PSI - Issue 84

Roberto Acerbis et al. / Procedia Structural Integrity 84 (2026) 765–772

769

2.5.1.3. Definition of the Reference Configuration In monitoring existing structures, analyses must be conducted with respect to a well-defined structural reference configuration. This configuration may not coincide with the original condition of the structure, as monitoring systems are typically installed when the structure has already been in service, often for many years. Since the structural history prior to installation (residual deformations, progressive degradation, extraordinary maintenance interventions) is generally unknown in detail, the initial reference configuration should be regarded as an operational baseline. It is defined based on the first datasets that successfully pass validation procedures and serves as the basis for analyzing the temporal evolution of measured quantities. Zero-referencing data with respect to this configuration serves multiple purposes: ensuring temporal comparability of measurements, correcting potential installation-related offsets (e.g., imperfect sensor alignment), and focusing analyses on variations with actual structural significance. A single reference configuration is preferably adopted for all installed sensors, even when different sensor types are used, with specific procedures defined to address subsequent system modifications (repairs, replacements, or sensor additions). 2.5.1.4. Standardization of Sign Conventions A critical aspect of pre-processing concerns verification and correction of sensor axis orientation. Sign inconsistencies may result from installation errors or incorrect software configurations and must be resolved to avoid misinterpretation of structural behavior. For biaxial clinometers installed on the deck, providing information on flexural and torsional rotations, unified conventions for measurement directions consistent with the global structural reference system must be adopted. Similarly, for displacement transducers installed across joints or specific elements, the meaning of positive (extension or increased distance) and negative (contraction or reduced distance) values must be clearly defined. Detection of axis inconsistencies is performed by cross-referencing as-built documentation with quantitative correlation analyses. In particular, Pearson correlation coefficients are computed between individual sensor measurements and reference quantities, such as structural temperatures or measurements from similar sensors installed in analogous positions. Correlations are computed on moving averages to attenuate high-frequency noise. Unexpected or inverse correlations relative to theoretical expectations indicate the need for sign corrections. 2.5.1.5. Filtering Techniques and Discontinuity Management To improve signal readability and reduce instrumental noise, smoothing techniques based on moving averages with different time window lengths are applied. The optimal window size is selected to preserve structurally meaningful information while reducing variability due to instrumental errors or non-relevant signal fluctuations. Excessively large windows may suppress significant structural variations, while excessively small windows retain excessive noise. For temperature-sensitive quantities, which constitute the majority of static measurements, window calibration is performed by evaluating correlations between raw data and temperature values over different moving window lengths. The minimum window size beyond which further increases do not significantly improve correlation is selected, ensuring an optimal balance between noise reduction and information preservation. Special attention is given to the removal of discontinuities and spikes using specific algorithms that identify three main anomaly types: isolated spikes, prolonged constant values (plateaus) typically indicating sensor blockage, and rapid but persistent shifts in the mean signal level. Once identified using objective statistical criteria, these anomalies are corrected through interpolation or replacement with estimated values, ensuring signal continuity and regularity. The effectiveness of applied corrections is always verified through graphical comparison between original and corrected signals to ensure that no artifacts are introduced and that data integrity is preserved. 2.5.2. Data Analysis and Interpretation Validation and pre-processing are methodologically complex but essential steps to ensure the reliability of structural monitoring. The adoption of standardized protocols based on objective and verifiable criteria represents a necessary prerequisite for conducting reliable diagnostic analyses and supporting informed decisions regarding infrastructure management and maintenance.