Dynamic testing and health monitoring via wireless sensor networks in the post-earthquake assessment of structural conditions at LAquila

E. Antonacci, A.M. Ceci, A. Colarieti, V. Gattulli, F. Graziosi, M. Lepidi, F. Potenza CERFIS, University of LAquila, Localit Monticchio 67040, LAquila, Italy, web: www.cerfis.it email: elena.antonacci@univaq.it, alfredo.ceci@dm.univaq.it, andrea.colarieti@univaq.it, vincenzo.gattulli@univaq.it, fabio.graziosi@univaq.it, marco.lepidi@univaq.it, francesco.potenza@email.it

ABSTRACT: The use of vibration measures during ambient dynamic testing or integrated within permanent structural health monitoring systems may play a fundamental role in urban areas rich of strategic and monumental buildings, especially in postearthquake scenarios, as the city of LAquila, struck by the devastating seismic events of April 2009. In particular, a careful use of different output-only identification procedures may help in extracting the structural signature from low-cost and easy-todeployed wireless networks of dynamic sensors. This valuable experimental information may significantly increase the general confidence in understanding the real dynamic behavior of the structures which suffered moderate or severe damage due to the seismic action. The paper presents a pair of case-studies related to a historical church (the Basilica di Collemaggio) and some modern buildings (the Engineering Faculty of LAquila), currently object of wide-spectrum analyses, including dynamic testing and structural model updating aiming to design retrofitting interventions. On this respect, the initial deployment and the current development of a permanent monitoring system based on a wireless network of accelerometers is illustrated. KEY WORDS: modal identification; structural health monitoring; wireless sensor network; smart sensors; cultural heritage. 1 INTRODUCTION Urgent requests currently pressing on the research areas of smart sensors and information technology deal with the technological development and subsequent testing of: (i) wireless sensing networks, designed to both reduce the operational disadvantages and costs of traditional wired networks, and to improve the data quality; (ii) smart sensors, with on-board microprocessors, for the decentralization of algorithmic elementary functions; (iii) low-cost, small-size, and low-consuming sensors for an autonomous, continuous and distributed employment [1,9,10]. Recently the use of a wireless sensor network (WSN) for permanent structural health monitoring (SHM) of historic buildings has been proposed [11] and a specific activity was conducted to customize the system for the continuous assessment of the integrity of frescos. The use of dynamic measurements either during ambient vibration testing or through permanent structural health monitoring may play an important role in an area densely populated of strategic and monumental structures after the occurrence of a devasting earthquake, such as the case of LAquila [12]. In particular, a careful use of different outputonly identification procedures may help in extracting the structural signature from low-cost and easy-to-deployed wireless networks of dynamic sensors. This valuable experimental information may significantly increase the general confidence in understanding the real dynamic behavior of the structures which suffered moderate or severe damage due to the seismic action. In the sense, the ongoing activities, involving the contributions from many different scientific disciplines available at CERFIS (www.cerfis.it), on structural health monitoring are illustrated in the following. A pair of case studies is presented related to different goals going from to the development of permanent monitoring of provisional scaffolding efficacy in preventing the degradation

The structural health monitoring is rapidly emerging as a valuable diagnostic tool for the early detection of unexpected reductions in the designed performance of facilities and infrastructure, in many civil engineering applications. The monitored quantities can in fact provide useful information on the progressive and inevitable decay of the resistant capacity of aging structures, in both the short and long period. The knowledge of the past time-evolution, even of a few significant data, can ensure the proper assessment of the current state, both in terms of safety with respect to the limit strength, and in terms of functionality with respect to the current serviceability. An adequate monitoring program is therefore among the essential prerequisites to suitably extent the structures' useful life, by means of an aware and sustainable management of the ordinary and extraordinary maintenance needs. In many countries worldwide, the strategic importance recognized to the topic by the most recent guidelines from national and supranational codes contributes towards increasing the demand [1]. Simultaneously, the recent and extraordinary technological developments in the areas of sensors, computers and communications make available nowadays the resources and knowledge sufficient to potentially overcome some limitations, even of economic nature, which have traditionally delayed the large-scale application of the structural health monitoring [2-8]. Despite emerging from a scattered scenario of proposals and solutions, the current frontier of theoretical and applied research in the field is clearly oriented to the use of smart and low-cost sensors, combined with advanced systems of communication, and completed by distributed elaboration strategies.

of damaged monumental historic structures to the dynamic testing and structural model updating usable for seismic retrofitting design. Permanent monitoring during reconstruction can add valuable information on the efficiency of the new structural solutions. 2 ONGOING ACTIVITIES TOWARDS SHM-WSN

of several stochastic state space models, built to reproduce its experimental response, and characterized by increasing order n. Therefore, the order of the model (or the subspace dimension), which better approximates the experimental response, is a matter of identification too. 2.2 Finite element modeling and updating

A group of researchers of CERFIS with complementary skills is conducting a wide plan of activities in the field of dynamic testing under environmental loading and structural health monitoring for a series of buildings, with strategic or historical value, at LAquila. In the following a synthetic description of the most challenging findings is reported. In order to achieve adequate level of confidence on the structural dynamic behaviour of the studied buildings a schedule of consequent activities are currently performed: (i) on-site dynamic testing under environmental actions with standard equipments [13-16]; (ii) finite element modelling based on exhaustive survey and material testing; (iii) definition of SHM-WSN sensor features; (iv) laboratory dynamic testing on 1:3 scaled frame in order to validate procedures and wireless monitoring sensors; (v) deployment of structural health monitoring systems with wireless smart sensors; (vi) development and installation by remote programming of modal and damage identification procedures taking into account temperature variation effects. All activities are at different stages of development, therefore in the following a synthetic description for each of them is presented, while the achieved results running two case studies are reported in further devoted sections. 2.1 On-site dynamic testing The clear comprehension of structural behavior is a consequence of a deep investigation of the different aspects involved. However dynamic testing in operational condition, conducted recording only absolute accelerations at different significant points, can be very helpful [15]. Within the group, the data-recording is generally conducted using a multichannel acquisition system. Servo-accelerometers (SA107LNColumbia) have been used in previous experiences [15,16]. The on-site experiences have been recently completed by a comparative studies conducted on real experimental data on the most popular output-only identification procedures for modal model and their use to identify finite element parametrical model [17]. On this basis, the identification of modal parameters from ambient vibration data is currently carried out using two main procedures: Enhanced Frequency Domain Decomposition (EFDD) and Stochastic Subspace Identification (SSI) The Enhanced frequency domain decomposition is a stochastic technique, operating in the frequency domain, based on the evaluation of the spectral matrix, collecting the frequency-depending power cross-spectral densities of the experimental structure response at different measurement points. The key point of the method is the assumption that, at a certain frequency, only a few significant modes (typically one or two) contribute to determine the spectral matrix. Instead, the data driven Stochastic Subspace Identification method, representing a time domain technique, allows the modal identification of a structure through the eigenproperties

The assessment of a representative physical model differs from modal identification in a few conceptual and procedural aspects. Modal models consist of global information, and a few frequencies and mode shapes are expected to capture the dominant structural behaviour. In contradistinction, physical models include local information, such as the stiffness and mass spatial distribution, which in principle should be wholly reconstructed. The simplifying hypotheses introduced in the modellization phase fix the model dimension, and rigidly determine the inherent structure of the stiffness and mass matrices. Such matrices can be initially evaluated according to nominal, or even estimated values of the mechanical parameters. Forcing the reference model to match the experimental frequencies and modes, the identification process reduces to the calibration, or updating, of the initial parameter values, while the model dimension and the structure of the governing matrices remain unchanged. Depending on the number, quality, and nature of the available information from the modal identification, different approaches to the physical model updating can be pursued [17]. Generally, the finite element models are used as a reference, taking advantage of the higher flexibility and computational efficiency of the numerical environment to explore different updating schemes [15], corresponding to different sets of free parameters. The data-to-unknowns redundancy is fully exploited, recurring to iterative techniques to minimize purposely defined objective functions, expressing the error of the updated model in emulating the experimental modal data. 2.3 Definition of SHM-WSN sensor features Vibration-based SHM requires sensed data that well represents the physical response of the structure both in amplitude and phase. The measurements must have sample resolution to characterize the structural response and must be recorded with a consistent sample rate that is synchronized with other sensed data from the structure. The sensor hardware needs for a sensor board with higher resolution and more accurate sampling rates designed specifically for SHM applications. The ST Microelectronics LIS344ALH capacitive-type MEMS accelerometer with DC to 1500 Hz measurement range, was chosen for the SHM-A board. This type of accelerometer utilizes the motion of a proof mass to change the distance between internal capacitive plates, resulting in a change of output voltage in response to acceleration. Though MEMS accelerometers are available with lower noise levels, the ST Micro accelerometer offers an excellent price/performance ratio. In addition, it provides three axes of acceleration on a single chip. The specifications for the accelerometer are given in Table 1. The SHM-A sensor board has been designed for monitoring civil infrastructure through

the Illinois SHM Project, an interdisciplinary collaborative effort by researchers in civil engineering and computer science at the University of Illinois at Urbana-Champaign [1]. Two hardware configurations of smart sensor nodes are required for the wireless communication and sensing: a gateway node for sending commands and receiving wireless data from network, and the battery powered nodes remote to the base station. To increase the communication range, both nodes are equipped with an antenna which covers the communication in a range of 30m and a SMA connector to install an external additional antenna. In the CERFIS configuration the boards are protected by a watertight partialgauzy box, allowing an in-the-distance visibility of light sensor to check the efficiency of the remote node. Each node is powered by an external cable connecting both the 220V electric web and an energy store box, composed by three rechargeable 1.5V battery IND alkaline D size with capacity of 20500mAh each, to assure a continuous registration procedure during earthquake events. The sensor location, inside historical monuments, does not allow an autonomous powered, as trough the well-known solar panels. An additional USB receptacle is installed to allow the link with a PC. The wireless communication is entrusted to an ADC converter.
Table 1. Accelerometer specification.

b)
Wireless sensor

A a) c)
Columbia, SA 107-LN

Figure 1. Experimental apparatus for wireless sensors characterization a) modular structural frame (1:3 scaling, 3m total height). b) the wireless sensor, c) wireless and wired sensors.

Parameter Axes Measurement range Resolution Power supply Noise density, x-and y-axes Noise density, z-axis Temperature range Supply current 2.4

Value 3 2g 0.66 V/g 2.4 V to 3.6 V 22 28 g/Hz 30 60 g/Hz -40 to 85C 0.85 mA

The investigation in the lab environment will be conducted on new sensor configurations fully developed by the CERFIS group. As is well known, one of the major limitations of wireless motes are the limited performances. Therefore, the idea is to use configurable hardware devices (e.g. FPGA) for the creation of hw/sw mixed service based architecture, with processing services directly implemented in hardware. In practice, we want to combine the mote processor with a set of ad-hoc developed co-processors specifically designed for the implementation of various processing modules. We think that this strategy will significantly increase monitoring efficiency, not only allowing a real-time processing, but also enabling the simultaneous support of different analysis techniques addressed to a wide range of application scenarios, from the pure structural health monitoring up to the emergency management, which imply often divergent specific requirements.
x 10
-4

Laboratory dynamic testing and wireless sensor characterization

0.04
[g]

Preliminary tests are conducted using a modular structural steel frame located at the CERFIS laboratory of University LAquila to characterize a SHM-WSN. In particular two different types of test have been performed. In the first series a direct comparison one single wireless sensor (the above described IMOTE 2 type) and one wired accelerometer (SA107LN-Columbia) has been conducted (see Figure 1). Within this configuration the frame responses both to a little impulse in longitudinal direction and under environmental noise have been recorded. Figure 2 displays the recorded acceleration with the two sensors in the impulsive test and the related Fast Fourier Transform of the signals. Others tests have been made using six wireless sensors, two for each slab, placed at diagonally opposite corners. This particular experimental setup has been used to identify the main modal frequencies, shapes and damping. Again both impulsive and ambient tests have been performed. The results are here not reported for sake of brevity. Moreover, in all tests, the wireless sensors, installed in the prototype structure, transfer the collected data to a single wireless node (gateway mode) linked to the acquisition card.

4 3 2 1

0.02 0 -0.02

a)
-0.04 0 0.04
[g]

b)
0 0 4 3 2 1 x 10
-4

20

[sec]

40

60

10

[Hz]

15

20

25

0.02 0 -0.02 -0.04 0

c)
20
[sec]

40

60

0 0

d)
5 10 [Hz] 15 20 25

Figure 2. Direct comparisons of accelerations recorded by two systems through time histories and FFT: a) b) wireless sensor, c) d) wired sensor.

OP AMP TI OPA4344

Accelerometer ST Microelectronics LIS344ALH

Light Sensor TAOS 2561

a)
External Input Connector Humidity & Temp. Sensor SHT11

b)

c)

Figure 3. Sensors for the monitoring of Basilica di Collemaggio: a) ISM400 (Imote 2 structural health monitoring board); b)wireless sensor assembled by CERFIS; c) sketch of the network.

2.5

SHM-WSN deployment on strategic and historical structures

Traditionally, a grid of sensor was deployed across a building and the measured data were conveyed via a cable connection to a central processing system (e.g. a personal computer). Recently, Wireless Sensor Networks (WSN) emerged as a possible attractive alternative solution, mainly due to the lower cost, lower size of the systems and ease of set up respect traditional wired systems thanks to the multi-hop connection capabilities which allow the nodes to organize themselves in a network where each node can be source, destination and also a router for the information flowing within the network. Current wireless monitoring systems are usually based on off the shelf sensor nodes equipped with new generation low cost, small sensors (e.g. MEMS accelerometers). Although these systems are not specifically designed for structural monitoring applications, they can still provide good performances. For example, Illinois Structural Health Monitoring Project (ISHMP) has shown the potential of WSN in several real monitoring scenarios [1]; they used a network of Imote2 motes equipped with a specifically design sensor board (ISM400) and an embedded processing software (ISHMP Toolsuite) based on TinyOS. Data processing is a key point in the future development of wireless monitoring systems. Many wireless implementations adopt a traditional processing paradigm, with data transmitted from the sensor nodes to a central gateway connected to a PC that performs the entire processing. However, modern sensor nodes are equipped with a microprocessor, allowing them to carry out local processing of data. In other words, data processing can be distributed across the network. The wireless systems, in fact, have progressed very rapidly in recent years and are now considered the enabling technology for realizing the pervasive ubiquitous computing environment that should support advanced distributed applications in many domains, especially for advanced distributed applications.

Therefore, owing to unprecedented design challenges and potentially large revenues, wireless sensor networks are calling huge interest in both the scientific and the industrial world. Besides a secure optimization of transmission (as shown by ISHMP work, whose software is already partially decentralized), processing de-centralization can bring the advantage of being able to quickly detect local phenomena, even in case of network splitting as a consequence of critical phenomena as an earthquake. This capability can be extremely useful in security systems or, generally, in the field of emergency management. In our view we will try to rethink structural modal analysis techniques, towards the goal of a distributed processing within the network which could efficiently support real-time monitoring and safety oriented services. Firstly, moving from the achievements and contributions of ISHMP, an iMote2based monitoring system was developed (see Figure 3). Moreover, the ISHMP software tools will be integrated with ad-hoc applications, in order to achieve an efficient distributed processing within our network. Moreover, optimizations of limited energy resources may be achieved through suited techniques of data compression and aggregation, providing reduced energy costs of communications and lower channel capacity for data delivery. The choice of the ISHMP software tools is not simply determined by the convenience of having a ready-to-use, decentralized-oriented middleware, but has a deeper reason. In fact, given the particular characteristics of the processing, the ISHMP Toolsuite was designed as a service-based software architecture. In other words, the various processing steps are implemented as services, and each application is just a collection of independent modules. 3 S. MARIA DI COLLEMAGGIO

The Basilica S.Maria di Collemaggio is one of the most attractive churches in Centre Italy. It dates from the XV century. The Basilica has a nave and two side aisles. The dimension of the nave is 61m in length and 11.3m in width; its height reaches 18.25m. The two side aisles are 7.8 and 8m in width; two external walls both 12.5m high delimit them. Seven columns, not evenly distanced, on each side separate the nave and two side aisles. The columns are about 5.25m high; a layer of well-laid stone, made of a calcareous material arranged irregularly in a poor quality mortar, encloses their core; the transverse section, approximately circular, is on average 1.00 m in diameter. The thickness of masonry varies from 0.95 m to 1.05 in the external walls; it is 0.9m in the two walls of the nave, over the columns. The four walls are connected on one side to the facade of the Basilica and, on the other side, to the transept. The facade is joined to a thick octagonal tower on the right corner; another masonry building is adjacent to a part of the wall, about 40% of it, behind the tower. The wooden roof is supported from trusses placed in a cross-sectional direction to the walls. Before the occurring of the 6th April 2009 earthquake A numerical and experimental study has permitted to characterize the dynamic behavior of the Basilica [13,14]. A parametric analysis was performed by FE models to predict and frame the response of the church.

a)

Figure 5. Temporary retrofitting interventions at Basilica of S. Maria di Collemaggio.

b)

c)

Figure 4. Finite element model of S. Maria di Collemaggio: a) top view, b) three-dimensional view, c) main facade.

Preliminary numerical analyses were carried out on the basis of several assumptions regarding: (1) mechanical parameters of masonry, (2) timber trusses of the roof, (3) restraints in walls and columns, (4) links among structural components. They permitted a reasonable prediction of the behaviour of the church to be made. Afterwards the Basilica was excited at a low level by an instrumented hammer and a mechanical vibration exciter (vibrodyne). Several tests have been carried out, with different positions of the instruments and impact locations, in order to excite and to measure as many modes as possible. The vibrodyne was located on the top of a lateral wall. The frequency responses were directly measured around the first two modes; these are the most important ones that describe the dynamic response of the church. Experimental data have been used to identify natural frequencies, modal displacements and damping factors. The first campaign of tests [13] have permitted to recognise at least four major resonance peaks in the range 0.83.0 Hz. The first two peaks are around frequencies values, about 1.25 and 1.7 Hz. Other peaks are present over 2 Hz. Two of them, around 2.5 and 2.7 Hz, are well defined in all tests. Secondary peaks, around 2.2, 2.3 and 2.6 Hz, are not always visible in all the responses; they indicate the occurrence of highly coupled modes. These peaks, however, are estimated to be less important: numerical analysis indicates that the participating mass of first two modes is at least 85% of total mass in the transverse direction of the church. After retrofitting, all peaks are shifted to higher frequencies. The first two are around 1.45 Hz and 2.12 Hz respectively. Other peaks are clearly visible around 2.6 and 2.95 Hz. Secondary peaks, which are not always visible in all the responses, are recognisable even in this case. Higher frequencies are a consequence of the increasing stiffness brought about by retrofitting. It is interesting to observe that now the responses of a pair of accelerometers are basically

identical, at least in the range of frequencies examined. This is a clear indication that the retrofitting had improved the link between the longitudinal walls. Other dynamic testing have been performed on the faade [14] which have permitted to evidence that out-of-plane local modes of this element are in a frequency range higher than the transversal mode of the nave. Recently, after the earthquake a strong effort has been made to use all the available data from the previous on-site dynamic campaign in order to develop a series of complete finite element models of the Basilica able to reproduce the main modal identified characteristics (see Figure 4). Starting from these models, a reproduction of the scenario after the collapse has been pursued. A campaign of numerical simulations is in progress to evaluate the dynamic response of the Basilica together with the temporary retrofitting (see Figure 5) under small earthquakes characterizing the numerous aftershocks at LAquila. The previous installation for the on-site dynamic testing campaigns together with the observation obtained by the modelling have driven the monitoring installation scheme reported in Figure 6.

Figure 6. Deployment of a smart wireless sensor network at Basilica S. Maria di Collemaggio.

THE ENGINEERING FACULTY BUILDINGS

The edifices of the Engineering Faculty have suffered particularly for seismic induced large structural displacements and accelerations which have brought them out of order due to the failure of non-structural elements (infills, false ceilings, door and window frames), the breakage of wiring and piping systems and the destruction of furniture and machineries. In particular, among the three recently-built buildings of the campus, erected in the early 90s, the so-called Edifice A presents the most critical damage scenario, which needs a significant rehabilitating intervention. The damage scenario is deeply described in [12], while the retrofitting procedure based on the dissipative coupling is presented at this conference in a companion paper [18]. Especially during the design stage of the retrofitting, the needs of a confident knowledge of the dynamic behavior of the structure has been significantly perceived. The following results synthesize the efforts made to reach such goal. Operational modal analysis was used to identify the spectral characteristics of the substructures A3 and A1 on the basis of standardized procedure, already used in previous studies [15,17] and thanks to the collaboration with a research group of Politecnico di Bari. The sensors, thirteen servoaccelerometers, were positioned on the structure and the facade of each substructure, to capture absolute accelerations at different significant points. The data-recording was conducted using a multi-channel acquisition system. The orientation of the uniaxial sensors was chosen to capture principally the out-of-plane motion of the planar frame sustaining the facade and the latero-torsional behaviour of the main three-dimensional structure. The most significant longduration registration of ambient vibrations in A3 substructure lasted for 2345s, during which the acceleration signals were recorded with a sampling frequency of 400Hz. The unique registration in A1 substructure lasted 2163s with a frequency of 400Hz. The identification of modal parameters from ambient vibration data has been carried out using the Enhanced Frequency Domain Decomposition (EFDD) and the Stochastic Subspace Identification (SSI). Figures 7a and 8a represents the singular values at each frequency. The curve peaks allow the identification of both well-separated and closely-spaced modes looking at the peaks of the singular value functions of frequency. The modal identification was carried out on the damaged substructures; differences with respect to the undamaged condition include the absence of some point connections between the facade and the main structure in the substructure A3, and a general reduction of the non-structural masses. The lower natural frequency of substructure A1 is equal to 3.90 Hz or 3.87 Hz, using EFDD or SSI technique, respectively. A roto-traslational shape characterizes the main mode, with prevalent translation in the y direction, and in-phase motion of the facade and the structure (Figure 7c). The modal properties until the sixth mode are reported in Table 2. The results of EFDD and SSI identification are characterized by similar frequencies, against more different damping values, especially in the more flexible modes. All the identified modes are characterized by a rototranslational shape prevalently in y direction. Substructure A3 is characterized by the lower natural frequency of 2.77 Hz or 2.76 Hz, using EFDD or SSI technique, respectively.

a)

b)

c)

Figure 7. Identification by EFDD of A1: a) singular values, b) experimental layout, c) modal shapes.

The first natural mode is characterized by a y-translational shape in the structure and a symmetrical roto-traslational shape in the two sides of the facade with a prevalent y direction. Where the interconnections are lacking, out-ofphase motion is possible. The first six modes, identified through EFDD and SSI techniques, are reported in Table 3. The second and third modes are characterized by rototraslational shape both in the structure and in the facade (Figure 8c). Due to the position of accelerometers, the different shape of facade are localized at the last floor. As evidenced in the substructure A1, the two different techniques, have furnished comparable frequencies values and more scattered modal damping values. Updating FE models for A1 and A3 substructures are currently under development.
Table 2. Identified spectral properties of substructure A1. (T=translational, R=rotational, L-F=local-facade).

Mode 1 2 3 4 5 6

Shape R-T R-T R-T R-T R-T R-T

Freq. [Hz] 3.90 4.47 6.16 8.86 12.67 15.84

EFDD Damping [%] 1.58 1.41 1.46 1.16 0.09 2.67

Freq. [Hz] 3.87 4.69 6.03 6.33 8.66 10.94

SSI Damping [%] 3.06 3.85 3.60 4.45 3.58 1.29

Table 3. Identified spectral properties of substructure A3 (T=translational, R=rotational, L-F=local-facade).

Mode 1 2 3 4 5 6

Shape T R-T R-T L-F L-F R

Freq. [Hz] 2.77 3.32 3.71 4.67 5.18 6.16

EFDD Damping [%] 2.03 2.28 1.57 4.15 8.04 2.29

Freq. [Hz] 2.76 3.34 3.71 5.19 5.30 6.21

SSI Damping [%] 3.43 4.02 3.61 3.42 1.22 3.40

b) a)

a) c)

Figure 10. Update finite element model of substructure A3: a) three-dimensional view, b) first mode, c) second mode.
b) c)

Figure 8. Identification by EFDD of A3: a) singular values, b) experimental layout, c) modal shapes.

Two different frame model are realized based on a deep investigation of various aspects, including global and local information on geometric characteristics and mechanical properties, completed by repeated visual inspections [12,16]. Both models are realized through frame elements, considering only the structural elements and diaphragm constraint at each floor of principal structure. The ground is simply modeled as rigid constraint. The load scenario during the dynamic measurements is nearly reproduced. A general overall agreement between the modal properties of the finite element models and the identified modal characteristics has been found in substructure A1. In Figure 9 are reported the modal shapes of the first and second modes. The shapes reproduce coherently the identified modes (Figure 7). A comparison between the identified frequencies and the ones predicted by FEM modeling is reported in Table 4, demonstrating a better correspondence with the SSI identified modes.
Table 4. Substructure A1: comparison between FEM and identified frequencies (ij=|(Freq.FEMj- Freq.ij)/ Freq.FEMj|).

The modal updating for substructure A3 represents a more complex operation. The modeling of A3 (Figure 10) does not allow to reproduce correctly the identified dynamic properties. A better approximation is founded taking into account the global behavior of both substructures A3 and A4, which are joined through a slab supported by a Gerber-type suspensions, as reported in Figure 11. The improvement of FE model is demonstrated in terms of frequencies, comparing them with the EFDD and SSI identified modes. Table 5 reports this comparison, highlighted the significant discrepancy between the identified and the FE model of A3 frequencies. The first and second mode shapes (Figure 11b,c) are substantially similar to those identified (Figure 8c). Indeed, in the second identified mode, is clearly demonstrable the stiffening effect on right side due to the supported slabs joining the two substructures. A mode-based comparison is developed through the modal assurance criterion (MAC) procedure, on the A3 data. Both the EFDD and SSI identified shapes are compared with the FE model A3-A4. Unlike the frequencies comparison (Table 5), the EFDD technique furnishes a better result on the principal diagonal and out of diagonal values.

Mode 1 2 3 4 5 6

Freq. [Hz] 3.86 5.51 6.66 7.37 16.13 18.88

EFDD [%] 1.0 18.9 7.5 21.2 20.8 42.3

SSI [%] 0.0 18.5 5.1 23.4 18.8 44.4

a)

b) b)

a) c)

c)

Figure 9. Updated finite element model of substructure A1: a) three-dimensional view, b) first mode, c) second mode.

Figure 11. Update finite element model of substructure A3-A4: a) three-dimensional view, b) first mode, c) second mode.

Table 5. Substructure A3: Comparison between A3 FEM, A3-A4 FEM and identified frequencies (ij=|(Freq.FEMj- Freq.ij)/ Freq.FEMj|).

Mode 1 2 3 4 5 6

Freq. [Hz] 2.40 2.64 3.25 3.78 3.80 5.71

FEM A3 EFDD [%] 15.4 25.8 14.1 23.5 36.3 7.9

SSI [%] 15.0 26.5 14.1 37.3 39.5 8.8

FEM A3 A4 Freq. EFDD SSI [Hz] [%] [%] 2.95 6.1 6.4 3.19 4.1 4.7 3.59 3.3 3.3 5.26 11.2 1.3 5.27 1.71 0.6 5.30 16.2 17.2

information can be used to explore the potentiality either for new seismic protection systems using innovative dissipative schemes or for the design of permanent monitoring system based on WSN. A specific path towards the installation of a SHM-WSN at the Basilica S. Maria di Collemaggio has been illustrated. The data of the monitoring systems will be used during the reconstruction phases to evaluate the efficiency of the proposed structural solutions. The multidisciplinary experience of a group of young researchers at University of LAquila in these challenging fields has been strongly synthesized in the paper and hopefully will be object of future developments. REFERENCES
[1] J. Rice, B. Spencer Jr., Flexible smart sensor framework for autonomous full-scale structural health monitoring, NSEL Report Series 018, 2009. F. Casciati, An overview of structural health monitoring expertise within the European Union, Struct. Health Monit. & Intell. Infrastruct. Lisse, 2003. D. Pines, A. Aktan, Status of structural health monitoring of long-span bridges in the United States, Progr. in Struct. Eng. & Mat. 4(4):372, 2002. A. Mufti, Structural health monitoring of innovative Canadian civil engineering structures, Struct. Health Monit. 1(1):89, 2002. Z. Wu, Structural health monitoring and intelligent infrastructures in Japan, Struct. Health Monit. & Intell. Infrastruct. Lisse, 2003. J. Ou, The state-of-the-art and application of intelligent health monitoring systems for civil infrastructures in mainland of China, Progr. in Struct. Eng., Mech.& Comp., London, 2004. H. Van der Auweraer, B. Peeters, International Research Projects on Structural Health Monitoring: An Overview, Struct. Health Monit. 2(4):341, 2003. A. Del Grosso, F. Lanata, A. Torre, Recent Structural Health Monitoring applications in Italy, Eur. Struct. Health Monit., Granada (Spain), 2006. T. Miller, B. Spencer Jr, j. Li, H. Jo, Solar energy harvesting and software enhancements for autonomous wireless smart sensor networks, NSEL Report Series 022, 2010. C. Fritzen, P. Kraemer, Self-diagnosis of smart structures based on dynamical properties, Mech, Syst. & Signal Processing 23(6):1830, 2009. D. Zonta, H. Wu, M. Pozzi, P. Zanon, M. Ceriotti, L. Mottola, G. P. Picco, A. L. Murphy, S. Guna, M. Corr, Wireless sensor networks for permanent health monitoring of historic buildings, Smart Structures and Systems 6(5-6):595, 2010. A.M. Ceci, A. Contento, L. Fanale, D. Galeota, V. Gattulli, M. Lepidi, F. Potenza, Structural performance of the historic and modern buildings of the University of LAquila during the seismic events of April 2009, Engineering Structures, 32(7), 1899-1924, 2010. E. Antonacci, G.C. Beolchini, F. Di Fabio, V. Gattulli, The dynamic behavior of the basilica S. Maria of Collemaggio, Proc. 2nd Intern. Congr. on Studies in ancient Structures, Instanbul, Turkey, 2001. E. Antonacci, G.C. Beolchini, The dynamic behaviour of the facade of the Basilica di S. Maria di Collemaggio. Proc. of Structural analysis of historical constructions, IV International Seminar, Padova, Italy, 2004. M. Lepidi, V. Gattulli, D. Foti, Bell-swinging resonance and their cancellation identified by dynamical testing in a modern bell tower. Engineering Structures, 31(7), 1486-1500, 2009. A. Ceci, L. Fanale, D. Galeota, V. Gattulli, M. Lepidi, F. Potenza, Seismic retrofitting of recently-built Engineering Faculty edifices at LAquila, Proc. Conference on Sustainable Development Strategies for Constructions in Europe and China, Roma, Italy, 19-20 April, 2010. E. Antonacci, A. De Stefano, V. Gattulli, M. Lepidi, E. Matta, Comparative study of vibration-based parametric identification techniques for a three-dimensional frame structure, J. of Struct. Control & Health monitoring, in press. V. Gattulli, M. Lepidi, F. Potenza, A.M. Ceci, Nonlinear viscous dampers interconnecting adjacent structures for seismic retrofitting, EURODYN 2011, Leuven, Belgium, 4-6 July, 2011.

Table 6. Substructure A3: mode-based comparison among the identified modes by EFDD and the updated FEM A3-A4.

0.98 0.57 0.63 0.09 0.17 0.54

0.23 0.17 0.07 0.28 0.29 0.00

0.81 0.69 0.57 0.24 0.03 0.43

0.12 0.01 0.02 0.94 0.11 0.42

0.07 0.13 0.18 0.37 0.98 0.17

0.10 0.20 0.06 0.44 0.19 0.05

[3] [2]

[4] [5] [6]

A flipping of fourth and fifth modes in FE model was evidenced in the MAC with SSI results. The out of diagonal elements positioning at (4,5) and (5,4) components, results extremely close to the unit, while the diagonal element (4,4) and (5,5) being complementary and closer to zero. Being the frequencies in the FEM model very similar, an inversion of order could be possible. In Table 7, the MAC with SSI results, reversing the fourth with the fifth mode in FEM A3-A4, is reported. However, the collected data are still under a deeper investigation in order to extract information on the effects of the damaged parts (not all the horizontal steel tubes were present in the A3 substructure) on the measured dynamic response.
Table 7. Substructure A3: Mode-based comparison among the identified modes by SSI and the updated FEM A3-A4.

[7]

[8]

[9]

[10]

[11]

[12]

0.98 0.55 0.62 0.10 0.16 0.55 5

0.04 0.07 0.01 0.29 0.46 0.03

0.84 0.66 0.56 0.26 0.03 0.47

0.18 0.03 0.05 0.91 0.07 0.50

0.24 0.26 0.33 0.09 0.93 0.50

0.07 0.15 0.04 0.48 0.24 0.04

[15] [13]

[14]

[16]

CONCLUSIONS
[17]

The available procedures of operational modal identification procedures, based on experimental dynamic testing, may be substantially helpful for a full comprehension of the dynamic behavior of strongly damaged structures after the occurring of a near-fault earthquake such as the one which struck the city of LAquila on 6 April 2009. In particular the acquired

[18]