Engineering Structures 31 (2009) 14471459

Contents lists available at ScienceDirect

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

Seismic analysis and strengthening design of a masonry monument by a rigid

body spring model: The Maniace Castle of Syracuse
Siro Casolo a, , Carlo Alberto Sanjust b
a
Politecnico di Milano - Dipartimento di Ingegneria Strutturale, Piazza Leonardo da Vinci, 32, 20133 Milano, Italy
b
Politecnico di Bari - Dipartimento di Scienze dellIngegneria Civile e dellArchitettura, Via Orabona 4, 70125 Bari, Italy

article info abstract

Article history: The seismic analysis of a large monument subjected to strong earthquakes is the object of the present
Received 8 February 2008 paper. As a case study, the response of the Maniace Castle of Syracuse has been investigated by a
Received in revised form multi-level approach which adopts traditional finite element modelling as well as a specific mechanistic
9 January 2009
computational model for the final non-linear seismic analyses. At first, the linear behaviour of the
Accepted 11 February 2009
Available online 10 March 2009
monument was studied by means of two 3D FE models in order to understand the global response of
the building, its points of weakness, and the kinematics of the corresponding collapse mechanisms. Then,
Keywords:
these results were used to calibrate a mechanistic 2D plane rigid body and spring model, RBSM, specifically
RBSM formulated with the aim of approximating the macroscopic behaviour of masonry walls with reduced
Rigid element degrees of freedom, and taking into account the influence of the masonry texture and energy dissipation
Masonry texture capacity of the material. Given the uncertainties and the variability of the material characteristics, an
Pushover extensive parametric non-linear static analysis and the dynamic analyses, with artificial and natural
Dynamics accelerograms, were made in accord with the EC8 seismic code, and compared together. These results
Damage were then used to formulate and verify a simple proposal for the restoration design.
Seismic analysis 2009 Elsevier Ltd. All rights reserved.

1. Introduction resources. This difficulty is exacerbated by the need to account

for the uncertainties and variability of the material characteristics,
The analysis of the seismic response of large monumental that encourage one to adopt a statistical approach. On the other
masonry buildings is the object of the present research work. The hand, the material models commonly available in the frame of
size and the geometric complexity of these structures, together finite element computer codes are generally not proper, because
with the presence of specific architectonic parts, such as large of the intrinsic limitations of the Cauchy continuum approach to
faades, slender columns, arches and vaults, are often at the describe a masonry-like solid with orthotropic damage evolution
origin of heavy seismic damage even in the presence of a under cyclic loadings. Luckily, it is often effective to subdivide these
monuments into macro-elements (e.g. the faade, the lateral
moderate earthquake. This situation is typical of a large part of
walls, the triumphal arch, the bell tower, the apse), since the
the Italian architectural heritage. In theory, a reliable estimation
seismic response of these single parts tends to be dominated by
of the seismic risk would require to perform full dynamical
recurrent damage mechanisms, and the mechanical interrelation
analyses on the whole monument, by describing the non-linear
between contiguous parts is limited [5,6]. In particular, when
material response and the effective transmission and dissipation performing full dynamical analyses it is often preferable to adopt
of the seismic energy coming from the ground motion into the a two-dimensional model rather than a three-dimensional one [7],
structure [14]. even if the definition of a simpler model often requires a process
Unfortunately, at present, a detailed modelling of the non-linear of tuning in order to approximate the features of the specific
mechanical behaviour of masonry monuments in the frame of kinematics that is of interest.
three-dimensional dynamical analyses is not practically feasible. In the present paper a multi-step strategy is applied to the
In fact, on one hand the step-by-step full dynamical analysis of seismic analysis of the medieval Maniace Castle, in Syracuse
large models would require a huge amount of time-computation (Sicily, Italy) by using three different numerical models:
(i) a global three-dimensional finite element model of the whole
castle;
(ii) a detailed three-dimensional finite element model of the
Corresponding author. Tel.: +39 02 2399 4356; fax: +39 02 2399 4220.
E-mail addresses: [email protected] (S. Casolo), [email protected]
architectural part of the building that suffers the main static
(C.A. Sanjust). problems;

0141-0296/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.02.030
1448 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459

Table 1
Earthquakes with local intensity Is 5.0 registered in Syracuse.
Date (dd/mm/yy) Epicentral zone Is (MCS)

11/1/1693 East Sicily 9.0

15/6/1717 Syracuse 5.5
20/2/1818 Catania 5.0
11/1/1848 Augusta 6.5
11/9/1934 Madonie 5.0

Fig. 1. View of the castle from a drawing of 1630.

(iii) a two-dimensional rigid element model to perform the

dynamical analysis by considering the degradation of the
material subjected to cyclic loading [8,9].
The first two models have been used with the purpose of
studying the global behaviour of the structure, and to detect
the weak points and the parts of the castle which need further
investigations. These preliminary analyses also gave an indication
on the natural modes of vibrations and of the kinematics that are
more excited by the seismic loading.
Then, the third model has been used to investigate seismic dam-
age to the pillars, the buttresses and the arches of the masonry
monument. This is a mechanistic model that allows full dynam-
ical analyses, taking into account the peculiar behaviour of the
masonry material under loads that cause heavy mechanical degra-
dation. It considers the masonry material as an assembly of rigid
macro-scale elements connected by specific elasticplastic dam-
ageable springs. This hypothesis greatly simplifies the compu-
tations in comparison with the continuum approach, since it
permits dealing with uniaxial stress states at once. As a conse-
quence, dynamical analyses on realistic masonry structures can
be performed with limited computational effort, even when hys-
teretic and degradating materials are involved. The material de-
scription is made at a meso-scale level, that lies between the
micro-scale, at which the masonry is described in terms of single
material components (blocks and mortar joints), and the macro-
scale at which the masonry can be described by means of homo- Fig. 2. Plan and sections of the castle in the present condition.

geneous averaged parameters.

of the castle, as shown in the plan and two vertical sections of
2. The Maniace Castle Fig. 2; after the explosion, other masonry walls were erected in the
courtyard to create three new rooms. The exterior walls and the
The Maniace Castle was built in the 13th century (probably four circular towers belong to the original Middle Age building.
1232-1239 [10]) by the Holy Roman Emperor Frederick II In recent years the deterioration of some elements of the
Hohenstaufen on the island of Ortigia, which is most ancient part building has caused concern for its structural safety when
of the town of Syracuse. The original structure consisted of a large subjected to a possible seismic event. In particular, the four central
square hypostyle hall, named Salone, enclosed by large walls columns of the hypostyle hall present a number of vertical cracks.
and surrounded by four towers at the corners (see Fig. 1). The As a consequence, steel hoops have been provisionally installed on
large hall was originally covered by a double symmetric system the drums and on the capitals, as shown in the photo of the interior
of 24 cross-vaults, disposed 5 5 minus the central span. The hypostyle hall reported in Fig. 3.
castle mainly acted as a military facility until the end of the XIX For the future, the Soprintendenza ai Beni Culturali e Ambientali
century. During its long lifetime, earthquakes have been a cause of di Siracusa1 is planning the restoration of the whole castle and
damage for the castle. The main known seismic events that have nowadays it is still waiting for a reuse project. A number of
likely struck the castle are reported in Table 1. The corresponding studies and diagnostic investigations on the structure have been
MercalliCancaniSieberg intensity scale Is is also reported. commissioned. In particular ISMES [12], in 1999, carried out
The military use of the castle also caused major damage to the some tests, including: flat jack measurements,2 identification of
structure. In fact, a few years after the earthquake of 1693 that
caused minor damage to the vaults, in November 1704 lightning
struck the castle during a storm and a huge explosion of the 1 In Italy, the agency supervising the preservation of cultural and environmental
ammunition dump completely destroyed 3/5 of the vaulted hall heritage.
which had never been rebuilt in its original shape [11]. Other minor 2 A compression stress component in the range 3.34.2 MPa was measured in the
changes were made in 15th, 16th and 17th centuries. Presently, column near to the N-E side, while 0.44 MPa and 0.73 MPa were measured in the
only two spans of the original Salone survive on the SE side present two buttresses.
 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459 1449

Table 2
Frequency of the first four natural modes measured by ISMES [12] and calculated by the numerical models.
Mode ISMES [12] (Hz) University of Padua [13] (Hz) Present 3D model (Hz)

1 5.16 5.4 5.78

2 9.55 8.9 8.08
3 9.98 10.1 8.81
4 10.6 10.4 9.03

Table 3
Mechanical characterisation of the materials employed in the whole and partial 3D FEM model (Figs. 4 and 11).
Material Global 3D FEM model Partial 3D FEM model
Density (kg/m3 ) E (MPa) Density (kg/m3 ) E (MPa)
Columns, capitals 2200 10 000 0.2 2200 250020 000 0.050.2
Infill of hypostyle hall vault 1300 1 500 0.1 1100 400 0.1
Hypostyle hall vault (exterior side) 2200 10 000 0.1
Hypostyle hall vault (courtyard side) 1500 5000 0.1
Arches 2200 10 000 0.2
Roof of hypostyle hall 1000 350 0.1
Hypostyle hall boundary walls 2000 4 400 0.1 1800 5000 0.1
Buttresses 1800 5000 0.1

Fig. 4. Different materials used in the 3D FE model of the whole castle (see Table 3).

3. The two finite element models

3.1. The three-dimensional model of the whole castle

As first, the global behaviour of Maniace Castle was studied

by means of a large three-dimensional finite element model with
a total number of 64 196 solid elements. Due to the complex
geometry of the monument, and to avoid mesh distortion, the
castle was modelled in separate parts, then assembled in a single
model using tie constraints [15]. The hypostyle hall was modelled
with quadratic 20-node solid brick elements (named C3D20R
into the adopted Abaqus computer code [15]), whereas 10-node
tetrahedral elements (C3D10M) were adopted for the parts of
Fig. 3. The hypostyle hall of the castle. minor interest.
The geometry of this model was simplified by disregarding
principal vibrating modes and frequencies (Table 2), georadar the exact three-dimensional shape of the internal vaults, as well
analysis of the ground and probing of the walls. A global 3D as many other structural details. On the other hand, the elastic
FE model of the castle was made by Modena et al. [13], and, characteristics given to the different construction materials are
based on these results, a new diagnostic investigation focused on assigned in substantial accord with the data already published
the ultrasonic analysis of the columns was made by Ipertec in about this building [13]. However, due to the limited geometric
2003 [14]. In the same year, the Soprintendenza also commissioned accuracy of the model, a process of tuning of these values was
to the Politecnico di Milano a study for a more detailed numerical needed to approximate the basic fundamental modes of vibration
determination of the local stress conditions in the columns and of the castle (Table 2); the values assigned to the materials stiffness
in the arches of the hypostyle hall when subjected to seismic and to the mass density are reported in Table 3 and Fig. 4.
loadings. The following step consists in the evaluation of different Images of the deformed shapes corresponding to the first four
proposals for the seismic improvement of the structure. For this natural modes of vibration that mainly involve the hypostyle hall
study, an innovative rigid body and spring model has been adopted, are shown in Fig. 5. Mode 1 is mainly flexural and involves the
with the aim of performing dynamic analysis of the building, taking lateral displacement of the roof along the NWSE direction; mode
into account the non-linear behaviour and the damage of the 2 involves the lateral displacement of the roof along the NE-SW
masonry material. direction; mode 3 involves the vertical displacement of the roof
1450 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459

Fig. 5. Deformed shapes of the first four natural modes of vibration that involve the hypostyle hall of the castle.

and the vaults. Mode 4, as well as the other higher modes are not
important from a seismic point of view, because of their reduced
participation factor and effective mass.
Thanks to the tuning process, the matching of the mean values
of stress due to the gravity load with those reported from the
previously published experimental and computational studies was
acceptable in the most critical parts of the building: the columns
and the buttresses.
The spectral analyses, performed using the response spectra
prescribed by the Italian seismic code for the site of Syracuse,
and taking into account both vertical and lateral acceleration,
highlighted peaks of stress close to the cracking values for good
limestone in the central columns of the hall and in the buttresses. Fig. 6. Different materials used in the partial 3D FE model (see Table 3). The grey
These results picked out the need for a more accurate model of scale is proportional to the Young modulus given to the material.
the most stressed parts of the monument, with higher geometric
accuracy and taking into account the heterogeneities in the
columns, due to the presence of damaged materials, as resulted
from the diagnostic surveys [14].

3.2. The three-dimensional model of the hypostyle hall

An accurate three-dimensional finite element model was

assembled in order to get a deeper knowledge of the kinematics
and of the stress situation in the columns and in the arches of the
hypostyle hall in the linear elastic field. In this case, the geometry
was modelled with great detail, and the real shape of the cross-
vaults, the ribs, and all the structural details that can be significant
for the stress analysis were implemented. Moreover, the different
properties of the materials in the real building could be assigned
with more accuracy within the limits of the linear elastic field.
This detailed model had a total number of 48 004 brick elements
(named C3D8 in the adopted Abaqus computer code [15]).
The damage and material heterogeneities of the masonry block
stones were modelled by considering a reduced macroscopic Fig. 7. On the left: detail of the 3D FE model of the column with indication of the
Youngs modulus in the corresponding volume. In Figs. 6 and 7 it different materials. The grey scale is proportional to the value of Young modulus.
is shown how the masonry material was modelled in order to be On the right: map of the vertical axial stress S33 [Pa] caused by the action of gravity
as much as possible in accord with the presence of weak materials plus a lateral acceleration equal to 2.5 m/s2 .
 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459 1451

size or larger than the unit cell i , i.e. equal or larger than the
representative volume element of the masonry [21,9].
The elastic characteristics of these springs are defined by a
specific procedure of identification with the objective to transfer
some characteristics of the internal texture to the macro-scale
model [9]. On the other hand, a simplified phenomenological
approach [22], based on the experimental tests available in the
technical literature [2325], has been followed in order to model
the response under cyclic loading.
The geometry and the kinematics of the RBSM are described
by considering a global Cartesian coordinate frame {O, x, y} that is
placed with the x-axis parallel with the horizontal bed-joints. The
domain is partitioned into m quadrilateral elements such that no
vertex of one quadrilateral lies on the edge of another quadrilateral.
As shown in Fig. 10, on the left, a local reference frame {oi , i , i }
is fixed in the element barycentre oi , with the i -axis initially
parallel with the global x-axis. The deformed configuration of the
discrete model is described by the variations of position of these
local reference frames with respect to the global one, and the 3m
Lagrangian coordinates are assembled into vector {u} as follows:

{u}T = {u1 , v1 , 1 , u2 , v2 , 2 , . . . , um , vm , m } . (1)

The external loads, applied in the undeformed configuration,
are assembled into a vector of generalised external loads {p}, work-
Fig. 8. Deformed shape of the first natural mode of vibration of the detailed model conjugated with {u}, as follows:
of the hypostyle hall (top) and distribution of the vertical axial stress S33 [Pa] caused
by the action of gravity plus a lateral acceleration equal to 2.5 m/s2 (bottom).
{p}T = {p1 , q1 , 1 , p2 , q2 , 2 , . . . , pm , qm , m } . (2)

in the innermost parts of the columns, as detected by ultrasonic Three connection points named P, Q and R are defined for
surveys [12]. each side in common between two adjoining rigid elements, as
The deformed shape that corresponds to the first natural mode shown in Fig. 10, on the right, and the corresponding measures
of vibration of this partial model, with a frequency of 4.6 Hz, is of axial and shear strain are assigned to a volume of pertinence
shown in Fig. 8 (top), while the map of the vertical axial stress opportunely associated to each of these points. The following
is reported in the same figure (bottom) for the case in which the vector of generalised strains {} is assembled in order to contain
model was loaded by the vertical acceleration of gravity plus a the mean strain measures:
lateral acceleration equal to 2.5 m/s2 . This analysis showed that
n o
the central columns of the hypostyle hall were subjected to high
{}T = 1P , 1Q , 1R , 2P , 2Q , 2R , . . . , rP , rQ , rR (3)
values of stress, locally up to 10 MPa, quite close to the limit for an being r the number of sides that connect the elements of the whole
ancient masonry block of limestone.3 In the actual configuration of discrete model. A vector of generalised stresses { }, conjugated
the castle, no longer symmetric because of the collapsed arches in in virtual work, is correspondingly assembled, together with a
18th century, the present buttresses are clearly no longer capable diagonal matrix [V ] that contains the volumes of pertinence of the
of contracting and limiting the lateral displacements that can cause connecting devices.
collapse due to the central columns because of the combined action
of flexure and compression in the case of strong earthquakes. { }T = {1P , 1Q , 1R , 2P , 2Q , 2R , . . . , rP , rQ , rR } (4)

4. The RBS model [V ] = Diag{V1P , V1Q , V1R , V2P , V2Q , V2R , . . . , VrP , VrQ , VrR }. (5)
Then, by assembling a diagonal matrix [D] that contains the
4.1. Mechanical formulation elastic stiffnesses of all the connecting springs, we have:

The proposed plane model considers the heterogeneous solid

{ } = [D]{}. (6)
material as a mechanism consisting of rigid masses connected At this point we can apply the principle of virtual work as follows:
by simple elasticplastic springs, in the spirit of the rigid body
spring model (RBSM) [1618], and also other discrete element
{u}T {p} = W E = W I
methods applied to structural masonry modelling [19,20]. The core = {}T [V ]{ } = {u}T [B]T [V ][D][B]{u} (7)
of the present model is the unit cell defined by four rigid elements
connected to each other by two normal springs plus one shear being W E and W I the external and internal virtual work done by
the discretised system for an arbitrary virtual displacement {u}.
spring at each side, as shown in Fig. 9, on the right. This unit
Given the arbitrary of {u}, the following balance equation holds:
cell is a discrete mechanism that approximates the macroscopic
behaviour of the heterogeneous masonry material. The specificity {p} = [B]T [V ][D][B]{u} = [K ]{u} (8)
of the present model is that it considers some micro-structure
features of the composite material while it is designed to work being [K ] the generalised stiffness matrix of the discrete system.
at the macro-scale. The element size should of the same order of The present formulation is implemented in the dynamical field,
thus in order to consider the inertial forces, the mass of each
element mi and the rotational inertia about the barycentre i , are
assembled in the following diagonal generalised mass matrix:
3 Few specimens of similar limestone material were tested and the compression
strength was very scattered, in the range 10.7112 MPa [14,12]. [M ] = Diag{m1 , m1 , 1 , m2 , m2 , 2 , . . . , mm , mm , m }. (9)
1452 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459

Fig. 9. On the left: scheme of a regular masonry texture. On the right: an example of the corresponding unit cell made of rigid elements connected by line springs.

Fig. 10. On the left: couple of rigid elements with the adopted notation for the displacements. On the right: notation adopted for the external forces and the springs.

Fig. 11. Rigid element models with indications of the 8 different materials M1 M8 (Tables 4 and 5) assigned to the elements. Present buttress (right) and proposed new
buttress (left).

A viscous damping that acts essentially on the first few modes at 4.2. The plane rigid element model
low frequency is adopted by defining the viscous damping matrix
[C ] as proportional to the mass matrix [C ] = a0 [M ].4 Finally, In order to perform the non-linear dynamic analyses with
in the general case of non-linear force-displacement relation, the rigid elements, the geometry of the proposed plane model should
linearised incremental form of the system of equations of the be simple. Moreover, since this model was designed to make
motion of the discrete system during a time interval can be computations at the macro-scale, as well as to limit the difference
expressed as follows: [22]: of size of the elements in the various parts of the model, we
renounced to describe the material heterogeneities in the columns.
[M ]{u} + [C ]{u} + {f 0 } + [K ]{u} = {p} (10) The two meshes of the plane model in the actual configuration
and with the proposed buttress are shown in Fig. 11. Eight
being {f 0 } the vector of the internal generalised forces, and [K ] different materials (M1 M8 ) were assigned to these discrete
the generalised tangential stiffness matrix. The implicit method models. The characteristics of these materials were assigned to the
of Newmark is adopted to integrate this system of equations, elements after performing a fine-tuning in order to obtain a good
performing full NewtonRaphson iterations until convergence is approximation with the previous three-dimensional models.
attained. The main parameters of these materials in the linear elastic
field are reported in Table 4, together with the thickness given to
the corresponding elements. The numerical response in the linear
4 In order to obtain a damping factor equal to 5% for the first natural mode of elastic field proved to be satisfactory with regards to the correct
vibration, we have a0 = 3.35 s1 and a0 = 3.77 s1 for the rigid element model in prediction of the static stress due to the gravity load as well as
the actual configuration and with the proposed intervention, respectively. in the presence of a lateral load. A reasonable agreement with the
 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459 1453

Fig. 12. Deformed shape of the first two modes of the RBS model that approximate the first mode (left) and third mode (right) of the 3D FE model (Fig. 5).

Table 4
Mechanical characteristics in the linear-elastic field and element thickness of the RBSM (Fig. 11).
Material Density (kg/m3 ) Ex /Ey (MPa) 2G (MPa) Thickness (m)

(M1 ) Columns 2200 8000/10 000 7500 0.75

(M2 ) Roof 1000 1000 833 8.7
(M3 ) Infill of the vault (main) 1000 700 650 3.5
(M4 ) Perimeters main wall 1800 4000/5000 3500 8.7
(M5 ) Buttress 1800 4000/5000 3500 1.2
(M6 ) Arches and vault (main) 2200 5000 4000 4.75
(M7 ) Arches and vault (bottom) 2200 5000 4000 3.0
(M8 ) Infill of the vault (top) 1000 700 650 5.0

Table 5
Main characteristics assigned to the six materials for defining the non-linear elastic behaviour. The central values are reported, while variations of 10% and 20% have also
been considered into the parametric study. The symbols make reference to Fig. 13.
Material xc /yc (MPa) xt /yt (MPa) xr /yr (MPa) Ex1 /Ey1 (MPa) vm /hm (MPa) vr /hr (MPa) vr /hr 1000
1
(M ) Columns 4.0/5.0 0.8/1.0 0.4/0.5 2000 0.5/0.4 0.05/0.04 0.05
(M2 ) Roof 1.0 0.2 0.1 200 0.1 0.01 0.05
(M3 , M8 ) Infill of the vault 0.7 0.14/0.20 0.07/0.10 140 0.07 0.007 0.05
(M4 ) External main wall 2.0/2.5 0.8/0.6 0.4/0.3 800 /1000 0.15/0.10 0.02/0.015 0.05
(M5 ) Buttress 2.0/2.5 0.8/0.6 0.4/0.3 800/1000 0.15/0.10 0.02/0.015 0.05
(M6 , M7 ) Arches and vault 2.0/2.5 1.0/0.8 0.5/0.4 2000 0.15/0.10 0.02/0.015 0.05

previous 3D finite element models was also obtained for the first input accelerograms compatible with the local seismicity, and the
natural mode of vibration, both in terms of frequency and of global convenience to perform a large number of analyses with different
shape (see Fig. 12). The agreement for the higher modes is less parameters. For these reasons, and considering the unwillingness
satisfactory, since the frequency of natural mode 2, which mainly of adequate commercial codes, the Italian and the European
involves the vertical displacement of the roof and of the vault, seismic design codes [26] allow one to use non-linear static
is significantly higher than the frequency of the corresponding procedures [27,28] (usually referred to as pushover analyses) as
natural mode 3 of the global 3D model. In any case, the estimated well as full dynamic analyses. In this paper both analysis strategies
values of stress were correct, in particular for what regards the are performed and compared by adopting the proposed RBSM.
vertical axial stress in the columns and in the buttress, that are the
more critical parts of this structure. 5.1. The parametric study
The hysteretic rules assigned to the shear and axial springs are
sketched in Fig. 13, [24,25], while the main values of the material The uncertainty in the determination of the material parame-
parameters that govern the post-elastic behaviour are reported in ters has been dealt with by a parametric study. To limit the num-
Table 5. The adoption of these specific and relatively simple rules ber of varying parameters, a variation of the whole skeleton curve
allowed to perform a large set of non-linear static and dynamic that defines the hysteretic response of the axial and shear springs
analyses whose results are presented in the following section. This has been considered. Fig. 14 shows an example of this proce-
was the basis for the design of the new buttress shown in Fig. 11, dure applied to the case of a shear spring: the positive branch
on the right, that is proposed as a possible strategy to improve the of the skeleton curve OEYUS is translated to OEn+ Yn+ Un+ Sn+ or to
seismic performance of the hypostyle hall. OEn Yn Un Sn by moving the points E, Y , U, S, along a direction
parallel with the initial stiffness of the spring to obtain an increase
5. Seismic analysis or a reduction of the stress equal to n% of the value at the elastic
limit. The analogue has been also done for the negative branch as
Assessing the seismic vulnerability of a masonry monumental well as for the axial springs.
building by carrying out full dynamic analyses in the non-linear The six rows of Table 5 report the central values given
elastic field is a challenging task, due to many factors: the to the material parameters in order to define their non-linear
uncertainties in defining the constitutive laws for the materials, the response. These parameters are varied in order to produce a
problems of convergence in solving non-linear problems, the large set of materials improved or worsened by 10% and 20%.
computational resources required, the difficulty in the choice of In this manner, for each material M i we have five cases,
1454 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459

Fig. 13. Scheme of the hysteretic rules of the axial (top) and of the shear spring
(bottom).

Fig. 15. Forcedisplacement curves obtained by parametric analyses. The solid line
corresponds to the model response with the central values given to all the materials.
The shaded area represents to the median curve the mean deviation from it.

5.2. Pushover analyses

In the pushover analysis, the numerical model is loaded by a

constant gravity load plus a monotonically increasing distribution
of horizontal loads with the aim of pushing the structure
beyond the linear elastic field [28]. The basic idea is that the
resulting response conveniently represents the envelope of a
Fig. 14. Variation of the skeleton curve in parametric analyses.
representative set of possible dynamical responses, and thus it can
be used as an alternative to a full non-linear dynamic study.
(Mi20 , Mi10 , Mi0 , Mi10+ , Mi20+ ), and a total number of 56 = 15 625 In our parametric study, a force-displacement curve was
possible combinations that unfortunately makes the running time obtained from each single analysis, then the median force value
to perform a complete parametric study too large5 even for and the mean deviation from this median were evaluated in
the present specific discrete model. In order to reduce the time correspondence of each value of displacement. Fig. 15 shows the
required to perform the parametric study, a significant subset resulting median force-displacement curve for the old and for the
new buttress. In the same figure the force-displacement curves
of all the possible combinations has been selected by separately
that correspond the best (+20%) and the worst (20%) set of
considering the (Mn10 , Mn , Mn10+ ) and (Mn20 , Mn , Mn20+ ) sets of
material are also plotted, while the shaded area shows the mean
materials. Thus, in this case the number of required analyses is
deviation from the median.
limited to 2 36 1 = 1 457.
By fixing the values for one single material and by considering
the set obtained by varying the other five materials we also made
a sensitivity analysis that is summarised in Fig. 16. In this figure,
5 In fact, given an average duration of about one hour for a single non-linear the average of the maximum reaction force reached during the
analysis performed on a CPU Intel Xeon E5365, 3.0 GHz, with Linux OS and Intel non-linear static analyses are plotted against the variation of the
Fortran compiler, this would require over 650 days of computation! parameter that remained fixed. It is apparent that the quality of
 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459 1455

The specificity of the present monument, and in particular

the fact that the heavy roof is supported by a relatively slender
structure made of columns and arches, suggests to cautiously
fix a limit state of near collapse (NC) in correspondence of the
global equilibrium loss as long as growing lateral accelerations
are applied. From these analyses, the results show that the
corresponding maximum lateral acceleration value of median
response of the original structure was 2.95 m/s2 (3.06 m/s2 for the
central curve), while with the new buttress the maximum lateral
acceleration of the median curve was 3.33 m/s2 (3.71 m/s2 for the
central curve). After all, this analysis pointed out a limit state of
near collapse (NC) for the two examined situations, characterised
by the limit value of the lateral acceleration triggering the collapse
mechanism (in other terms, the maximum base shear Fb ) and
the corresponding maximum available displacement dm . Hereafter,
this specific limit state will be referred to in order to illustrate and
discuss the applications.
In what follows, the force-displacement relationship F d for
the equivalent SDOF system is calculated on the basis of the median
curves:
Fb dn
F = d = (11)

being the modal participation factor6 :
m
mi i
P
i=1 m
= m
= m
(12)
mi i2 mi i2
P P
i=1 i=1

where is the first eigenvector, normalised on the displacement

of the control node. The curve F d was then transformed into
Fig. 16. Sensitivity curves obtained on the basis of the parametric study.
an elastic-perfectly plastic relationship, by choosing the stiffness
of the elastic branch in such a way that the areas under the
the central column (M1 ) has the main impact on the maximum actual and the bilinear force-deformation curves are equal. In the
lateral force sustainable by the model. The material of the external equivalent bilinear curve, the yield force Fy , which represents also
main wall (M4 ) and of the buttress (M5 ) have also a significant the ultimate strength of the idealised system, is given by:
effect that is more pronounced in the case of the proposed new
Fy = Fy / (13)
buttress. From this study, it seems also interesting to note that a

large improvement of a single material quality does not necessarily and the corresponding yield displacement dy is calculated by the
imply a corresponding enhancement of the global performance. following expression, provided by Eurocode 8 [26]:
This is due to the problems related with the conflicting interaction
!
Em
of high quality and low quality materials. dy = 2 dm (14)
The pushover methodology, according with Eurocode 8 [26], Fy
followed these fundamental steps: where Em is the actual deformation energy up to the formation
of the plastic mechanism, and dm = dm / is the maximum
(i) determination of the global force-displacement relationship
displacement corresponding to the NC state.
for the structure (pushover or capacity curve);
At this point, the actual seismic demand for the inelastic
(ii) transformation of the capacity curve into a bilinear force- system is needed, that is to say according to a displacement
displacement relationship for an equivalent SDOF system; based approach the displacement demand (target displacement)
(iii) definition of the target displacement, i.e. the displacement should be determined according to the code standards.7
demand for the SDOF under the design spectrum in the The procedure can be graphically visualised in the Acceleration-
acceleration-displacement (AD) format; Displacement (AD) plane8 where the design spectrum9 is plotted
(iv) transformation of the target displacement for the SDOF
system into the corresponding displacement demand for the
original structure; 6 = 1.3 for the old buttress and = 1.29 for the new buttress.
(v) verification of the displacement compatibility of the structure 7 In both cases T < T , causing the structures to fall in the short period range
C
d
 
with the displacement demand, according to the seismic ([26] B.5). Thus, we have Fy /m Se (T ) dt = et
1 + (qu 1) TC , where
T
qu
design code.
det is the target displacement for a SDOF elastic system.
8 In this plane the design spectrum is represented by a parametric curve
In the present example, due to the squat shape of the structure,  
S (T )
a uniform pattern of applied lateral accelerations was applied. The Se (T ), e2 , where the period of the equivalent SDOF system is T =
r
average displacement dn (control node displacement) of two points 2
m d
y
.
F
(P 1 and P 2 in Fig. 11) and the value of the base shear force Fb were y

9 The spectrum was calculated for rock soil (type A) according to Eurocode 8
monitored during the process in order to obtain the non-linear
(Section 3.2.2.2). The values used for parameters were: S = 1.0, TB = 0.15,
force-displacement curves shown in Fig. 15. It is worth observing TC = 0.40, TD = 2.0. The design ground acceleration ag = 0.25 g, corresponding
that in both the cases there is a sudden decrease of the strength to a return period T = 475 years, was provided by the Italian sesmic code (O.P.C.M.
after the peak base reaction force is attained. 3274) for seismic zone 2 [29].
1456 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459

Table 6 recorded in Colfiorito (Umbria, Italy) on September 3rd, 1997, as

Target displacement dt of the MDOF systems compared with the maximum
well as by two couples of artificial accelerograms, are presented in
available displacements dm obtained by pushover analyses.
some detail. The linear elastic response spectra of these signals and
Old buttress New buttress
the corresponding acceleration time-histories are shown in Fig. 18.
dt (mm) 9.4 6.6 The artificial accelerograms were selected with the criterion of
dm (mm) 6.7 7.4 being spectrum-compatible with an estimated earthquake with
a return period T = 475 and T = 140 years for the site of the
Castle. These accelerograms were generated according with the
method proposed by Sabetta and Pugliese [30], in case of seismic
events with a magnitude Ms of 5.5. The intensity of the couple of
accelerograms of Colfiorito can be considered as reasonable for a
return period of T = 140 years for this site.
The time-history responses, in terms of average displacement of
points P 1 P 2 , are shown in Fig. 19. These dynamic analyses begin
with a ramp application of the vertical gravity acceleration, and
then, after few seconds, the analyses continue with the application
of the seismic acceleration while the gravity is kept constant. As
first, from Fig. 19 (left), we observe that in the case of Colfiorito
earthquake the model with the old buttress suffers a permanent
horizontal displacement of about 0.9 mm, similar to the result
Fig. 17. Verification of the displacements in the AD format.
obtained with the couple of artificial accelerograms with T =
140, while the dynamical analysis could not be completed in the
together with the (F /m , d ) bilinear curves for the present struc- case of the couple of artificial accelerograms with T = 475, due
ture and for the new proposed structure (Fig. 17). to the collapse of the column. The time-history of the average
The target displacement dt (for the SDOF system) and dt = vertical displacement of points P 1 P 2 is shown in the right side of
dt (for the MDOF structures) are finally compared with the Fig. 19. The permanent lowering is clearly related with the crushing
displacement resources obtained by the non-linear static analyses. of the masonry material of the central column that is subjected
From the values reported in Table 6 the results show that the old to the repeated compression-flexure cycles of the loading, and
buttress does not verify the NC limit state, while the new proposed it is evident the collapse due to compression with the artificial
buttress appears to be safe. accelerograms of T = 475, at time t = 4.5 s. Thus, the global
lowering is a consequence of the interaction between the flexural
5.3. Dynamic analyses and the axial loading. It is worth noting that the proposed new
buttress avoids the collapse in the case of T = 475, it reduces the
The dynamic analyses in which the proposed RBS models were lateral displacements, and it also gives a significant benefit in terms
loaded by the couple of horizontal (NS) and vertical accelerograms of reduction of the crushing in central column. It is noticeable that

Fig. 18. Elastic response spectra at 5% of damping (first row), and time histories of the horizontal (left) and vertical (right) accelerograms used for dynamic analyses.
 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459 1457

Fig. 19. Time history of the mean horizontal (left) and vertical (right) displacement of points P 1 P 2 (see Fig. 12) subjected to recorded and artificial generated earthquakes.

Fig. 20. Deformed shape and vertical axial stress under simple gravity load (a), after the application of Colfiorito accelerograms (b), after the couple of artificial accelerograms
with T = 475 years (c), just before the collapse of the column, and after the couple of artificial accelerograms with T = 140 years (d).

these non-linear dynamic results are in substantial accord with the at the end of the application of the earthquake accelerograms. As
results previously obtained with the pushover analysis. expected, the two most critical elements are the buttress at the
Figs. 20 and 21 compare the maps of the deformed shape right side of the model, and the central column. It is remarkable
and of the vertical axial compression at the end of the simple that the buttress mainly suffers damage due to the shear action that
application of the vertical gravity load with the maps obtained produces a permanent shear deformation as well as a reduction
1458 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459

Fig. 21. Deformed shape and vertical axial stress under simple gravity load (a), after the application of Colfiorito accelerograms (b), after a couple of artificial accelerograms
with T = 475 years (c), and after a couple of artificial accelerograms with T = 140 years (d).

of the lateral stiffness. As a consequence, the central column also intervention that increases the seismic strength and ductility,
suffers from flexural deformation due to the lateral displacements without a significant change of the stress state under the static
of the roof, that combines with the peaks of vertical axial stress gravity conditions, and avoiding having to modify the present
due to the vertical seismic acceleration. This is the main reason masonry structure of the vaulted hypostyle room.
for the reaching of compression stress values that are close the A noticeable outcome of this work is also the substantial
strength of the masonry material. It appears that the proposed new agreement between the non-linear dynamic results and the results
buttress seems to be capable of reducing the lateral displacements obtained with the pushover analysis.
without suffering significant shear damage. It is also worth noting
that the proposed intervention does not significantly modify the Acknowledgments
stress regime with the simple gravity loading.
This work was carried out thanks to an agreement between
6. Concluding remarks the Politecnico di Milano and the Soprintendenza ai Beni Culturali
e Ambientali di Siracusa. The Authors wish to thank, in particular,
The seismic analysis and the strengthening design of the arch. Mariella Muti from the Soprintendenza di Siracusa, prof.
Maniace Castle, in Syracuse, has been approached by means Vincenzo Petrini from the Politecnico di Milano, and prof.
of a complex multi-step strategy that integrates the diagnostic Giuseppina Uva from the Politecnico di Bari, for their valuable
investigations, the traditional FE modelling and a new specific rigid comments and suggestions.
body and spring model (RBSM) capable of performing non-linear
dynamic analyses. References
The first part of the work was devoted to the global analysis of
the building in the linear elastic field, with the aim of individuating [1] Bertero VV. Innovative approaches to earthquake engineering. In: Oliveto G,
editor. Innovative approaches to earthquake engineering. WIT press; 2002.
the main damage mechanism that can be activated by seismic
p. 154 [Chapter 1].
excitation. Then, the non-linear seismic response was investigated [2] Housner GW. Earthquake engineering. Strong ground motion, Prentice-Hall;
in order to propose a strengthening strategy that should have 1970. p. 7591 [Chapter 1].
the minimum architectural impact on the existing monumental [3] Petrini V, Casolo S. Vulnerability of historical and monumental buildings:
Significant ground motion parameters and evaluation of the seismic
structure. This is desirable when dealing with historical buildings, performance. In: Oliveto G, editor. Innovative approaches to earthquake
because in this case it is often necessary to take into account the engineering. WIT press; 2002. p. 20328 [Chapter 4].
effective hysteretic behaviour, with the energy dissipation and the [4] Uang CM, Bertero VV. Evaluation of seismic energy in structures. Earthq Eng
Struct Dyn 1990;19(1):7790.
mechanical degradation during the seismic excitation. [5] Doglioni F, Moretti A, Petrini V. Le Chiese e il Terremoto. Trieste: Lint press;
The RBSM proved to be effective in this context, since it allowed 1994 [in Italian].
to enhance our awareness by allowing to perform pushover as well [6] Petrini V, Casolo S, Doglioni F. Models for vulnerability analysis of monuments
and strengthening criteria. In: Proceedings of the XI European conference on
as non-linear dynamic analyses, in accord with the procedures of earthquake engineering. Volume of invited lectures, 1999. p. 17998.
the European seismic design code [26]. As an application example [7] Loureno PB. Computations on historic masonry structures. Prog Struct Eng
and numerical test, we have proposed and verified a simple Mater 2002;4(3):30119.
 S. Casolo, C.A. Sanjust / Engineering Structures 31 (2009) 14471459 1459

[8] Casolo S, Pea F. Rigid element model for in-plane dynamics of masonry walls [20] Lemos JV. Discrete element modeling of masonry structures. Internat J Archit
considering hysteretic behaviour and damage. Earthq Eng Struct Dyn 2007; Heritage 2007;2(1):190213.
36(8):102948. [21] Casolo S. Modelling in-plane micro-structure of masonry walls by rigid
[9] Casolo S. Macroscale modelling of the micro-structure damage evolution by a elements, International Journal of Solids and Structures. Internat J Solids
rigid body and spring model. J Mech Mater Struct [in press]. Structures 2004;41(13):362541.
[10] Bottari S. Monumenti svevi di Sicilia. Palermo: Societ Siciliana per la Storia [22] Casolo S. Macroscopic modelling of structured materials: relationship
patria; 1950. [in Italian]. between orthotropic cosserat continuum and rigid elements. Internat J Solids
[11] Agnello G. Larchitettura Sveva in Sicilia Tivoli; 1935 [in Italian]. Structures 2006;43(3):47596.
[12] ISMES (Enel Group). Castello Maniace - Siracusa, indagini diagnostiche sulle [23] Atkinson RH, Amadei PB, Saeb S, Sture S. Response of masonry bed joints in
strutture murarie. 1999 [in Italian]. direct shear. ASCE J Struct Engrg Div 1989;115(9):227696.
[13] Modena C, Franchetti P, Zonta D, Menga R, Pizzigalli E, Ravasio F, Muti M, [24] Magenes G, Calvi GM. In-plane seismic response of brick masonry walls. Earthq
Meloni R, Bordone G. Static and dynamic analyses of Maniace Castle in Eng Struct Dyn 1997;26(11):1091112.
Siracusa-Sicily. In: Loureno PB, Roca P, editors. Historical constructions. [25] Naraine K, Sinha S. Cyclic behaviour of brick masonry under biaxial
Guimaraes; 2001. p. 93343. compression. ASCE J Struct Engrg 1991;117(5):133655.
[14] Ipertec srl. Indagini strutturali sulle murature del Castello Maniace. 2003 [in [26] European standard approved by CEN on 23 April 2004. EuroCode8: Design of
Italian]. structures for earthquake resistance - Part 1: General rules, seismic actions
[15] Hibbitt, Karlsson & Sorensen, Inc. ABAQUS manuals, version 6.5. 2005. and rules for buildings.
[16] Griffiths DV, Mustoe GGW. Modelling of elastic continua using a grillage of [27] Fajfar P. Capacity spectrum method based on inelastic demand spectra. Earthq
structural elements based on discrete element concepts. Internat J Numer Eng Struct Dyn 1999;28:97993.
Methods Engrg 2001;50:175975. [28] Fajfar P. A nonlinear analysis method for performance based seismic design.
[17] Kawai T. New discrete models and their application to seismic response Earthq Spectra 2000;16(3):57392.
analysis of structures. Nucl Eng Des 1978;48:20729. [29] Presidenza del Consiglio dei Ministri Ordinanza 3274 come modificata
[18] Ostoja-Starzewski M. Lattice models in micromechanics. Appl Mech Rev ASME dallOPCM 3431 del 3/5/05 Italian seismic code. 2005 [in Italian].
2002;55(1):3560. [30] Sabetta F, Pugliese A. Estimation of response spectra and simulation of
[19] Chen S-Y, Moon FL, Yi T. A macroelement for the nonlinear analysis of in-plane nonstationary earthquake ground motions. Bull Seismol Soc Amer 1996;(86):
unreinforced masonry piers. Engrg Struct 2008;30(8):224252. 33752.