Textile spinning-frame roller fault diagnosis based on improved wavelet analysis and support vector machine

For roller fault diagnosis problems in textile spinning-frame, a method which combines an improved wavelet algorithm and support vector machine is presented. This method extracts feature information of the failure and classifies the failure. Based on the analysis of the frequency overlapping phenomenon encountered in traditional wavelet algorithm, the improved wavelet algorithm combines wavelet transform and FFT analysis to reduce the frequency overlapping phenomenon. The SVM classifiers are adopted because the sample available in the problem of roller fault diagnosis is small. Simulation results demonstrate the improved wavelet algorithm is superior to the common wavelet algorithm for the problem of the spinning-frame roller fault diagnosis, and SVM is very effective for fault identification.