Seismic Vulnerability Assessment of a Historical Church : Limit Analysis and Nonlinear Finite Element Analysis

The seismic vulnerability of a historical Basilica church located in Italy is studied by means of limit analysis and nonlinear finite element (FE) analysis. Attention is posed to the failure mechanisms involving the façade of the church and its interaction with the lateral walls. In particular, the limit analysis and the nonlinear FE analysis provide an estimate of the load collapse multiplier of the failure mechanisms. Results obtained from both approaches are in agreement and can support the selection of possible retrofitting measures to decrease the vulnerability of the church under seismic loads.


Introduction
Recent Italian seismic events, such as the Emilia (2012) and the L' Aquila earthquakes (2009), have demonstrated how churches are susceptible to damage when subjected to earthquake motions [1][2][3].In fact, designed to withstand vertical loads, churches in general present slender walls, lack of horizontal structures, weak or nonexistent connections among structural elements, absence of effective tie-rods to absorb arch thrusts [4], and irregular stone texture.These aspects, added to poor material performance especially in tensile stress mode [5], are among the reasons for the structural collapses of historical monumental buildings when subjected to seismic forces.
Cases of damaged or collapsed churches after earthquake events (sometimes also immediately after their renovations) are numerous.It is worthwhile to remember the collapse of St. Possidonio church (Italy) in 2012, the partial collapse of the church of Santiago in Lorca (Spain) in 2011, the collapse of the Catholic Cathedral of Port au Prince in Haiti in 2010 (right after its total restoration), and the partial collapse of the main church in St. Pietro di Coppito (Italy) in 2009, after consistent work of conservation.
Among these methods, the most suitable to determine the seismic vulnerability of masonry buildings are the latters that correspond to limit analysis and nonlinear FE analysis [15,16].Both approaches aim at providing the value of the load that activates the failure mechanisms of the macroelements in which the church can be subdivided.In fact, the behavior at collapse of a masonry church under a seismic event is rarely global since several out-of-plane collapses of macroelements may occur.The experience suggests to consider a limited number of possible failure mechanisms involving the main elements of the church such as the fac ¸ade, the apse, the naves, the triumphal arch, and the navy walls.
Within this context, the present study presents a nonlinear FE analysis coupled with a limit analysis applied to Madre Santa Maria del Borgo, an Italian church located in San Nicandro Garganico.In particular, the collapse mechanisms involving the fac ¸ade are analyzed in detail, since reputed potentially critical after a survey.Three mechanisms are identified: an out-of-plane mechanism of the fac ¸ade, a mechanism involving the fac ¸ade and the peripheral lateral walls, and a mechanism considering the interaction between the fac ¸ade, the peripheral lateral walls, and the navy walls.
The seismic vulnerability of the church has been analysed by means of the following: (i) a preliminary linear dynamic analysis of the whole structure performed to check if a global dynamic behavior of the structure exists and the corresponding global structural interaction among the single elementary parts; (ii) nonlinear FE analyses of the main structural macroelements interacting with the fac ¸ade performed to identify the collapse loads.Results are integrated with those provided by a limit analysis of each mechanism.
The comparison of the results obtained from the above studies shows a good agreement and provides an estimate of the seismic vulnerability of the church, in particular of the main fac ¸ade.
The main relevant contribution of the present study is the application of the limit analysis to a masonry church and the validation by a nonlinear analysis of a model prepared according the FEM.As a consequence, the present approach can be readily extended to analyze the seismic vulnerability of other similar churches.Thus, the results of the analyses can be useful to design appropriate retrofitting and restoration measures.
The paper is organized as follows.In Section 1, the church structure is described, highlighting the conservation state through an analysis of the existing cracking pattern.Section 2 presents a first investigation through a linear dynamic analysis of the whole structure.In Section 3, a detailed study of the main mechanisms involving the church fac ¸ade is presented using a nonlinear FE analysis coupled with a limit analysis.Some conclusive remarks end the paper.

Description of the Church Structure
Madre Santa Maria del Borgo is the main church of San Nicandro Garganico (Italy); see Figure 1.The construction began between 1573 and 1580 and was completed around 1693, when lateral small buildings and chapels were added to the original project.
The church measures 30 m and 25 m of length and width, respectively.The lateral walls, near the fac ¸ade, are 11 m high, while the fac ¸ade height is 14 m.
In general, monumental buildings, such as churches, are characterized by a massive thickness of the walls with few and comparatively small openings.In this case, the wall thickness varies approximately between 0.8 m along the main nave and 1.1 m at the apse; see Figure 2(a).These walls are made by double shell tuff stone with different schemes: opus incertum, opus mixtum, and opus quadratum, filled with rubble of the same material; see Figure 3.
The three naves are subdivided by two lines of five columns each.The columns have squared cross-section made by regular bricks of tuff stone.The roof of the church is made by timber except for the roof of the chapel Immaculate, C3, which is realized by masonry (Figure 2(a)).The chapel C4 is the only one that has a ribbed-slab concrete floor, underneath the roof structure.The main fac ¸ade is made by regular prismatic stone blocks.
It is interesting to note that the church has been built on an artificial horizontal basement higher than the other buildings around the church; see Figures 2(b) and 2(c).The crypt is located between the internal pavement level and the actual external ground level.

Conservation Status of the Church: Cracking Pattern.
The church presents a diversified cracking pattern characterized by horizontal and vertical damages (Figure 4).In particular, the area near the crypt shows a severe damage/cracking pattern; see Figures 4(b) and 4(c).Possible causes of this situation can be listed as follows: (i) a localized subsiding of the walls located over the crypt due to a settlement of the foundation walls, or (ii) a seismic interaction of the fac ¸ade and the nave walls associated with an out-of-plane mechanism of the fac ¸ade.In fact, the church was built on a seismic area and several earthquake events were recorded in this area from the III level to the IX level of the MCS intensity scale.

Linear Dynamic Analysis
A linear dynamic analysis has been carried out on a 3D finite element model of the church using ABAQUS [17][18][19].The 3D FE model has been realized using shell elements to model the midplane masonry walls and using truss elements to model the timber roof structures of the nave and aisles.Shell elements are also employed to model the main vault of the nave, the vault of the apse, the vaults of the C1 and C2 chapels, and the vault roof structure of the C3 chapel; see Figure 5.The model consists of 49781 shell elements and 30 truss elements for a total of 49674 nodes.
The analysis aims at finding the dynamic behavior of the church as-built.For this reason, a perfect matching of the walls without any assumed cracking pattern or reduction in wall stiffness is considered.
Boundary conditions have been applied according to the characteristics of the terrain and church pavement.In particular, nodes located at the ground level, see Figure 2(b), have been clamped, whereas only the horizontal displacement has been restrained for the nodes located at the pavement level.To the authors' knowledge, no experimental investigations have been conducted to assess the mechanical properties of the church constituent materials.As a consequence, in this study, the masonry constitutive parameters are derived from the literature [20] for similar materials: Young's modulus  = 1000 MPa, Poisson's ratio ] = 0.07, and mass density  = 2000 kg/m 3 .
Results of the analysis are collected in Table 1, where the frequencies of the main mode shapes, characterized by a mass participation factor greater than 3% in the longitudinal or in the transversal direction, are listed.These results show mainly local modes and just one global mode, no. 7, that corresponds to a frequency equal to 5.334 Hz.Mass participation sum of the first 30 eigen-modes results is equal to 55.5% and 55.2% of the total mass in the transversal and longitudinal directions, respectively.It should be noted that this proportion accounts also for the belfry mass.Belfry mode shapes are characterized by lower frequencies if compared with those related to the church structure.As a consequence, the mass participation is not improved significantly if belfry mass is removed from the previous sums.In fact, the sum of the mass participation of the first 30 eigen-modes is equal to 63.5% and 64.9% of the total mass in the transversal and longitudinal directions, respectively.This is a common result for monumental religious buildings, where the particular geometry does not allow to reach high mass participation factors even for relatively simple structures; see, for instance, [11,21].
Figures 6 and 7 show the transversal and longitudinal sections of the selected eigen-modes, respectively.It can be seen that modes are characterized mainly by local transversal behavior: excitation of the main vault or transversal excitation of lateral and nave walls.It is interesting to note that the longitudinal mode shape, no. 7, see Figure 6(d), involves the fac ¸ade and the nave walls.Detailed view of the longitudinal section of this mode shape is reported in Figure 8.In the following, this mode is employed to suggest the failure mechanism of the fac ¸ade macroelement for the nonlinear FE analysis.

Limit Analysis and Nonlinear FE Analysis
In this section, attention is focused on the study of the seismic vulnerability of the fac ¸ade of the church conducted by    The church has been subdivided into five main macroelements that are involved in the three mechanisms: the fac ¸ade, the lateral nave wall (wall between C4 and N3), the lateral nave wall (wall between N2 and C1-C2-C3-S2), the nave wall (wall between N3 and N1), and the nave wall (wall between N1 and N2); see Table 2.
The analysis of the failure mechanisms is based on the calculation of the parameter which corresponds to the activation of the mechanism, the so called collapse load multiplier.The approaches employed to determine the collapse load multiplier are described in the following.
Limit Analysis.The limit analysis represents a simple and effective tool for deriving an estimate of the ultimate strength capacity of masonry structures [15,22,23].This approach considers no tensile strength and infinite compression strength for the masonry with absence of sliding at failure.In particular, these hypotheses lead to the definition of the term "mechanism": the structure undergoes failure through the formation of cylindrical hinges (corresponding to disconnections and localized cracking) that subdivide the structure into macroelements subjected to inelastic deformations.
With reference to a generic mechanism, see Figure 10, the collapse load multiplier  0 can be evaluated by means of the principle of the virtual works: where   is the self-weight of each macroelement part ( entries) composing the kinematic mechanism;   is the weight transmitted to the macroelement by adjacent structures ( entries);  ℎ is the generic external force applied to a macroelement part ( entries);  , and  , are the horizontal virtual displacements of each macroelement centroid;  , and  ℎ are the vertical virtual displacements of each point of application of   and  ℎ , respectively, while   represents the work done by internal forces.
The true collapse multiplier can be found in the class of all the collapse multipliers that are obtained by varying the position of each hinge.Once the collapse multiplier  0 has been computed for the considered mechanism, the seismic acceleration of the activation of the mechanism  * 0 is defined as [24]  where  * is the effective mass that reads as where  +  is the number of applied forces   that cause horizontal displacements  , and  is the gravitational acceleration.
The investigated mechanism is verified comparing the spectral acceleration  * 0 with the seismic demand  * SLU , [25], that is, where  is a factor depending on the ground type.In this case,  is assumed to be equal to 1.0 because it refers to lithoid soil formations or homogeneous very rigid soil, including alterations in superficial layers of maximum thickness of 5 m (type A in [25]).   represents the spectral acceleration for the elastic design (period  = 0), being   the peak ground acceleration (PGA) for the specific location.In this case,   equal to 0.25 is considered. is the height of the centroid of masses activated by the mechanism and  is the total height; see Figure 10.
According to the Italian standards [24], for masonry structures  is equal to 2. For the church under consideration made of unreinforced stone masonry, a more severe choice of  is desirable.As a consequence, the parameter  is set to equal to 1.5.
Nonlinear FE Analysis.Finite element analyses are carried out using the commercial computer program ABAQUS.The tool allows to solve a variety of highly nonlinear structural problems for concrete [26], metallic [27], and masonry [28] structures.In order to model the nonlinear behavior of the masonry material, the smeared crack model is adopted.The smeared crack (or concrete model) is a multicrack model based on a yield surface with isotropic hardening and associated flow when the state of stress is predominantly compressive and it uses damaged elasticity to account for the cracking.The cracking phenomenon is modeled by means of the so-called crack detection surface: a simple coulomb line written in terms of the first and second stress invariants.Designed to predict the nonlinear behavior of concrete, the model has some limitations when applied to masonry.The model is not capable of predicting cyclic response and the use of associated flow usually leads to an overprediction of the volume strain.Nevertheless, the ABAQUS smeared crack model has proven to reasonably predict the masonry behavior in monotonic loading, as long as proper material definition is provided [29].
The model requires the definition of few parameters.In this study, the stress-strain curve in compression is taken according to experimental campaign done on similar materials [20].Moreover, the shape of the failure surface and the postcracking tensile behavior are required to calibrate correctly the model.Setting the failure ratios option (FR), compression strength   , tension strength   , and the tension stiffening option (TS), the failure surface and the postcrack behaviour are defined.The selected values result in FR = (1.12,0.1, 1.33, 0.2),   = 3.3 MPa, and   = 0.33 MPa.Threedimensional elements, such as hexahedral and tetrahedral elements, are employed to model the macroelements.This choice allows to better represent the nonlinear behavior of the material during the crack evolution [30][31][32].
The nonlinear FE analysis provides the load proportionality factor (LPF), that corresponds, at the last converged step, to the value of the load multiplier  0 that activates the failure mechanism.It is computed by means of a Riks static analysis that is constructed by imposing a horizontal set of forces generated by applying a horizontal acceleration to the whole structure, with increasing magnitude from 0 to , using a function, (), to shape the acceleration according to the Italian standards [24].The current magnitude of the load  total on the structure is given by where  ref is the load pattern defined in the current step and  0 is the dead load (weight load applied in the previous step).
The LPF may increase or decrease as the solution proceeds.

Mechanism 1: Out-of-Plane Mechanism of the Fac ¸ade.
Here, the out-of-plane mechanism of the fac ¸ade, as shown in Figure 9(a), is investigated.The fac ¸ade thickness is 1.1 m, the width is 17.2 m, and the maximum height is  = 11.7 m.The centre of gravity of the whole fac ¸ade is located at  = 6.11 m from the ground level.The limit analysis yields to the collapse multiplier for the out-of-plane mechanism:  0 = 0.09.The corresponding seismic acceleration  * 0 = 0.116 does not correspond to a safe condition, since  * SLU = 0.297.Besides, subdividing the church fac ¸ade according to the geometry of the openings, it is possible to identify different local mechanisms.They are not reported because they possess higher values of the collapse load multiplier.In particular, the collapse multiplier of the mechanism associated with the higher part of the fac ¸ade is  0 = 0.534.Moreover, a detailed description of the local mechanisms of the fac ¸ade is beyond the aim of the present study.For further details about this aspect and about the out-of-plane behaviour of unreinforced masonry walls see [33].
Figure 11 shows the comparison between  0 and LPF obtained from the nonlinear FE analysis.The load proportionality factor LPF is plotted versus the nondimensional displacement parameter /, where  is the displacement of the reference point of the mechanism, indicated with the letter A in Figure 11.The computation of LPF is carried out for values of the tension stiffening parameter TS ranging from 0 to 0.003 but, for clarity purpose, only two values of TS are plotted: TS = 0.001 and TS = 0.003.

Mechanism 2: Interaction between the Fac ¸ade and the
Peripheral Lateral Walls.Here, the failure mechanism of the  fac ¸ade with the lateral nave walls, as shown in Figure 9(b), is considered.
A preliminary in-plane FE analysis is performed.Figure 12(a) shows the resulting deformed configuration for the macroelement no. 3 (Table 2).The representation of the minimum (compression) and maximum (tensile) principal A 3D FE model has been constructed and a nonlinear analysis is performed by considering perfect matching between the fac ¸ade and the lateral walls.This assumption provides a mechanism of the as-built condition to compare with the assumed mechanism for the limit analysis shown in Figure 12(b).Figure 13 shows the results obtained using the nonlinear FE analysis and the limit analysis.The comparison between  0 = 0.175 and LPF = 0.18 confirms that results are in good agreement.The corresponding seismic acceleration  * 0 = 0.201 does not correspond to a safe condition with respect to  * SLU = 0.291 computed via (4).A value of 0.001 for the tension stiffening parameter TS was considered in the computations.

Mechanism 3: Interaction between the Fac ¸ade, the Peripheral Lateral Walls, and the Nave Walls.
Here, a further study of the potential failure of the fac ¸ade is carried out considering the interaction between the fac ¸ade macroelement and all the longitudinal macroelements (peripheral lateral walls and nave walls); see Figure 9(c).
As for the previous case, these macroelements (nos.4 and 5 in Table 2) are studied preliminarily by means of an inplane FE analysis.In Figure 14, the deformed configuration with the representation of the minimum (compression) and maximum (tensile) principal stresses is reported for the nave wall (N1-N2).This representation allows to identify the hinges to be used in the limit analysis.
A 3D FE model was constructed and the nonlinear FE analysis was performed by considering perfect matching between the fac ¸ade and the lateral party walls.Figure 15 shows the comparison between  0 = 0.33 and LPF = 0.34 obtained from the nonlinear FE analysis.Results are in good agreement.A value of 0.001 for the tension stiffening parameter TS was considered in the computations.Figure 16 shows the comparison between the results obtained from the nonlinear FE analysis for two control points: one in the middle of the fac ¸ade, denoted as point A, and one at the top of the fac ¸ade, denoted as point B.
The deformed configuration of the structure is illustrated in Figure 17.The mechanism is characterized by a pronounced horizontal displacement at the center of the fac ¸ade.The corresponding cracking pattern shows a severe damage localized at the top of the fac ¸ade, points E and Q, and the formation of two hinges along the lines F-O and A-G.This configuration is very similar to failure mechanism experienced by similar churches after earthquakes [34,35].Figure 18 summarizes the results obtained for all the investigated mechanisms.

Conclusions
In this paper, the analysis of the seismic vulnerability of a church structure has been presented throughout the study of three collapse mechanisms of its fac ¸ade.The analysis has been carried out by means of nonlinear finite element models and according to the Italian standards on monumental and historical masonry buildings.Investigations based on results obtained from limit analysis and nonlinear finite element analysis have been conducted on some macroelements with special attention to those that interact with the fac ¸ade.
The obtained results showed the primary role of the interlocking effect of lateral walls on the fac ¸ade behavior and the role of the conservation status of the church, its cracking pattern and its previous damage state.Analyses have allowed to obtain basic information about the structural behavior of the church under seismic loads highlighting that some of the studied mechanisms do not correspond to safe conditions.As a consequence, proper retrofitting actions can be designed.

Figure 1 :Figure 2 :
Figure 1: (a) Photograph of Madre Santa Maria del Borgo church, San Nicandro, Italy, and (b) fac ¸ade of the church.
Advances in Civil Engineering(a) Regular stone masonry: opus quadratum (b) Irregular stone masonry: opus incertum
means of limit analysis and of nonlinear FE analysis.For this purpose, three mechanisms are analyzed: (i) the out-of-plane failure mechanism of the fac ¸ade considering no interaction with the lateral walls (Figure9(a)); (ii) the failure mechanism due to the interaction between the fac ¸ade and the peripheral lateral walls (Figure9(b)); (iii) the failure mechanism due to the interaction between the fac ¸ade, the peripheral lateral walls, and the nave walls (Figure9(c)).

Figure 6 :
Figure 6: Linear dynamic analysis: horizontal section of the horizontal section of the main six eigen-modes.Vaults and roof structures are hidden to view.

Figure 7 :
Figure 7: Linear dynamic analysis: transversal section of the of the main six eigen-modes.

Figure 10 :
Figure 10: Limit analysis: representation of a generic mechanism due to the application of  ℎ force and self-weights.

Figure 11 :
Figure 11: Mechanism 1: comparison between nonlinear FE analysis and limit analysis for the out-of-plane collapse of the fac ¸ade.

Figure 12 :Figure 13 :
Figure 12: Nonlinear FE analysis: (a) principal stresses with the locations of crack openings for the lateral wall N2-(C1,C2,C3,S2).White and black colours correspond to compression and tension areas, respectively, and (b) illustration of the lateral section (see Figure 9(b)) used to compute the collapse load multiplier ( = 5.81 m and  = 11.7 m).

Figure 14 :
Figure 14: Nonlinear FE analysis: principal stresses with the locations of crack openings for the nave wall N1-N2.White and black colour corresponds to compression and tension areas, respectively.

Figure 15 :
Figure 15: Mechanism 3: comparison between nonlinear FE analysis and limit analysis for the in-plane collapse of the nave walls.

Figure 16 :Figure 17 :
Figure 16: Mechanism 3: nonlinear FE analysis for the out-of-plane collapse of the fac ¸ade considering the contribution of lateral and nave walls.Points A and B correspond with the location, where maximum and minimum displacements are measured, respectively.

Figure 18 :
Figure 18: Mechanisms involving the fac ¸ade macroelement.(a) Mechanism 1: no connection between the fac ¸ade and the longitudinal walls; (b) mechanism 2: connection between the fac ¸ade and the peripheral lateral walls; (c) mechanism 3: connection between the fac ¸ade and all the transversal walls.

Table 1 :
Linear dynamic analysis: frequencies of the main mode shapes of the St. Maria del Borgo Church (bell tower modes are not reported in the present table).