A fault diagnosis approach for diesel engines based on self-adaptive WVD, improved FCBF and PECOC-RVM

Abstract

Targeting the cross-term interference of Wigner–Ville distribution (WVD) and the redundancy control problem of fast correlation-based filter (FCBF), a novel approach based on self-adaptive WVD, improved FCBF and relevance vector machine (RVM) is proposed for the identification of diesel engine fault in this study. The approach primarily consists of three stages. Firstly, the self-adaptive WVD method is used to generate the time–frequency images of the vibration signals of the diesel engine. Secondly, four types of commonly used image features, including moment invariants, gray statistical characteristics, textural features and the differential box-counting fractal dimension, are calculated for all of the time–frequency images. Next, the improved FCBF method is proposed and used to select the relevant but non-redundant fault features. Finally, the probability-based error correcting output codes (PECOC) method is used to determine the structure of the multi-RVM model, and the fault diagnosis results can be obtained by inputting the selected fault features into the PECOC-RVM classifier. The simulation and experimental results demonstrate that the proposed fault diagnosis approach can overcome the cross-term interference of WVD, effectively extract the relevant but non-redundant fault features and accurately identify the fault types of diesel engines.