PSI - Issue 78

Elisa Tomassini et al. / Procedia Structural Integrity 78 (2026) 1831–1838 Author name / Structural Integrity Procedia 00 (2025) 000–000

1833

3

• Geometry Definition: Allows users to upload a simplified geometric model of the structure, including node positions, lines, colored surfaces, sensor placements, channel orientations, and boundary conditions. This visual model is crucial for the mode shape visualization module in the next steps. • Quality Check and Measurement Chain: Facilitates the configuration of quality thresholds to evaluate the reliability of the measurement chain. It checks for missing data, synchronization errors, and connectivity issues, and flags any defective recordings or sensor malfunctions within the wired system. • Signal Preprocessing: O ff ers a variety of preprocessing tools to condition raw data. Available filters include detrending, bandpass filtering, and moving average removal. Users can view time histories and power spectral density (PSD) plots before and after filtering to verify data quality. • Modal Identification: Three OMA algorithms are o ff ered: i) Frequency Domain Decomposition (FDD), ii) Automated Frequency Domain Decomposition (AFDD) and iii) Automated Covariance-Driven Stochastic Sub space Identification (Cov-SSI). The FDD method provides a computationally e ffi cient modal identification, while Cov-SSI o ff ers higher precision and automation capability. The Cov-SSI strategy can be personalized by adjusting parameters such as time lag, stabilization diagram limits, and model order step. Stability is assessed by comparing frequencies, damping ratios, and mode shapes of poles between consecutive model orders. Further thresholds can be defined in order to check the consistency of the identified modes, such as maximum damping ratio and minimum Mode Phase Collinearity (MPC) factor and minimum cluster size for hierarchical clustering. • Frequency Tracking: Supports both static and dynamic tracking of modal parameters over a selected training period. Static tracking compares newly identified modes with fixed reference modes, while dynamic tracking updates references in response to significant structural changes. • Statistical Pattern Recognition: Permits the construction of statistical models to account of the variability due to environmental or operational e ff ects. Estimators such as resonant frequencies and Modal Assurance Criterion (MAC) values can be used. Available techniques include Principal Component Analysis (PCA) and Multiple Linear Regression (MLR).(see Ubertini et al. [2018], Yan et al. [2005]). • Continuous Structural Health Monitoring: Enables automated, real-time updating of control charts using the Hotelling’s T 2 (Kullaa [2003]) distance metric computed on residuals between predicted and measured fea tures. These charts support damage detection through threshold-based warning and alarm levels. Visualizations include predictor time series, estimated-vs-predicted modal frequencies, and control chart values over time. • Seismic Analysis: Complements the SHM system by enabling assessment of structural response during seismic events. It facilitates comparison between accelerations recorded at the abutment bases and pier tops to evaluate transmission e ff ects and pier drifts. When pier masses are known, the module also estimates lateral forces acting on vertical elements, supporting post-event evaluation. The Marmore Bridge represents a modern integration of structural engineering and long-term monitoring. Located along the SS79 near the Marmore Falls in Umbria, it was completed in 2010 to enhance connectivity between Lazio and Umbria, addressing the Valnerina region’s challenging topography and aging infrastructure. The structure fea tures a 170 m main steel arch spanning the Nera River, a secondary semi-arch embedded into the rock at the Valnerina tunnel entrance, and a total length of approximately 300 m, with the deck rising 70 m above the valley floor. Designed for both landscape integration and seismic resilience, the bridge was instrumented in 2024 under the national SHM program by ANAS S.p.A. The monitoring system includes 9 triaxial and 54 uniaxial Dewesoft ® MEMS accelerom eters, strategically deployed to capture global dynamic behavior. Specifically, 46 uniaxial vertical sensors are aligned along the deck edges, while each arch crown and pier top hosts a triaxial sensor on one side and a vertical uniaxial unit on the other (Fig. 1(a)). The accelerometers operate with ± 2 g range, 4 kHz sampling rate, 96 dB dynamic range, and 25 µ g / √ Hz noise density. A high-range triaxial sensor at the right abutment base serves as a seismic trigger. The sys tem comprises 81 accelerometric channels and 3 temperature sensors distributed across key deck sections. All devices are wired to an onboard edge computer. Static data (temperature, inclinometers) are logged every 15 minutes, while accelerometers record 30-minute time histories every two hours (12 per day) at 100 Hz. Preprocessing includes linear detrending, low-pass Chebyshev Type I filtering (8th-order, 10 Hz cuto ff ), and downsampling to 20 Hz for subsequent modal analysis. 3. Continuous SHM on a landmark arch bridge: the Marmore bridge