Firstly, according to nongaussian and nonlinear characteristics of rotor fault vibration signals, combining higher order statistics analysis,singular spectrum analysis, information entropy and multi-scale analysis, a conception of MSHOSSE was presented and applied to rotor fault feature extraction. Secondly, VPMCD was a new class discriminate approach, which was of excellent learning ability for small samples and multi_classification, however, the choice of model type existed subjectivity. Thus, an improved class discriminate method based on GA and VPMCD (GA_VPMCD) was presented in the course of rotor fault diagnosis using VPMCD by the global optimization performance of genetic algorithm (GA) and fusion diagnosis method. Finally, a novel intelligent fault diagnosis method based on MSHOSSE and GA_VPMCD was put forward. Simultaneously, the method was applied for rotor fault diagnosis. The experimental results show its effectiveness and superiority.
