Research On Rotating Machinery Fault Diagnosis Based On Variational Mode Decomposition And Neural Network

With the application of rotating machinery in the industry becoming more and more common and deeper,bearing and gear as the key parts of rotating machinery operation,the possibility of its failure is also increasing.Fault diagnosis based on vibration signal is widely used at present,and the time-frequency analysis method can analyze and process nonlinear and non-stationary signals,so it is widely used in practice.In the analysis of nonlinear and non-stationary signals,the traditional time-frequency analysis method has its shortcomings.Therefore,this thesis proposes an optimization for the deficiency of parameter setting in Variational Mode Decomposition,and VMD is used in time-frequency analysis to obtain time-frequency expression of bearing and gear faults.At the same time,the VMD and neural network is combined to study the new methods of fault diagnosis of rotating machinery under variable speed condition and compound condition.First,the theory of VMD is introduced.The performance of VMD in noise resistance and decomposition in a signal is verified by the ECG signal and simulation signal.The parameters of the VMD are optimized by Cuckoo Search,and the performance of the optimized algorithm is verified by simulation signal and MFPT fault signal.Secondly,for the fault diagnosis of bearings,combining the advantages of VMD in the decomposition of multi-component signals and Polynomial chirplet transform in the time-frequency expression of nonlinear signals,a fault diagnosis method based on CS-VMD and PCT bearings is presented.The single fault bearing,compound fault bearing and bearing whole life fault data are used to verify the effect of the proposed method.Then,for fault diagnosis of the gears,the advantages of Transient-extracting transform in time-frequency energy aggregation were verified by comparison.The theory of Resonance Sparse Signal Decomposition is expatiated and use it for sparse decomposition of gear signals to obtain high and low frequency components.On this basis,the gear fault diagnosis method based on CS-VMD and RSSD-TET is proposed and verified by experiment.Finally,the principle of Stacked sparse auto-encoder is presented.Combining VMD and SSAE,using sample entropy as a characteristic index,a gear fault diagnosis method based on CS-VMD-sample entropy and SSAE is proposed.At the same time,the qualitative and quantitative diagnosis of constant speed bearing,the diagnosis of compound fault of constant speed bearing,the fault diagnosis of variable speed bearing and the diagnosis of compound fault of variable speed bearing is studied emphatically.This paper presents a qualitative and quantitative fault diagnosis method of constant speed bearing based on CS-VMD-sample entropy and SSAE,a compound fault diagnosis method of constant speed bearing based on RSSD、CS-VMD-sample entropy and SSAE,the fault diagnosis method of variable speed bearing and compound fault diagnosis of variable speed bearing based on COT、CS-VMD-sample entropy and SSAE are also presented.The proposed methods are well validated by experiments.