Application of Bearing Diagnosis Based on Wavelet Preprocessing Neural Network

Abstract

According to compactly supported characteristics of bearing local damage diagnosis based on wavelet analysis, the construction of compactly supported orthogonal function method are studied. With the wavelet multi-resolution analysis and multivariate time series analysis as the foundation, the parameters of weights, scaling and translation in the model are modified. Based on MATLAB, the non-stationary signal of fault bearing acceleration wave is numerical simulated and calculated. Through the comparison of convergence between the parameters revising before and after, it shows that the error on the amplitude and the convergence efficiency of parameter modification are better than the fixed before, which can significantly improve the accuracy of bearing fault diagnosis and signal processing capacity.