Research On Fault Diagnosis Method Of Rolling Bearing Based On Ensemble Soft Competitive Yu’s Norm ART

As the most commom and important general components in all kinds of mechanical system,rolling bearing’s running state directly affects the performance of the whole system.Therefore,it is of great significance to diagnose bearing faults.On the basis of summarizing and absorbing the previous research results,this thesis puts forward the fault diagnosis of rolling bearing based on ensemble learning and soft competition Yu’s norm ART neural network.First,Fault sample points of fuzzy interface would be categorized wrongly by adaptive resonance theory(ART)with hard competition.Then a new fault diagnosis method by Yu’s norm ART based on soft competition is proposed.With the soft competition method of fuzzy competitive learning(FCL)being introduced into ART which is based on Yu’s norm,the neural nodes in the competition layer are trained according to the degree of membership between the mode node and the input,then fault samples are classified in turn.Bearing fault data is used to validate the fault diagnosis model,which proves that the fault diagnosis model can not only distinguish different faults,but also distinguish different fault degrees under the same fault type.Comparing with other methods for fault diagnosis,such as hard competition ART and fuzzy Cmeans clustering,the proposed method has higher diagnostic accuracy.Secondly,because of the weak performance of single model with less training samples,and the instability performance in the diagnosis of different fault signals,this thesis introduces the ensemble learning to further improve the performance of soft competition Yu’s norm ART.Sets of feature parameters are extracted from the fault signal with the distance evaluation technique,then different features are used to train each model as the input of the network,the final result of the fault diagnosis is identified by the ensemble learning with majority voting.The proposed fault diagnosis method is applied to the fault diagnosis of rolling element bearing and shows the better performance with less training samples.And the strong robustness of the proposed f method is proved with bootstrap method.According to the analyses above,the rolling bearing fault data is acquired in the rolling bearing fault vibration test,then the fault data enters the fault diagnosis model as the input of the network,after comparing with the single model,the ensemble soft competition Yu’s norm ART shows the better performance with less training samples.The theory research on Yu’s norm ART based on soft competition in thesis is of high value,and provides a new idea for the performance optimization of ART;besides,this thesis has a certain guiding value for the popularization and application of ensemble learning in neural networks.