Electric transmission fault localization based on PSO-MVMD and weighted Gaussian membership model

Abstract

Electric transmission is a key component in the powertrain of heavy vehicles. Due to its highly compact component layout and complex operational environment, fault features in vibration signals are easily overwhelmed by noise. To address this issue, this paper proposes a fault localization method that combines modal decomposition and weighted Gaussian membership model. Firstly, considering the spatial layout characteristics of the electric drive device, we use PSO algorithm to search for the optimal parameter combination of multivariate variational mode decomposition, with the goal of minimizing the overall envelope entropy of multi-channel vibration signals. Secondly, we extract principal component features representing the main energy structure of the signal and sparse features sensitive to fault features from the decomposed multivariate intrinsic mode functions, and construct a mixed feature vector with high discriminability and low redundancy. Finally, we construct a weighted Gaussian membership model by calculating the global weighted similarity between sample features and various state template libraries, achieving accurate identification of fault patterns and localization of faulty components. The bench test results show that this method can effectively identify six typical working conditions, including normal state, bearing inner ring cracks, and gear tooth fractures, with an overall average diagnostic accuracy of 94.33%. This provides an efficient and reliable engineering solution for fault localization of electric transmission in heavy vehicles.