Research On Fault Diagnosis Method Of Wind Turbine Gearbox Based On Unsupervised Learning

As the core component of the wind turbine,the gearbox is a high-load,fast-speed working condition for a long time.Under continuous impact,the components in the gearbox are prone to failure,and due to the complicated internal structure,Fault diagnosis is difficult,so it is one of the hot topics in recent years.Aiming at the difficulty of gear fault feature extraction in fan gearbox and the failure of adaptive classification of traditional classification methods,this paper proposes a fault diagnosis method based on set empirical mode decomposition(EEMD)energy entropy and self-organizing map(SOM)neural network..First,the EEMD method is used to adaptively decompose the original vibration signal of the gear under various working conditions into several eigenmode functions(IMF),calculate the energy value of each IMF and the energy entropy of the signal;then,select valuable The IMF energy ratio and the signal energy entropy form a feature vector,and the dynamic feature vector capable of reflecting the fault vibration signal is input into the SOM neural network for adaptive classification.The simulation results show that the proposed method can effectively extract the fault characteristics of the fan gearbox and has high accuracy and adaptability in fault identification.Since the adaptive resonance neural network(ART2)solves the dilemma of other artificial neural networks in terms of "adaptation" and "stability",it has been widely used in pattern recognition.However,the ART2 network uses the "hard competition" method for classification,which leads to the degradation of classification accuracy.The Gaussian mixture model(GMM)Clustering algorithm is proposed to "soften" the clustering results of the ART2 neural network.The simulation data of wind turbine gearbox is simulated.The results show that the correct diagnosis rate of GMM-ART2 model is 90%,which is obviously improved compared with the traditional ART2 neural network classification results.The unsupervised learning is applied to the research of gearbox fault diagnosis.The simulation experiments show that the two diagnostic models proposed in this paper can provide a new effective means for gearbox fault diagnosis.