Bearing fault signal feature extraction based on SVD and generalized S-transform module matrix

In order to extract the weak fault information from rolling bearing vibration signals, a method of feature extraction of bearing vibration signal based on singular value decomposition (SVD) and generalized S-transform module matrix was proposed. Firstly, mutant information is separated from the noise background and smooth signal by using SVD, according to the distribution state of singular value, selecting the transition stage of the singular value to extract mutant signal; then using the mean value of sum of squares of generalized S-transform module matrix amplitude to locate mutant information and the fault feature of bearing vibration signals are extracted for fault diagnosis. This method is used for representing characteristics of the bearing outer circle and inner circle partial fault, and through the fundamental frequency information can accurately detect and identify the type of fault. The result shows that this method proposed here is feasible and effective.