PSI - Issue 84

284 Andy Duarte-Taño et al. / Procedia Structural Integrity 84 (2026) 280–287 Fig. 4 shows a representative reconstruction result at = 8 , comparing the original and reconstructed acceleration responses for a selected degree of freedom. The strong agreement observed in both the time and frequency indicates that the proposed compression strategy preserves waveform characteristics and the spectral content relevant for modal identification. To assess whether signal-level fidelity translates into preserved structural dynamics, OMA was performed on the reconstructed signals at above mentioned CR. The covariance-driven stochastic subspace identification algorithm (CoV-SSI), implemented in the software MOVA-MOSS [García-Macías and Ubertini (2020)], was employed using model orders from 2 to 160 and a time lag of 4.31s. Fig. 5 compares six representative mode shapes identified from the original and reconstructed signals. All matched modes exhibit MAC values above 0.98 , confirming the preservation of the spatial modal characteristics under high compression.

Fig. 4 . Synthetic vibration case: original and reconstructed acceleration response for a representative degree of freedom at CR = 8 and window length N = 256. (a) Full time-domain signal, (b) zoomed-in time window highlighting local waveform reconstruction accuracy, and (c) corresponding power spectral density (PSD) estimated using Welch’s method.

Fig. 5. Synthetic vibration case: Comparison of six selected vibration mode shapes identified by OMA from the original signals (blue solid lines) and the reconstructed signals (orange dashed lines) at = 8 and =256 ( ℎ : theoretical frequency; : original signal frequency; : reconstructed signal frequency). 3.2. Case study 2: San Jerónimo bell-tower monitoring The proposed framework was further validated using ambient vibration data recorded in the San Jerónimo bell tower in Granada, Spain. The structure is a sixteenth-century masonry tower [Hernández-Montes et al. (2023)] instrumented with two triaxial and two uniaxial high-sensitivity piezoelectric accelerometers located at two elevation levels. At each level, sensors were placed at opposite corners along the diagonal of the cross-section to enable the