PSI - Issue 64

Filippo Andreose et al. / Procedia Structural Integrity 64 (2024) 40–47

42

3

sensor network, and monitoring records of certain time duration are transferred and stored in a server or in the cloud, where data are automatically elaborated by P3P (Garcìa-Macìas et al. (2023)).

Fig. 1. Schematic of a permanent SHM system installed in a bridge structure using P3P (Garcìa-Macìas et al. 2023).

Acceleration signals are introduced in a sequential process involving signal processing of acceleration signals, system identification through automated OMA, and frequency tracking. At this stage, the user can identify subgroups of accelerometers corresponding to separate spans in multi-span bridges, which allows to conduct span-wise OMA and frequency tracking. In parallel, probabilistic-based features (mean and rms) are extracted from both acceleration and non-acceleration data. This process constitutes the feature extraction phase in P3P. Finally, once the user selects the set of desired damage-sensitive features, P3P conducts data normalization and anomaly detection. The outcome of this process is the generation and automated updating of multiple control charts assessing the desired sets of damage-sensitive features. A stand-alone version of P3P has been developed in C++ with a compact GUI. From this interface, the user can access 10 main modules, namely: • Project configuration: local or FTP directories where acceleration data are stored, together with file configuration properties, log folder, naming protocols, and outputs settings. • Modal geometry: geometry text files including nodes, lines, colour planes, acceleration channels and kinetic equations among nodes. The code also allows to identify sets of channels belonging to different spans in multi span bridges as well as setting certain channels dedicated to seismic analysis. • Signal pre-processing: visualization and pre-processing of the acceleration time signals via a library of filters to minimize the effects of noise and the presence of abnormal events. • System identification - OMA: manual or automated Frequency Domain Decomposition (FDD) for fast assessment and verification, with results presented in the shape of tabulated data, stabilization diagrams, exportable reports, and the histogram representation of the MAC matrix. • Static/environmental data: data derived from environmental and static sensors are here treated by providing a reference tag, measurement unit, naming protocol, and the directory that contains the record files. • Modal tracking: frequency tracking from collected data, via the specification of tolerances on relative frequency variations and minimum MAC values. Here the time series of resonant frequencies, damping ratios, and mode shapes of the structure, as well as time series of MPC and MAC values of the mode shapes can be derived with no limitations. • Statistical pattern recognition: statistical models are here introduced to detect potential damage conditions through autoregressive modelling, estimators/predictors definition, and clustering analysis to create cluster-wise non-linear SPR models. • Continuous SHM: this module provides the control charts that can be used to to manage permanent SHM systems in real-time and in a completely autonomous way by plotting real-time graphs of the time series of the estimators and acceleration data, as well as the corresponding control charts.