An approach of on-board multistage health monitoring for heavy haul railway bridge using bogie acceleration

Abstract

Recently on-board damage detection becomes a promising strategy in the field of bridge structural health monitoring. Due to the randomness of track irregularity and the weak vehicle vibration feedback from the bridge damages, it is difficult to effectively extract the damage characteristics from vehicle responses. This paper proposes an innovative multistage health monitoring approach with bogie accelerations for heavy haul railway bridge, which considers the influence of track irregularity and realizes the localization and quantification of bridge damage. Firstly, the bogie accelerations are preprocessed by the presented time-domain correlation analysis denoising method (TCADM) to suppress the unfavorable interference of track irregularity. Secondly, the Empirical Mode Decomposition (EMD) is exploited to decouple the effect of vibration coupling among bridge spans on bogie accelerations, and a confidence classification threshold related to Mahalanobis distance is defined by the Gaussian inverse cumulative distribution function to detect damage. Thirdly, the sliding window strategy is utilized to localize damaged region, realizing a trade-off between time cost and localization accuracy by the custom window length. Fourthly, the classification threshold and the k-means clustering analysis are combined to quantify the damage severity. The effectiveness and robustness of the proposed framework are appraised using blind tests, where one monitor operator simulates various damaged conditions by combining a 2D train-track-bridge coupling model with the Monte Carlo Sampling Method, and the other estimates the bridge health state using the provided bogie dataset without any prior knowledge. The results indicate that the proposed framework performs well in detecting, localizing and quantifying damage.