Sparsity-Enhanced Model-Based Method for Intelligent Fault Detection of Mechanical Transmission Chain in Electrical Vehicle

Abstract

In today’s world, smart electric vehicles are deeply integrated with smart energy, smart transportation and smart cities. In electric vehicles (EVs), owing to the harsh working conditions, mechanical parts are prone to fatigue damages, which endanger the driving safety of EVs. The practice has proved that the identification of periodic impact characteristics (PICs) can effectively indicate mechanical faults. This paper proposes a novel model-based approach for intelligent fault diagnosis of mechanical transmission train in EVs. The essential idea of this approach lies in the fusion of statistical information and model information from a dynamic process. In the algorithm, a novel fractal wavelet decomposition (FWD) is used to investigate the time-frequency representation of the input signal. Based on the sparsity of the PIC model in the Hilbert envelope spectrum, a method for evaluating PIC energy ratio (PICER) is defined based on an over-complete Fourier dictionary. A compound indicator considering kurtosis and PICER of dynamic signal is designed. Using this index, evaluations of the impulsiveness of the cycle-stationary process can be enabled, thus avoiding serious interference from the sporadic impact during measurements. The robustness of the proposed approach to noise is demonstrated via numerical simulations, and an engineering application is employed to validate its effectiveness.