Railway bearing and cardan shaft fault diagnosis via an improved morphological filter

Railway faults are usually observed as impulses in the vibration signal, but they are mostly immersed in noise. To effectively remove noise and identify the impulses, an improved morphological filter is proposed in this article. The proposal focuses on two aspects: a novel gradient convolution operator is proposed for feature extraction, and a new fault sensitivity measurement algorithm is proposed for scale selection because a morphological filter’s effectiveness is mainly determined by these two elements. The performance of the improved morphological filter is evaluated with real vibration signals measured from train’s axle bearings and cardan shafts. From the analysis of three sets of railway faults, the results indicate that the proposed morphological filter effectively detects the faults. Compared with three reported morphological filters, the proposed method has better diagnosis effectiveness.