Study On Fault Diagnosis Of Wind Turbine Gear Based On Vibration Signal Analysis

In wind turbine systems,gears are a frequently failing component and account for a very high proportion of wind turbine failures.The fault diagnosis of wind turbine gears is conducive to the timely maintenance of gears,which can effectively reduce economic losses and has certain practical significance.Three aspects of feature extraction,feature selection and fault diagnosis were investigated,and the signal was decomposed using an improved parametric adaptive optimization-seeking variational mode decomposition(VMD)method to filter out high-quality feature vectors and combine with an improved least squares support vector machine(LSSVM)to perform diagnosis.The main research and conclusions of this paper are as follows:(1)In terms of feature extraction,the collected gear vibration signals were decomposed,and experiments showed that VMD can effectively avoid the endpoint effect and modal mixing problems compared with Empirical mode decomposition(EMD),and had higher decomposition accuracy.The decomposition performance of VMD was affected by the number of modes k and the penalty parameter α.In order to avoid the blindness of parameter selection,an improved grey wolf optimizer(IGWO)was proposed to optimize the parameters of VMD,which further improved the decomposition accuracy of the signal.In IGWO,a non-linear convergence strategy and a dynamic weight position update strategy were proposed to balance the local and global search.The IGWO algorithm was verified by comparative experiments to be more efficient in finding the best.(2)In terms of feature selection,in order to remove irrelevant information from the extracted features and reduce the complexity of the model,the Competitive adaptive reweighted sampling(CARS)algorithm,Iteratively retains informative variables(IRIV)algorithm and Bootstrapping soft shrinkage(BOSS)algorithm were used for feature selection of the extracted feature variables,and it was experimentally verified that the IRIV had better feature selection performance.(3)In fault diagnosis,gear vibration data from the QPZZ-Ⅱ rotating machinery vibration analysis platform system was selected for study.Firstly,the gear vibration signal was decomposed using IGWO-VMD,then several features such as energy entropy,sample entropy and multi-scale alignment entropy were extracted,followed by feature vector screening by IRIV.To address the problem that the classification performance of the LSSVM model was affected by the kernel function parameter σ and the penalty parameter γ,the parameters of the LSSVM with an improved grey wolf optimisation algorithm was optimised,and the feature data were input into the optimised IGWO-LSSVM for fault identification,and the fault diagnosis results were compared with those of the SVM,GWO-SVM,IGWO-SVM,LSSVM and GWO-LSSVM algorithms.The experimental results showed that the established IGWO-LSSVM model had the best diagnostic performance,and the four types of gear faults(normal,pitting,broken tooth and wear)with an accuracy of 97.5% and the recognition rate of fault degrees reached 96.7%.Therefore,the gear fault diagnosis method based on IGWO-VMD and IGWO-LSSVM proposed in this paper can effectively identify different faults and gears with different degrees of the same fault,and also provide a reference for fault diagnosis of other components of gearboxes,which has certain engineering application value.