Fault feature extraction of cylinder-piston wear in diesel engine with EMD

Abstract

Aiming at the characteristics of the vibration signals measured from the diesel engine, a novel method combining empirical mode decomposition (EMD) and lifting wavelet denoising is proposed, and is used for feature extraction and condition evaluation of diesel engine vibration signals. Firstly, the original data was preprocessed using the lifting wavelet transformation to suppress abnormal interference of noise, and avoid the pseudo mode functions from EMD. Obtaining intrinsic mode functions(IMFs) by using EMD, the instantaneous frequency and amplitude can be calculated by Hilbert transform. Hilbert marginal spectrum can exactly provide the energy distribution of the signal with the change of instantaneous frequency. The vibration signals of diesel engine piston-liner wear were analyzed. The analysis results show that the method is feasible and effective in fault feature extraction and condition evaluation of diesel engine. © 2012 Springer-Verlag GmbH.