60 D. La Mazza et al. As such, the data collection, filtering, and pre-processing should be supported by a detailed and comprehensive analysis process of experts in assessing the structural state of health and conservation. FE numerical model updating procedures based on the monitoring data, detailed modal analysis, and reliability evaluation are needed to explore critical scenarios, identify threshold limits, and develop algorithms to perform long-term data interpretation, to efficiently support the asset management process. 7.4 Anomaly Detection and Data Analytics The bridge has been subject to monitoring since August 2019, and at present, it’s still under continuous control, with over 2 years of recorded data available, as a base to support the structural diagnostics and response interpretation. After the monitoring system installation, a diagnostic load test has been performed on the instrumented bridge to identify the static and dynamic structural response under given load conditions, with the aim to calibrate the preliminary finite element (FE) model, based on the available design information and the results of the periodic inspections. The measured responses have been compared with those derived from the theoretical model under the same loading condition and used as the basis for validating the FE model. The so-called model-updating procedure consists of an ad hoc optimization multi-step deterministic algorithm, developed by Sacertis, that aims to minimize an objective function expressing the residuals between numerical and experimental data. The updated FE model has been used to define thresholds for the key performance indicators (frequency shifts, STD, instantaneous or residual rotations, etc.) as a result of the simulation of relevant damage scenarios for the viaduct. Additional statistical thresholds have been defined, based on the initially recorded behavior, considered as the standard baseline, to perform a near-real-time anomaly detection at the gateway level. Anomalies arose early, just after the initial baseline behavior characterization and threshold settings, as the monitoring systems identified a sudden change in the natural frequencies and longitudinal mode shapes, as well as a threshold exceedance of the average vibrational levels (STD) of the sensor group near the north joint. The anomaly persisted, showing an unstable dynamic response of the bridge, until it reached a new steady condition from the end of November up to May, when a second process of anomaly detection was followed by the return to the baseline behavior already observed in spring. There was clear evidence of a seasonal dependence of the bridge behavior from the environmental conditions; thermal actions were considered as the main actor playing a role in determining the anomaly, causing a variable joint behavior. This assumption was confirmed during summertime, when a third static scheme became apparent, characterized by a significant frequency shift (increase) of the lateral in-plane modal shape. Prompt analysis of the acquired data as well as of the FE models was performed to understand and justify the measured structural response (Fig. 7.3). The first simplified analyses were carried out at the sensor level to offer the bridge management prompt and reliable surveys, on all the accelerometers belonging to the network, processing data in both the time and the frequency domains. In the frequency domain, operational modal analysis (OMA) algorithms were implemented, since under operational conditions the structure excitation, mainly given by traffic and ambient vibrations, has been reasonably assumed as white noise. The structure natural frequencies were derived analyzing the power spectral densities (PSDs) calculated on accelerations recorded by each sensor and each measurement axis, identifying the spectral peaks and later connecting information coming from the Fig. 7.3 Comparison of the three different bridge dynamic behaviors: green is autumn (this is the baseline), blue is winter, and red is summer; power spectral densities from one accelerometer are given of the left (each of them represents a stable and repeatable condition), and the first mode shapes are represented on the right
RkJQdWJsaXNoZXIy MTMzNzEzMQ==