PSI - Issue 84

Vincenzo Gattulli et al. / Procedia Structural Integrity 84 (2026) 41–48

46

and improving data interpretability. This is particularly relevant for underground infrastructure, where complex geometries and limited accessibility hinder traditional interpretation of monitoring data. Overall, embedding sensor data within a consistent geometric framework supports dynamic characterization, model validation, and more informed inspection and maintenance strategies. 5. Real-Time Data Processing, Dynamic Characterization and Model Validation Data collected through the sensor-based monitoring infrastructure described in Sections 3 and 4 are transmitted to a centralized server, where they are processed to extract quantitative indicators of system behaviour. Beyond acquisition and visualization, the framework integrates data processing with structural dynamics analysis, enabling the identification of modal properties and their preliminary comparison with numerical models. In this way, the Digital Twin operates not only as a data platform, but as a tool for engineering-level assessment of structural behaviour. 5.1. Data Processing and Preliminary Frequency Analysis Incoming data streams from distributed sensors are handled through a processing workflow supporting both online and offline analysis. Online processing enables monitoring of current operating conditions and detection of variations in the measured response, while historical datasets are retained for offline analysis aimed at identifying long-term trends and repeatable structural features. For vibration-related measurements, acceleration time histories are processed to extract dynamic features representative of structural behaviour. The workflow includes signal acquisition, preprocessing, and transformation into the frequency domain through power spectral density (PSD) analysis. The PSD results, shown in Fig. 3, are obtained from acceleration data acquired during short-term dynamic tests in the COSINUS experimental setup within Hall B, highlighting clear frequency peaks and coherent structural dynamics under operational conditions (Farrar and Worden, 2007). The stability of these peaks over time represents an indicator of structural consistency, while also providing preliminary estimates of modal frequencies. These results serve as a consistency check for the subsequent SSI-based identification and enable comparison with numerical model predictions.

Fig. 3. Power spectral density (PSD) of acceleration signals measured during dynamic testing, COSINUS experiment

5.2. Experimental Modal Identification (SSI) To further characterize the dynamic behaviour of the monitored structures, stochastic subspace identification (SSI) techniques are employed to extract modal parameters from measured data. The SSI analysis is performed on acceleration data acquired during short-term dynamic tests conducted in the COSINUS experimental setup within Hall B. The method is based on a state-space representation of the system, reported in the system of Equations (1): x ( +1)= Ax ( )+w( ), (1) y ( )= Cx( )+v ( ) where x ( ) is the system state, y ( ) the measured output, and w( ), v ( ) stochastic noise components. SSI enables the identification of modal frequencies, damping ratios, and associated displacement patterns under