LAUSR

Abstract This paper presents a novel scheme for fault detection, isolation and diagnosis of multi-axle speed sensors on high-speed trains. Firstly, the steady features are extracted from dynamic signal measurement of multi-axle speed sensors, and then principal component analysis (PCA) is utilized for condition monitoring. Once an anomaly is detected, fault isolation is conducted using a reconstruction-based contribution (RBC) method. Moreover, a modified fault Petri net (PN) model is developed, and an indicator quantification idea is proposed to deduce an incidence matrix and discover diagnostic rules, which is practically suitable for diagnosing intermittent or time-varying fault. This proposed approach was demonstrated by cases study through test-rig experiments and actual fault records of train operations. Compared with traditional strategies, the results show that the developed approach can not only guarantee effective fault detection but also provide reliable fault isolation and diagnosis of multi-axle speed sensors even if when suffering axle-lock.