Abstract: The acoustic signal of the valve gear train in a locomotive diesel engine is proved non-stationary and carries intense background noise. In view of these characteristics, empirical mode decomposition (EMD) and Laplace wavelet are used in acoustic signal analysis of locomotive diesel engines. The principles and algorithms of EMD are briefly introduced. Applying EMD to the real signal measured from locomotive diesel engines, the ability of signal denoising is demonstrated. EMD acts essentially as an adaptive filter bank. Acoustic signal acquired from diesel engine valve gear train is decomposed into a number of intrinsic mode functions (IMFs), with each IMF corresponding to a specific range of frequency components contained within the signal. The power spectrum analysis is applied to the intrinsic mode function which contains the gear fault characteristics for detecting the gear faults. Due to the fact that the signature of a damaged bearing consists of exponentially decaying sinusoids and the characteristic frequency is of very low energy level masked by noise, the correlation filtering based on Laplace wavelet is used to identify the signature and extract the characteristic defect frequencies of the bearings. The proposed method is effectively applied to the operation tests of locomotive diesel engines, and the faults of the valve gear train are diagnosed successfully.

Key words: Empirical mode decomposition, Fault diagnosis, Gear, Laplace wavelet, Rolling bearing