Hybrid intelligent fault diagnosis based on quotient space

Aiming at the problem that existing hybrid intelligent models do not take into account the advantages and limitations of different diagnostic methods and fail to achieve complementary advantages of different classifiers, a new model of hybrid intelligent fault diagnosis based on quotient space is proposed. In this model, samples are granulated and granular layers are constructed by calculating equivalence and cluster analysis. Meanwhile, core features set (CFS) in every layer is extracted by features reduction algorithm. Then, support vector machine and anfis classifier are trained by CFS as sub-classifiers in corresponding layer. Finally, all results of sub-classifiers are integrated by weighted voting method as the output of hybrid model. This model is applied to fault diagnosis of roller bearing in bearing test bench. The application results show that the classification accuracy of hybrid model reaches to 100%, which is 2.2% higher than the highest accuracy of all sub-classifiers.