Study on Fault Diagnosis Method of Emulsifier Based on Empirical Wavelet Transform and SVM

Abstract

For the problem that the signals measured in the fault diagnosis of the emulsifier are non-stationary, non-linear signals and contain a large amount of background noise. This paper presents an emulsifier fault diagnosis method based on empirical wavelet transform (EWT) and Support Vector Machine (SVM). Empirical wavelet transform is a new adaptive signal decomposition method, which inherits the respective advantages of Empirical Mode Decomposition (EMD) and wavelet analysis methods. By extracting frequency domain maximum points to adaptively split the Fourier spectrum to separate different modalities. Then, a band-pass filter bank is adaptively constructed in the frequency domain to construct an orthogonal wavelet function to extract an AM-FM component with a Compact Support Fourier spectrum to better extract fault features. At the same time combining the advantages of SVM in small sample classification and identification, a fault diagnosis model based on EWT and SVM is proposed. The experimental results show that this method can extract the rotor rub-impact fault characteristics better. Compared with the EMD method, the improved EWT algorithm is better.