Adaptive Singular Value Decomposition and its Application to the Feature Extraction of Planetary Gearboxes

Abstract

Since the vibration signal of the planetary gear box is usually submerged by the noise, it is necessary to develop a method of weak fault characteristics extraction. Traditional singular value decomposition method is unable to select the number of the effective singular value automatically, a new fault diagnosis method of the planetary gear box based on adaptive singular value decomposition is, therefore, proposed. Firstly, according to certain conditions, several effective singular value numbers are selected and different reconstructed signals are obtained by the method. Secondly, the optimal reconstructed signal is automatically selected on the basis of the skewness of these reconstructed signals. At last, the envelope spectrum of the fault signal is acquired with the envelope analysis. The results of simulation and experiment show that this method performs better in eliminating noise and extracting the weak fault characteristics of the vibration signal in the planetary gear box compared to the traditional singular value decomposition.