Internal Combustion Engine Fault Diagnosis Method Based MICEEMD-PWVD

Abstract

Abstract As to multi-signals, the figure of time-frequency distribution becomes difficult to understand due to cross-terms in the Wigner-Ville distribution (WVD). At present time, cross-terms suppression of time-frequency analysis is one of the top interests in signal processing, and more and more research had been put on this topic, but the present research on cross-terms suppression exist conflicting problem that these methods couldn’t suppress the cross-terms while holding high time-frequency resolution. A novel method of cross-terms suppression in the WVD based on Complementary Ensemble Empirical Mode Decomposition (CEEMD) was proposed to solve the above problem. Firstly, multi-component signals were decomposed into intrinsic mode functions (IMFs) which were single component signal by CEEMD; Secondly, the Mutual Information values of IMFS and original signal were calculated to judge and delete the CEEMD false components; Thirdly, these Wigner-Ville distributions of true component of IMFs were computed; Finally, WVDs of each single component signal were added linearly to reconstruct the WVD of original signal. Test on synthetic data shows the effectiveness of the proposed method, which can suppress the cross-terms while holding high time-frequency resolution. Apply MICEEMD - PWVD in internal combustion engine valve clearance fault diagnosis, useTD-2DPCA method to extract MICEEMD-PWVD time-frequency image feature, then classify the characteristic parameters with the nearest neighbor classifier. The results show that the IC Engine Fault Diagnosis Method Based MICEEMD-PWVD and TD-2DPCA can accurately diagnose IC engine valve fault.