PSI - Issue 62

E. Tomassini et al. / Procedia Structural Integrity 62 (2024) 903–910 Author name / Structural Integrity Procedia 00 (2019) 000–000

904

2

1. Introduction In recent years, road infrastructural engineering has prioritized enhancing bridge network safety. Moved from the recent bridge collapses, the global governments are increasing investments and instituting monitoring regulations. The United States allocated funds through the Department of Transportation Appropriations Act (May 5, 2022), informed by the Report Card of America's Infrastructure (2021). In the European context, the Italian government enacted the Guidelines on Risk Classification and Management (2022). Evolving regulations underscore the crucial role of Structural Health Monitoring (SHM), with Operational Modal Analysis (OMA) deserving a special attention as a powerful non-destructive and low intrusive vibration-based monitoring technique. Particularly, OMA allows the extraction of modal characteristics (i.e. natural frequencies, damping ratios and modeshapes) from ambient accelerations recorded on structures. Such modal characteristics, in turn, can be used as damage sensitive features within monitoring strategies based on statistical pattern recognition approaches. In the past decade, much research has focused on automation of OMA algorithms, enabling the structural monitoring in real-time. Nevertheless, since the natural frequencies are particularly sensitive to temperature fluctuations, the removal of environmental effects from the tracked modal features is of pivotal importance. In that field, many authors presented different techniques able to remove the environmental effects through the creation of statistical models, able to reproduce the fluctuations due to the temperature and humidity effects. Peeters and De Roeck (2001) identified damages through residuals between OMA-derived resonant frequencies and the predictions of an autoregressive model. With a similar aim, Yan et al. (2005) used Principal Component Analysis (PCA) and Magalhães et al. (2012) employed Multi Linear Regression (MLR) models. Therefore, variations in the statistical pattern of the residuals between the tracked frequencies (or other damage-sensitive features) and the statistical model predictions are able to detect damages. Kullaa (2003) proposed the use of control charts, as a suitable way to highlight damages in a data-driven perspective. Indeed, control charts represent efficient statistical control from econometrics, capable of automatically triggering alarms when sample statistics deviate from an in-control reference, making them ideal for real-time SHM (Montgomery, 2020). Increasing awareness of SHM has led the above presented advanced techniques to complementing traditional visual inspections by management authorities. Nevertheless, full technological integration requires consolidating monitoring procedures into robust comprehensive software. García-Macías et al. (2020, 2022) provided a contribution to this need by developing MOVA/MOSS and P3P software packages, making significant strides in real-world large-scale applications. Worldwide road infrastructure management authorities are currently equipping numerous bridges with advanced sensor systems for highly efficient monitoring. These sensors, connected via wired connections to data acquisition systems, record huge amounts of data which are then stored on onboard remotely controlled computers and processed to extract significant features. This task is especially complicated given that many monitored bridges have large dimensions and their static configurations can be of many different types, such as long multi-span simply supported bridges, continuous girder bridges with half-joints, or other specialized designs like cable-stayed and suspension bridges. Therefore, the whole bridge dynamic identification of these instrumented structures involves managing high computational loads and necessitates high-performance hardware for making it compatible with continuous SHM schemes. For this reason, the modal identification of long multi-span bridges is often conducted by splitting the whole identification in many different subgroups of channels. Against this backdrop, this work proposes a critical methodology to define the subgroups of channels in the case of densely equipped bridges and to conduct their continuous SHM process. The methodology is applied to the real case study of a nine-span box girder highway bridge: the San Faustino bridge, located in the municipality of Perugia, Italy. All steps to set the parameters for running the damage detection process are critically discussed using the afore mentioned P3P software, starting from signal processing and extending to the development of control charts for damage detection purposes. 2. Theoretical background and methodology The setting of the Continuous SHM procedure can be discretized in four steps: (i) definition of a reference OMA identification, (ii) modal identification and frequency tracking of a training set of acceleration time histories, (iii) definition of the statistical models and (iv) running of the continuously updated control charts.