Research On Degradation State Identification Methods Of Planetary Gear

Manufacturing industry is the main part of the national economy and the foundation of the country.As China's economic development enters a new normal,the development of manufacturing industry is facing new challenges.Its development direction will be the deep integration of information technology and manufacturing technology,which means that mechanical and electrical equipment will be intelligent in the future.As an important part of the transmission system of mechanical and electrical equipment,planetary gear is one of the main maintenance components of mechanical and electrical equipment.In order to prevent gear from degenerating from normal state to serious state and cause serious consequences,it is of great practical significance to study the identification method of planetary gear degradation state.Planetary gear transmission is a more complex gear transmission,usually in a bad,complex working environment,so its vibration signal analysis is more complex than fixed-axis gear signal analysis.Because the vibration signal produced by planetary gear meshing is generally non-stationary and nonlinear.And there is little difference in the characteristics of various degraded states when the same fault occurs,which result in the recognition accuracy of planetary gear degradation state is not high.Therefore,this paper takes the planetary gear solar gear as the research object,and studies a method which can accurately identify the degradation state of the planetary gear.The main research contents are as follows:(1)Firstly,the traditional discrete wavelet analysis(DWT)signal processing method is introduced.Then the selection of wavelet basis is introduced,which does not have the disadvantage of adaptability.After that,the empirical mode decomposition(EMD)method,which can be used for adaptive time-frequency analysis of nonlinear non-stationary vibration signals,is introduced,but EMD has the problems of modal aliasing and endpoint effect.In order to solve the modal aliasing problem,an improved algorithm of EMD,the ensemble empirical mode decomposition(EEMD)is proposed,and the other improved algorithm,local mean decomposition(LMD),is proposed to solve the endpoint effect problem.And there are still error accumulation and endpoint effect.Although both EEMD and LMD inherit the adaptability of EMD and improve the decomposition accuracy to a certain extent,the two algorithms are still recursive pattern decomposition in essence,and algorithms still exist error accumulation and endpoint effect.In order to avoid the decomposition error caused by recursive decomposition,variational modal decomposition(VMD)is introduced in this paper.This decomposition method can not only complete adaptive time-frequency decomposition,but also accurately analyze the signal components of near frequency.Compared with DWT,EMD,EEMD and LMD,VMD is more suitable for dealing with nonlinear and non-stationary gear vibration signals.(2)In view of the fact that the planetary gear is in the same fault mode in the degradation process,the feature difference between different degradation states is small,and the common feature extraction methods are only for a single angle of feature extraction.In this paper,a feature extraction method for degraded state of planetary gear based on VMD-multi-angle feature fusion is proposed.The vibration signals of different degraded states of planetary gears are analyzed by VMD,and then the feature information of degraded gears in time-frequency domain,time domain and frequency domain is extracted by using multi-scale fuzzy entropy,energy entropy and marginal spectrum frequency band energy feature extraction method.The feature vectors extracted from these three angles are fused into a high-dimensional eigenvector which reflects the degraded planet gear from multi-angle features.(3)In view of the fact that the high-dimensional eigenvector may cause the problem of over-fitting in the training process of the classifier recognition,the local dimension reduction algorithm is introduced to reduce the dimension of the high-dimensional data set.In this paper,three local dimension reduction methods,locally preserving projection(Locality Preserving Projections,LPP),Hessian local linear embedding algorithm(Hessian Locally Linear Embedding,HLLE)and linear local tangent space arrangement(Linear Local Tangent Space Alignment,LLTSA),are used respectively.The high-dimensional features extracted from the degraded state of planetary gears based on VMD-multi-angle feature fusion are reduced in dimension.The degraded feature of low-dimensional star wheel obtained from dimension reduction is recognized by ANFIS,and the dimension reduction method with the highest recognition accuracy is selected and applied to the recognition of degraded state of planetary gear.(4)In order to improve the recognition effect of the degraded state of the planetary gear,the particle swarm optimization algorithm(Particle Swarm Optimization,PSO)is used to optimize the penalty factor and kernel parameters of the support vector machine(Support Vector Machine,SVM).It is used as a classifier to identify the low-dimensional features of the planetary gear obtained by the optimal dimension reduction method.At the same time,the degraded eigenvector of planetary gear after dimension reduction is identified by kernel limit learning machine(Kernel Extreme Learning Machine,KELM).The recognition efficiency of the two recognition algorithms is compared,and the classifier to find the optimal recognition effect is used to identify the degraded state of planetary gears.Through the analysis,it is fiinally found that the planetary gear degradation state recognition method based on VMD-multi-angle feature fusion-LLTSA and PSO-SVM can more accurately identify the different degradation states of solar wheels.At the end of this paper,the research content is summarized,and the related technology research is prospected.