Research On Multi-domain Fault Diagnosis Of Gearbox Of Wind Turbine Based On AVMD-ELM Algorithms

The traditional method of equipment fault diagnosis is based on spectrum analysis.However,with the improvement of computer hardware level and the development of machine learning algorithm,the equipment fault diagnosis method based on Intelligent data-driven has become the mainstream.By processing and analyzing the collected signal of the equipment,the signal feature quantities that can represent the operation status of the equipment are extracted,and the pattern recognition is carried out by using these feature quantities to realize the fault diagnosis of the equipment.In this paper,based on the above background,the signal processing technology,basic mathematical theory,machine learning algorithm and other multi domain knowledge are integrated to optimize and improve the existing technology,and a new multi-domain fault diagnosis method is proposed,which is applied to the fault diagnosis of wind turbine.The specific research work is as follows:Firstly,the existing wind turbines are classified,the similarities and differences between Direct-driven wind turbines and Doubly fed Induction wind turbines are analyzed,the mainstream Doubly fed Induction wind turbines are introduced emphatically,whose working principle is expounded.Aiming at the problem ofhigh failure rate,the common failures of Doubly fed Induction wind turbines are analyzed.The analysis of the statistical results shows that the fault of gearbox is an important reason for the economic loss of Doubly fed Induction wind turbines.Theref ore,the various faults of gears and bearings in the gearbox are analyzed,and the fault mechanism and vibration signal features are studied.Secondly,in the aspect of signal processing methods,the minimum mean envelope entropy MAEE)is used as the fitness function,and the fast grey wolf optimization algorithm(F-GWO)s introduced to optimize the variational mode decomposition(VMD).The effectiveness of AVMD is verified by establishing simulation signal in MATLAB,and its superiority is proved by comparing with EMD and other recursive decomposition methods.Thirdly,signal feature extraction is the core of diagnosis,which is the root cause of affecting the final accuracy.Therefore,a new multi domain fault feature extraction method is proposed in this paper.According to the characteristics of time domain,frequency domain and time-frequency domain of signals,the feature quantities of each domain are extracted,and principal component analysis(PCA)is used to simplify these feature quantities.The feature vectors which can fully characterize the operation status of equipment is designed,they are input to the extreme learning machine(ELM)for training.Then,the mult-domain fault diagnosis model of DFIG gearbox is constructed.Finally,the vibration signal of rolling bearing in actual gearbox is taken as an example.Firstly,the effectiveness of AVMD method in traditional spectrum analysis is verified;Secondly,through the comparative analysis of three design schemes and various methods,AVMD-ELM multi domain fault diagnosis method has the highest diagnostic accuracy among the four methods,and the diagnostic accuracy of different fault schemes reaches 98.8%,which fully illustrates the extraordinary significance of this method in wind turbine fault diagnosis.