Feature Extraction for Vibration Signal of Asynchronous Motor Based on Quaternion K-L Transform

This paper proposed a new method of feature extraction for vibration signals of asynchronous motors based on quaternion K-L transform. The vibration signals were detected by the vibration detector with three channels X, Y, Z, that were orthogonal axes. We constructed quaternion sequence based on signals of the three channels, and than the quaternion sequences were transformed by quaternion K-L method to extract features for asynchronous motors with faults. In order to verify it´s effectiveness, this method was applied combined with SVM classifier, the experimental results had proved its efficiency and feasibility, and dominated that it could be well applicable for asynchronous motor fault diagnosis.