Vibration-Based Novelty Detection of Masonry Towers Using Pattern Recognition

During the last decades, the increased availability of continuously monitored structures has attracted the attention of the Structural Health Monitoring (SHM) community towards the development of automated techniques capable of continuously providing useful information to timely assess the health state of a structure. Over the years, especially the SHM procedures based on Operational Modal Analysis (OMA) have proved to be effective tools for the continuous assessment of large infrastructures and ancient constructions. Within this context, the paper presents the development and validation of a vibration-based novelty detection strategy based on the application of pattern recognition models to the identified natural frequencies, with the latter being used as damage-sensitive features. The methodology presented herein is based on the forming of a decision boundary through the use of a Support Vector Machine (SVM) model: hence, SVM is exploited to separate data into two classes, associated to two different structural conditions (i.e., undamaged and damaged), without any prior assumptions on the propriety of the data. The robustness of the developed approach is exemplified using the natural frequencies automatically identified during the continuous monitoring of a historic masonry tower. Due to the occurrence of a far-field earthquake, the tower underwent structural damage demonstrated by a slight permanent variation in the natural frequencies. The obtained results highlight the capability of the proposed approach to automatically reveal slight damages in structures without any user interaction and without performing any removal of environmental and operational effects.