Fault diagnosis of rotating machines for rail vehicles based on local mean decomposition—energy moment—directed acyclic graph support vector machine

Abstract

In order to accurately perform fault diagnosis of key rotating machines of rail vehicles, a new method for diagnosis was proposed, based on local mean decomposition—energy moment—directed acyclic graph support vector machine. The vibration signals of rotating machines were decomposed by local mean decomposition to obtain the signal components, and then energy moment is calculated for each state feature component for feature extraction. For state identification, the directed acyclic graph support vector machine is established, multiple classical support vector machine were trained, and then multi-state identification was completed using directed acyclic graph support vector machine. The proposed method was tested on a train rolling bearing. Experimental results show that the method has nearly 95% identification accuracy and verified the feasibility and advantages of this method.