PSI - Issue 62

E. Tomassini et al. / Procedia Structural Integrity 62 (2024) 903–910 Author name / Structural Integrity Procedia 00 (2019) 000–000 5 matrix of the measurements, ( ) , commonly computed using the Welch’s method. Basic theory of linear time invariant dynamical systems yields the following property for ( ) : ( ) = ( ) ( ) ( ) � , (8) where: ( ) and ( ) � are the Frequency Response Function matrix and the corresponding Hermitian transpose, and ( ) is the PSD matrix of the input. Assuming the input to be a white noise spatially uncorrelated random process (matrix is diagonal), Eq. (8) resembles a SVD and suggests that natural frequencies of the structure can be identified by mean of a pick picking operation on the first Singular Value (SV) plot of ( ) and mode shapes as their corresponding singular vector. Following the definition of subgroups based on the modal coupling characteristics of the structure, the reference SSI-cov modal identification needs to be conducted for each subgroup. Subsequently, the parameters from the reference identification are employed for automated identification during the training period. A statistical model is then established for each subgroup, taking into account the residuals between the predictions made by each statistical model and the tracked modal features. Finally, a control chart, capable of monitoring the health state of the bridge, is computed for the training period and continuously updated whenever a new acceleration time history is acquired. 3. Case study: San Faustino Bridge The San Faustino bridge is a nine-span post-tensioned concrete box-girder bridge located in the municipality of Perugia, Italy. With a total length of 365.4 m, the first and ninth spans are 33.6 m each, while the internal spans measure 42.6 m (Fig. 1). As one of a pair of parallel twin bridges serving the same roadway, it is oriented towards Rome, with its twin directed towards Florence. The bridge features 10 half-joints, one for each span, except for the third span, which has two. 907

Fig. 1. (a) Longitudinal geometry; (b) cross section of the deck of the San Faustino bridge; (c) views of the structure; (d) views of the half-joints. The comprehensive monitoring system is composed by 66 MEMS accelerometers ( ± 2 g, 24-bit ADC, noise density 22.5 μ g/ √ Hz) strategically placed throughout the structure, resulting in 114 measurement channels. Each span is equipped by 2 biaxial accelerometers in the y- (transversal) and z-directions (vertical), along with 2 additional uniaxial sensors in the z-direction (vertical) and the piles are equipped with a biaxial accelerometer on the top, with the first and last piles featuring a triaxial accelerometer at the base. Furthermore, the half joints are outfitted with a triaxial accelerometer and a uniaxial accelerometer in the z-direction. The monitoring system comprises a wired connection to the onboard computer, continuously storing 12 acceleration files per day, each with a two-hour duration and a sampling acquisition frequency of 125 Hz. The recorded acceleration time series from April 25th, 2023, at 2 p.m., are used in this paper for illustrative purposes, undergoing a simple filtering sequence involving the removal of linear trends, a low-pass filter with a cut-off frequency of 20 Hz, and decimation to a sampling frequency of 40 Hz. In this study, the filtered acceleration data is utilized for automated SSI-cov modal identification.