Signal Processing Techniques For Structural Health Monitoring In Civil And Mechanical Engineering Applications

Impact-based structural health monitoring is based on sensor measurements to gain a structural behavior understanding, although complex wave propagation effects often make relationships between measured quantities and impact location without a direct causation. This work examines signal processing techniques for feature-based impact localization using multi-sensor transient response data in civil and mechanical engineering applications. Time-domain impact responses were preprocessed, segmented, and converted into statistical feature representations for each sensor. Feature standardization and exploratory principal component analysis were used to examine spatial trends in the feature space, while Pearson correlation analysis assessed the sensitivity of individual features to impact location. Impact localization was formulated as a regression problem using Ridge regression, and performance was evaluated with a group-aware holdout strategy to reduce information leakage across repeated experiments. The results show that spatial information is reflected through gradual variations in the feature space rather than explicit separation. Individual features exhibited moderate sensitivity to location, while feature fusion enabled consistent localization performance, with a mean coefficient of determination of 0.696 across validation splits. These findings indicate that interpretable, feature-based signal processing approaches can support impact localization in controlled settings and provide a practical baseline for further SHM studies.