**5. Structural health monitoring research activities at university of L'Aquila**

A group of researchers of CERFIS (www.cerfis.it) 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 L'Aquila. 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 [5,9,11,30]; (*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 program‐ ming of modal and damage identification procedures taking into account temperature variation effects.

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

Advanced Applications in the Field of Structural Control and Health Monitoring After the 2009 L'Aquila Earthquake

http://dx.doi.org/10.5772/55438

225

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-tounknowns redundancy is fully exploited, recurring to iterative techniques to minimize purposely-defined objective functions, expressing the error of the updated model in emulating

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

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 5. The SHM-A sensor board has been designed for monitoring civil infrastructure through the Illinois SHM Project, an interdisci‐ plinary collaborative effort by researchers in civil engineering and computer science at the

Two hardware configurations of smart sensor nodes are required for the wireless communi‐ cation 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 commu‐ nication 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 a watertight partial-gauzy box, allowing an in-the-distance visibility of light sensor to check the efficiency of the remote node, protects the boards. An external cable connecting both the 220V electric web and an energy store box, composed by three recharge‐ able 1.5V batteries IND alkaline D size with capacity of 20500mAh each, to assure a continuous registration procedure during earthquake events, powers each node. The sensor location, inside historical monuments, does not allow an autonomous powered, as trough the well-

unchanged.

the experimental modal data.

for SHM applications.

**5.3. Definition of SHM-WSN sensor features**

University of Illinois at Urbana-Champaign [31].

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 for the structural health monitoring of the Basilica di Collemaggio are finally reported.

#### **5.1. 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 [30]. Within the group, the data-recording is generally conducted using a multi-channel acquisition system. Servo-accelerometers (SA107LN-Columbia) have been used in previous experiences [5,30]. The on-site experiences have been recently completed by a comparative studies conducted on real experimental data on the most popular output-only identification proce‐ dures for modal model and their use to identify finite element parametrical model [11]. 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 frequencydepending 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 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.

### **5.2. Finite element modeling and updating**

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-tounknowns 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.

### **5.3. Definition of SHM-WSN sensor features**

procedures and wireless monitoring sensors; (*v*) deployment of structural health monitoring systems with wireless smart sensors; (*vi*) development and installation by remote program‐ ming of modal and damage identification procedures taking into account temperature

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 for the structural health

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 [30]. Within the group, the data-recording is generally conducted using a multi-channel acquisition system. Servo-accelerometers (SA107LN-Columbia) have been used in previous experiences [5,30]. The on-site experiences have been recently completed by a comparative studies conducted on real experimental data on the most popular output-only identification proce‐ dures for modal model and their use to identify finite element parametrical model [11]. 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)

The Enhanced frequency domain decomposition is a stochastic technique, operating in the frequency domain, based on the evaluation of the spectral matrix, collecting the frequencydepending 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 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

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

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

approximates the experimental response, is a matter of identification too.

spatial distribution, which in principle should be wholly reconstructed.

monitoring of the Basilica di Collemaggio are finally reported.

224 Engineering Seismology, Geotechnical and Structural Earthquake Engineering

variation effects.

**5.1. On-site dynamic testing**

and Stochastic Subspace Identification (SSI)

**5.2. Finite element modeling and updating**

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 5. The SHM-A sensor board has been designed for monitoring civil infrastructure through the Illinois SHM Project, an interdisci‐ plinary collaborative effort by researchers in civil engineering and computer science at the University of Illinois at Urbana-Champaign [31].

Two hardware configurations of smart sensor nodes are required for the wireless communi‐ cation 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 commu‐ nication 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 a watertight partial-gauzy box, allowing an in-the-distance visibility of light sensor to check the efficiency of the remote node, protects the boards. An external cable connecting both the 220V electric web and an energy store box, composed by three recharge‐ able 1.5V batteries IND alkaline D size with capacity of 20500mAh each, to assure a continuous registration procedure during earthquake events, powers each node. The sensor location, inside historical monuments, does not allow an autonomous powered, as trough the wellknown 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 5.** Accelerometer specifications.

#### **5.4. Laboratory dynamic testing and wireless sensor characterization**

Preliminary tests are conducted using a modular structural steel frame located at the CERFIS laboratory of University L'Aquila 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 (Figure 10). Within this configuration the frame responses both to a little impulse in longitudinal direction and under environmental noise have been recorded. 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 per‐ formed. 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.

**Figure 10.** Light model (scale 1:3) of modular steel-made three-dimensional frame: (a) basic configuration, (b) sensor-

Advanced Applications in the Field of Structural Control and Health Monitoring After the 2009 L'Aquila Earthquake

http://dx.doi.org/10.5772/55438

227

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 setup 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

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 [31]; they used a network of Imote2 motes equipped with a specifically design sensor board (ISM400) and an

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

node of wireless network; (c) comparison with sensor of traditional wired network

**5.5. SHM-WSN deployment on strategic and historical structures**

also a router for the information flowing within the network.

embedded processing software (ISHMP Toolsuite) based on TinyOS.

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.

Advanced Applications in the Field of Structural Control and Health Monitoring After the 2009 L'Aquila Earthquake http://dx.doi.org/10.5772/55438 227

**Figure 10.** Light model (scale 1:3) of modular steel-made three-dimensional frame: (a) basic configuration, (b) sensornode of wireless network; (c) comparison with sensor of traditional wired network

### **5.5. SHM-WSN deployment on strategic and historical structures**

known solar panels. An additional USB receptacle is installed to allow the link with a PC. The

Preliminary tests are conducted using a modular structural steel frame located at the CERFIS laboratory of University L'Aquila 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 (Figure 10). Within this configuration the frame responses both to a little impulse in longitudinal direction and under environmental noise have been recorded. 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 per‐ formed. 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

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

**Parameter Value** Axes 3 Measurement range ±2g

Resolution 0.66 V/g Power supply 2.4 V to 3.6 V Noise density, x-and y-axes 22 – 28 μg/Hz Noise density, z-axis 30 – 60 μg/Hz Temperature range -40 to 85°C Supply current 0.85 mA

wireless communication is entrusted to an ADC converter.

226 Engineering Seismology, Geotechnical and Structural Earthquake Engineering

**5.4. Laboratory dynamic testing and wireless sensor characterization**

node (gateway mode) linked to the acquisition card.

often divergent specific requirements.

**Table 5.** Accelerometer specifications.

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 setup 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 [31]; 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.

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

Advanced Applications in the Field of Structural Control and Health Monitoring After the 2009 L'Aquila Earthquake

http://dx.doi.org/10.5772/55438

229

Before the occurring of the 6th April 2009 earthquake a numerical and experimental study has permitted to characterize the dynamic behavior of the Basilica [32-34]. The experimental data were firstly used to identify a modal model and then to determine suitable FE models able to predict and frame the dynamical response of the church. 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. 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

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

**Figure 11.** Drawings for the locations of the 16 smart sensors mounting tri-axial MEMS accelerometers, humidity and

temperature measuring instruments, installed at the Basilica di S. Maria di Collemaggio, at L'Aquila, Italy.

placed in a cross-sectional direction to the walls.

measure as many modes as possible.

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 envi‐ ronment 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 insecurity systems or, generally, in the field of emergency management.

A series of activities are still under development to rethink structural modal analysis techni‐ ques, towards the goal of a distributed processing within the network, which could efficiently support real-time monitoring and safety oriented services [10]. Firstly, moving from the achievements and contributions of ISHMP, an iMote2-based monitoring system was devel‐ oped. 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 servicebased software architecture. In other words, the various processing steps are implemented as services, and each application is just a collection of independent modules.
