Unsupervised Learning Techniques for Vibration-Based Structural Health Monitoring Systems Driven by Data

Abstract

Structural damage detection is a crucial issue for the safety of civil buildings, which are subject to gradual deterioration over time and at risk from sudden seismic events. To prevent irreparable damage, the scientific community has directed its attention toward developing innovative methods for structural health monitoring (SHM), which can provide a timely and reliable assessment of structural conditions. In this domain, the significance of unsupervised learning approaches has grown considerably, as they enable the identification of structural irregularities solely based on data obtained from intact structures to train statistical models. Despite the importance of studies on unsupervised learning methods for structural health monitoring, no reviews are specifically dedicated to this topic, considering the application part. The review of studies, therefore, made it possible to highlight the progress achieved in this field and identify areas where improvements could still be made to develop increasingly accurate and effective methods for structural damage detection.