Condition monitoring of Internal Combustion engines using empirical mode decomposition and Morlet wavelet

Abstract

The process of detecting deterioration in the performance of any system termed as condition monitoring and fault diagnosis is at the heart of the condition monitoring procedure. Use of acoustic signatures of Internal Combustion (IC) engines for the condition monitoring procedure is the basic motivation of this paper. Acoustic signatures of IC engines always carry relevant information. However, in many cases, these acoustic signatures might be corrupted by the surrounding noise resulting in a low signal-to-noise-ratio (SNR). Extracting features from the signals having low SNR becomes highly difficult. Therefore, those signals corrupted by noise should be preprocessed before extracting features from them. In this paper, a denoising method based on empirical mode decomposition (EMD) and Morlet wavelet is presented. This denoising method is an advanced version of “soft thresholding denoising method” proposed by Donoho and Johnstone [1,2] and “generalized soft thresholding method” proposed by Jing Lin [4]. Morlet wavelet based denoising eliminates the noise and improves the SNR significantly and Back Propagation (BP) is used further for classification of faulty and healthy IC engines. Results obtained by using these techniques for condition monitoring of IC engines are promising.