Fault isolation of Light Rail Vehicle suspension system based on D-S evidence theory

This paper presents an innovative approach for the fault isolation of Light Rail Vehicle (LRV) suspension system based on the Dempster-Shafer (D-S) evidence theory. The considered LRV has three rolling stocks and each one equips three sensors for monitoring the suspension system. A Kalman filter is applied to generate the residuals for fault diagnosis. For fault isolation purpose, both fault feature in the time domain (mean, standard deviation, skewness and kurtosis) and frequency domain (frequency centre, root mean square frequency and root variance frequency) are used and a fault feature database in the time and frequency domain is built in advance. The norm distance between the fault feature of the new occurred fault and the one in the feature database is applied to measure the similarity of the feature which is the basis for the basic belief assignment to the fault. After the seven pieces of basic belief assignments are obtained, they are fused by using the D-S evidence theory. The fusion of the basic belief assignments increases the isolation accuracy significantly. The efficiency of the proposed method is demonstrated by a case study and its simulation results.