 2005;19(8):634–45.
simulation of the anisotropic mechanical behavior of shale. Int J Numer Anal
[36] Modena C, Valluzzi M, Folli RT, Binda L. Design choices and intervention
Methods Geomech 2017;41(14):1494–522.
techniques for repairing and strengthening of the monza cathedral bell-tower.
[69] Ceccato C, Salviato M, Pellegrino C, Cusatis G. Simulation of concrete failure
Constr Build Mater 2002;16(7):385–95.
and fiber reinforced polymer fracture in confined columns with different cross
[37] Anzani A, Binda L, Carpinteri A, Invernizzi S, Lacidogna G. A multilevel
sectional shape. Int J Solids Struct 2017;108:216–29.
approach for the damage assessment of historic masonry towers. J Cult Heritage [70] Lale E, Rezakhani R, Alnaggar M, Cusatis G. Homogenization coarse graining
2010;11(4):459–70. (HCG) of the lattice discrete particle model (LDPM) for the analysis of reinforced
[38] Carpinteri A, Invernizzi S, Lacidogna G. In situ damage assessment and nonlinear concrete structures. Eng Fract Mech 2018;197:259–77.
modelling of a historical masonry tower. Eng Struct 2005;27(3):387–95. [71] Rezakhani R, Scott DA, Bousikhane F, Pathirage M, Moser RD, Green BH,
[39] Carpinteri A, Invernizzi S, Lacidogna G. Historical brick-masonry subjected to Cusatis G. Influence of steel fiber size, shape, and strength on the quasi-static
double flat-jack test: Acoustic emissions and scale effects on cracking density. properties of ultra-high performance concrete: Experimental investigation and
Constr Build Mater 2009;23(8):2813–20. numerical modeling. Constr Build Mater 2021;296:123532.
[40] Lourenço P. Computational strategy for masonry structures. Delft University of [72] Han L, Pathirage M, Akono A-T, Cusatis G. Lattice discrete particle modeling of
Technology and DIANA Research; 1996. size effect in slab scratch tests. J Appl Mech 2020;1–32.
[41] Riva P, Perotti F, Guidoboni E, Boschi E. Seismic analysis of the Asinelli Tower [73] Alnaggar M, Cusatis G, Di Luzio G. Lattice discrete particle modeling (LDPM)
and earthquakes in Bologna. Soil Dyn Earthq Eng 1998;17(7–8):525–50. of alkali silica reaction (ASR) deterioration of concrete structures. Cem Concr
[42] Bernardeschi K, Padovani C, Pasquinelli G. Numerical modelling of the structural Compos 2013;41:45–59.
behaviour of Buti’s bell tower. J Cult Heritage 2004;5(4):371–8. [74] Pathirage M, Bousikhane F, D’Ambrosia M, Alnaggar M, Cusatis G. Effect of alkali
[43] Casolo S. A three-dimensional model for vulnerability analysis of slender silica reaction on the mechanical properties of aging mortar bars: Experiments
medieval masonry towers. J. Earthquake Eng 1998;2(04):487–512. and numerical modeling. Int J Damage Mech 2019;28(2):291–322.
[44] Casolo S, Milani G, Uva G, Alessandri C. Comparative seismic vulnerability [75] Pathirage M, Bentz D, Di Luzio G, Masoero E, Cusatis G. The ONIX model:
analysis on ten masonry towers in the coastal Po Valley in Italy. Eng Struct a parameter-free multiscale framework for the prediction of self-desiccation in
2013;49:465–90. concrete. Cem Concr Compos 2019;103:36–48.
[45] Bocciarelli M, Barbieri G. A numerical procedure for the pushover analysis of [76] Angiolilli M, Gregori A, Pathirage M, Cusatis G. Fiber reinforced cementitious
masonry towers. Soil Dyn Earthq Eng 2017;93:162–71. matrix (FRCM) for strengthening historical stone masonry structures: Experiments
[46] Gregori A, Castoro C, Mercuri M, Angiolilli M. Numerical modelling of the and computations. Eng Struct 2020;224:111102.
mechanical behaviour of rubbercrete. Comput Struct 2021;242:106393. [77] Angiolilli M, Pathirage M, Gregori A, Cusatis G. Lattice discrete particle model
[47] Gregori A, Castoro C, Mercuri M, Angiolilli M. Modeling the mechanical response for the simulation of irregular stone masonry. Journal of Structural Engineering
of rubberised concrete. In: Developments and Novel Approaches in Biomechanics 2021;147(9):04021123.
and Metamaterials. Springer; 2020, p. 341–52. [78] Cusatis G, Zhou X. High-order microplane theory for quasi-brittle materials with
[48] Marra AM, Salvatori L, Spinelli P, Bartoli G. Incremental dynamic and nonlinear multiple characteristic lengths. J Eng Mech 2013;140(7):04014046.
static analyses for seismic assessment of medieval masonry towers. J Perform [79] Cusatis G, Bažant ZP, Cedolin L. Confinement-shear lattice model for
Constr Facilities 2017;31(4):04017032. concrete damage in tension and compression: I. Theory. J Eng Mech
[49] Milani G, Russo S, Pizzolato M, Tralli A. Seismic behavior of the San Pietro di 2003;129(12):1439–48.
[80] Cusatis G, Bažant ZP, Cedolin L. Confinement-shear lattice model for concrete
Coppito church bell tower in L’Aquila, Italy. Open Civil Eng J 2012;6(1).
damage in tension and compression: II. Computation and validation. J Eng Mech
[50] Peña F, Lourenço PB, Mendes N, Oliveira DV. Numerical models for the seismic
2003;129(12):1449–58. http://dx.doi.org/10.1061/(ASCE)0733-9399(2003)129:
assessment of an old masonry tower. Eng Struct 2010;32(5):1466–78.
12(1449).
[51] Decreto del Ministro delle Infrastrutture 17 gennaio 2018. Aggiornamento delle
[81] Nguyen HT, Pathirage M, Cusatis G, Bažant ZP. Gap test of crack-parallel stress
‘‘Norme tecniche per le costruzioni’’, Gazzetta Ufficiale Della Repubblica Italiana,
effect on quasibrittle fracture and its consequences. J Appl Mech 2020;87(7).
Supplemento Ordinario n. 42.
[82] Nguyen H, Pathirage M, Rezaei M, Issa M, Cusatis G, Bažant ZP. New perspective
[52] Brandi C. Teoria Del Restauro. Ed. di storia e letteratura; 1963.
of fracture mechanics inspired by gap test with crack-parallel compression. Proc
[53] Giuffrè A. Sicurezza e conservazione dei centri storici. Il Caso Ortigia 1993;279.
Natl Acad Sci 2020.
[54] Bartoli G, Betti M, Vignoli A. A numerical study on seismic risk assessment [83] Decanini L, Liberatore L, Mollaioli F, Monti G, Al Shawa O. Terremoto aquilano
of historic masonry towers: a case study in san gimignano. Bull Earthq Eng del 6 aprile 2009 studio preliminare della domanda sismica elastica ed anelastica
2016;14(6):1475–518. in termini di energia, spostamenti e forze (rel. 1.0). 2009.
[55] Sepe V, Speranza E, Viskovic A. A method for large-scale vulnerability assessment [84] Decanini L, Liberatore L, Mollaioli F. Damage potential of the 2009 L’Aquila,
of historic towers. In: Structural Control and Health Monitoring: The Official Italy, earthquake. J Earthquake Tsunami 2012;6(03):1250032.
Journal of the International Association for Structural Control and Monitoring [85] per i Beni M, et al. Linee Guida Per la Valutazione E Riduzione Del Rischio
and of the European Association for the Control of Structures. 15, (3):Wiley Del Patrimonio Culturale: Allineamento Alle Nuove Norme Tecniche Per Le
Online Library; 2008, p. 389–415. Costruzioni. Gangemi Editore spa; 2011.
[56] Bazant ZP. Fracture and Size Effect in Concrete and Other Quasibrittle Materials. [86] Rainieri C, Fabbrocino G. Il periodo elastico delle torri in muratura: correlazioni
Routledge; 2019. empiriche per la previsione. In: XIV Congresso Nazionale L’Ingegneria Sismica
[57] Cusatis G, Cedolin L. Two-scale study of concrete fracturing behavior. Eng Fract in Italia, Bari. 2011, p. 18–22.
Mech 2007;74(1–2):3–17. [87] Applicativa delle NTC18 C. Circolare del Ministero delle Infrastrutture e dei
[58] Mercuri M, Pathirage M, Gregori A, Cusatis G. Computational modeling Trasporti 21 gennaio 2019, n7 recante istruzioni per l’applicazione delle nuove
of the out-of-plane behavior of unreinforced irregular masonry. Eng Struct norme tecniche per le costruzioni di cui al decreto ministeriale 20 febbraio 2018.
2020;223:111181. Suppl Ord 2019.
[59] Gregori A, Salem H, Harak T, Fassieh K, Khalil A. Collapse analysis of Santo [88] Magenes G, Penna A, Senaldi IE, Rota M, Galasco A. Shaking table test of a
Stefano tower using applied element method. In: Proceedings of 16th World strengthened full-scale stone masonry building with flexible diaphragms. Int J
Conference on Earthquake Engineering, 16WCEE, 2017:. Arch Heritage 2014;8(3):349–75.

13
M. Mercuri et al. Engineering Structures 246 (2021) 113046

[89] Magenes G, Penna A, Galasco A. A full-scale shaking table test on a two- [93] Modena C, Bettio C. Experimental characterisation and modelling of injected
storey stone masonry building. In: 14th European Conference on Earthquake and jacketed masonry walls, In: Proceedings of Italian-French Symposium
Engineering. 2010. Strengthening and Repair of Structures in Seismic Area, Nizza, France, 1994:
[90] ASTM. Standard test method for diagonal tension (shear) in masonry p. 273–282.
assemblages. 2010. [94] Tomazevic M. The influence of rigidity of floors on the seismic resistance of
[91] de Normalisation CE, et al. Eurocode 6: Design of Masonry Structures—Part 1- of old masonry buildings: shaking-table tests of stone-masonry houses: summary
1: General Rules for Reinforced and Unreinforced Masonry Structures. Brussels, report. Ljubljana: Institute for Testing and Research in Materials and Structures;
Belgium: Comité EuropÉen de Normalisation; 2005. 1992.
[92] Modena C. Interpretazione dei risultati ottenuti dalle prove in sito nell’ambito [95] Angiolilli M, Gregori A. Triplet test on rubble stone masonry: numerical as-
delle tre convenzioni con gli istituti di ricerca di firenze e milano e modellazione sessment of the shear mechanical parameters. Buildings 2020;10(3):49. http:
del comportamento strutturale dei componenti rinforzati. In: Convenzione Di //dx.doi.org/10.3390/buildings10030049.
Ricerca Tra la Regione Toscana E Il Dipartimento Di Costruzioni E Trasporti [96] Degli Abbati S, Rossi M, Lagomarsino S. Out-of-plane experimental tests on
Dell’Università Degli Studi Di Padova. 1999. masonry panels, In: Proceedings of the 2nd European Conference on Earthquake
Engineering and Seismology, 2014: p. 25–29.

14