Time-Frequency-Based Bridge Scour Identification by Trend-Change Detection

Abstract

A bridge scour identification method was developed based on a time-frequency analysis that was combined with trend change detection. The basic concept behind this method is to probabilistically consider the interference-induced random variations in the natural frequencies to reveal their scour-induced change trends. First, the natural frequencies were instantaneously analyzed based on the bridge structural health monitoring (BSHM) data by time-frequency analysis. The probability distribution of the frequencies is obtained by kernel density estimation (KDE) and normalized by the probability integral transform theory. Then, based on the normal fluctuation in the frequencies that were defined by the control limits, an anomaly diagnosis for the frequencies is conducted using the control chart method. Continuously diagnosed anomalies are further regarded as a detected trend change in the frequencies, which is the key indicator to warn of scour in terms of the probability (P). Finally, the Jintang Bay Bridge, Zhoushan, Zhejiang Province, China, which is a 1,210-m-long cable-stayed bridge, was selected as a case study to comprehensively illustrate the application of this method using numerical simulations. The long-term BSHM field data from the Anqing Yangtze River Bridge, Anqing, Anhui Province, China, which is a 1,038.5-m-long cable-stayed bridge, were used as another case study to effectively prove the validity of this method in practice. These applications showed that the proposed bridge scour identification method that was based on time-frequency analysis and probabilistic trend change detection is effective. The developed methodology does not require any underwater devices and operations and could be conveniently integrated into a routine BSHM system.