Fundamental study on outlier detection in the passing sound of a vehicle over road-bridge joints

Abstract

This paper reports on the outlier detection in the passing sound of a vehicle over the bridge-joint using Chaos prediction analysis as well as frequency analysis. First, on the basis of frequency analysis of the passing sound, it was found that the irregular sound over the damaged joint can be identified by detecting the peaks of power in the band of 500 to 800Hz. Then, by assuming that the passing sound is Chaos series, Chaos prediction analysis was applied to the passing sound for outlier detection. Finally it was found that the irregular sound also can be detected by Chaos prediction error which is the difference between the prediction based on the intact sound and the measured one.