Vision-based Vibration Measurement and Damage Detection using 2D Autoregressive Models

Vision-based vibration measurement has emerged as a powerful tool in the field of structural health monitoring. The camera, as one of the non-contact sensors, can capture the structural dynamics with high resolution and without any mass loading effect. The defect detection through modal parameters, such as operational deflection shape and mode shape, can be further improved using a camera. In this study, a phase-based displacement measurement with an optimal complex steerable filter is used to accurately measure the structural vibrations. Based on the measured vibration data, a damage detection technique using autoregressive (AR) model is proposed that aims to detect different types of damage. First, a one-dimensional (1D) AR model is applied to the obtained signals to examine the feasibility of applying time series analysis techniques for data-based structural health monitoring. Then, a two-dimensional (2D) AR model that fully utilizes the high-resolution measurement provided by a camera is proposed for more effective damage detection. The experiment was conducted on a five-story structure with three different types of damage, and damaged detection and classification were efficiently performed using 1D and 2D autoregressive models.