Research On Fault Diagnosis Method Of Gearbox Based On Variational Mode Decomposition

For gearbox contains the characteristic of stable transmission ratio,large transmission torque and the compact structure,its role in adjusting the speed of rotating machinery equipment and transferring power is irreplaceable.Since the probability of a gearbox failure is very high,when the gearbox is in a fault condition,the overall functionality of the rotary machine equipment will be severely affected.So it is very meaningful to study the fault diagnosis method of gearbox.This article researched the gearbox fault diagnosis method as follows:1.The problem of the number of decomposition modalities need to be presetted empirically for the Variational Mode Decomposition(VMD),a Particle Swarm Optimization(PSO)algorithm based on amplitude spectrum to optimize the VMD parameters is proposed and applied to extract the feature parameters of gear fault.Firstly,The PSO is used to optimize the two parameters K and α of the VMD;secondly,gear data is decomposed using VMD of optimized parameter to obtain K modal components;then the singular value decomposition is performed on the matrix constructed by K modal components to obtain K singular values and form the features vector;finally,the Euclidean distance classifier is constructed with the feature vector to diagnose the gear fault.This method is applied to the rotating machinery and fault simulation test platform,the simulation results showed that the proposed method can effectively extract the feature of gear fault and perform troubleshooting.2.In order to solve the problem that the single evaluation criteria is difficult to fully evaluate the fault sensitivity of the gear feature parameters,feature parameter selection method of multi-criterion fusion is proposed based on D-S evidence theory.First of all,three evaluation criterias consisting of the ReliefF algorithm,the dispersion ratio and the improved distance method were respectively used to evaluate the gear feature parameters.Then,according to the evidence theory of D-S,feature evaluation results using the above three methods are fused to obtain the comprehensive weight vector.Finally,the optimal subset of features is determined according to the size of the integrated weight.The method is verified used the measured data,the proposed method compared with three kinds of single-criterion feature selection method and ranking feature fusion method,the results show that the proposed method effectively improves the accuracy of fault diagnosis.3.In order to further improve the diagnostic accuracy,improved KNN classification method with gaussian weight is applied to gears fault diagnosis.Because the algorithm determines the weight of each neighbor sample according to the distance between each neighbor sample and the test sample which can effectively amplify the distance between different types of samples,so the algorithm improved the classification accuracy.Finally,the results show that the improved KNN algorithm is superior in diagnosis compared with KNN and LDA.