Research On Mechanical Bearing Fault Diagnosis Algorithm Based On Probability Envelope And Deep Belief Network

As one of the most common and important parts of mechanical equipment,mechanical bearings play an indispensable role in the safe operation of mechanical equipment.With the complexity of mechanical equipment and the diversification of monitoring methods,the close relationship between equipment and equipment,as well as the uncertainties and information fusion problems caused by the multi-source of bearing status information,it directly increases the difficulty of fault diagnosis.In addition,how to effectively utilize the massive equipment status data to improve the diagnostic accuracy is also the hot and difficult point of current fault diagnosis.For this reason,starting from the uncertainty of bearing state information and the direction of deep feature extraction,the fault diagnosis method of mechanical bearing based on probability envelope and depth belief network is studied.Firstly,this thesis introduces the research background of this topic and the related research of mechanical bearing fault diagnosis,and fully elaborates the necessity of this topic research.Secondly,the basic theories of uncertainty modeling and feature extraction of probability envelope,as well as the basic principles and infrastructure of deep belief network are introduced.Then,aiming at the uncertainties and lack of information in mechanical bearing fault signal acquisition,a fault diagnosis method based on probability envelope feature extraction is proposed.The specific process of probability envelope modeling and the feature extraction method based on probability envelope are described.Experiments show that the bearing fault diagnosis method based on probability envelope is reasonable and feasible,and good results are obtained.Aiming at the problem that it is difficult to further improve the accuracy of diagnosis after the data scale reaches a certain level,a method of bearing fault diagnosis based on probabilistic envelope and deep belief network is proposed.The model construction and training process of deep belief network are elaborated in detail.The uncertainty model of bearing state information is established by using probabilistic envelope,and the single index and multi-source information after modeling is carried out.The probabilistic envelope feature is fused,and the extracted probabilistic envelope feature is used as the input of deep belief network for further feature extraction and classification diagnosis.The experimental results show that the proposed method is effective in fault diagnosis of mechanical bearings.