A Novel Wavelet Transform – Empirical Mode Decomposition Based Sample Entropy and SVD Approach for Acoustic Signal Fault Diagnosis

Abstract

An advanced and accurate intelligent fault diagnosis system plays an important role in reducing the maintenance cost of modern industry. However, a robust and efficient approach which can serve the purpose of detecting incipient faults still remains unachievable due to weak signals’ small amplitudes, and also low signal-to-noise ratios (SNR). One way to overcome the problem is to adopt acoustic signal because of its inherent characteristic in terms of high sensitive to early stage faults. Nonetheless, it also suffers from low SNR and results in high computational cost. Aiming to solve the aforesaid problems, a novel wavelet transform - empirical mode decomposition (WT-EMD) based Sample Entropy (SampEn) and singular value decomposition (SVD) approach is proposed. By exerting wavelet analysis on the intrinsic mode functions (IMFs), the end effects, which decreases the accuracy of EMD, is significantly alleviated and the SNR is greatly improved. Furthermore, SampEn and SVD, which function as health indicators, not only help to reduce the computational cost and enhance the SNR but also indicate both irregular and periodic faults adequately.