Research On Machinery Fault Diagnosis Based On The Variational Mode Decomposition And Depth Learning

China is currently in the stage of transformation and upgrading of intelligent manufacturing.Mechanical and electrical equipment is being upgraded to large-scale,complex,and intelligent directions.Its structure and composition are becoming more complex.Once a failure occurs,if it is not detected and eliminated in time,it can easily cause serious consequences.Rolling bearings are the most widely used components in mechanical equipment and they are also components that are prone to failure.Therefore,in order to eliminate safety hazards,ensure the normal operation of machinery and equipment,and improve the economic benefit of the enterprise,the real time state monitoring and fault diagnosis of the bearing of the mechanical equipment must be carried out.Aiming at the diagnosis of mechanical fault,this paper uses the method of combining variational mode decomposition and sample entropy to extract the characteristics of the vibration signal,and introduces the method of deep learning to complete the pattern recognition of the mechanical fault.Firstly,the fault mechanism and common diagnostic methods of rolling bearings are analyzed,and the problems existing in feature extraction and pattern recognition methods used in fault diagnosis are summarized,which lays a theoretical foundation for the research of new methods.Secondly,in order to solve the problem of modal aliasing when other methods are dealing with nonlinear and non-stationary mechanical fault signals,the basic principle and algorithm flow of the variational mode decomposition method are studied,and a mechanical fault feature extraction method based on the variational mode decomposition is proposed.After that,the variational mode decomposition method and the empirical mode decomposition method are simulated.Through the comparison and analysis,the variational mode decomposition method is proved to be very good in the anti mode aliasing and filtering.Finally,through the decomposition experiments of different parts of the rolling bearing and the fault signal of different damage degree,the validity of the variational mode decomposition method in feature extraction is verified.Thirdly,in view of the problems of poor classification stability and low recognition precision in traditional mechanical fault identification methods,it is found that deep learning has unique advantages in pattern recognition,and is widely used in the fields of speech recognition and machine vision.Therefore,deep learning is introduced into the field of mechanical fault diagnosis,which provides a new method and idea for mechanical fault diagnosis.Finally,a new method of mechanical fault diagnosis based on variational mode decomposition and deep learning is proposed.Mechanical failure vibration signal is decomposed into a series of modal components by variational mode decomposition method,then sample entropy of each modal is obtained and the feature vectors are formed to complete the feature extraction.Then the feature vectors are input into deep belief network to identify the mode of the mechanical fault and finally complete the fault diagnosis.The pattern recognition of mechanical failures will eventually complete the diagnosis of the failure.The experimental data of the laboratory rolling bearing of Case Western Reserve University was tested,and compared with other methods.It was verified that the method based on variational mode decomposition and deep learning can quickly and accurately complete the bearing fault diagnosis.