Probabilistic cable condition monitoring for cable-stayed bridges using Gaussian process regression

Abstract

Structural condition monitoring techniques based on sensor data have been studied, and they have been applied to various civil infrastructures including cable-stayed bridges for assessing their structural safety. However, such a technique typically requires pre-defining a threshold to detect anomalies, which can be heuristic. This study suggests a new probabilistic method for the cable condition monitoring of cable-stayed bridges using Gaussian process regression (GPR). To monitor the condition of multiple cables in parallel, the GPR analysis is conducted based on multi-input multi-output (MIMO), which enables to set the threshold values for anomaly detection in a reasonable manner. To demonstrate the proposed method, it is applied to an actual cable-stayed bridge in the Republic of Korea, and the corresponding analysis results are presented.