Fault Diagnosis Of Wind Turbine Rolling Bearing Based On Data-drive

As an important component of wind turbine,the rolling bearing has been working in the high-speed running state,which makes it easy to fail.Therefore,it is of great significance to research the fault diagnosis of wind turbine rolling bearing.In this thesis,the rolling bearing of the wind turbine drive system is taken as the research object.Data processing,feature extraction and fault classification are performed on vibration signal of the rolling bearing,and the data-driven fault diagnosis strategy of the wind turbine rolling bearing is studied.The main research contents are as follows:(1)Aiming at the problem that the vibration signal of wind turbine bearing has the characteristics of non-stationary and non-linear,and the fault feature extraction is difficult,the fault diagnosis method based on deep belief network(DBN)is proposed.The thesis processes the vibration signal of the rolling bearing through the improved EEMD-AR,and gets the coefficient of AR model.The coefficients are input into the DBN to mine the fault characteristics of bearings and realize fault classification.Taking Western Reserve University rolling bearing data as experimental data in the simulation,the results show that this method can extract the fault features of different faults of wind turbine rolling bearing,and has higher accuracy than the back propagation neural network(BPNN)and support vector machine(SVM).(2)In order to solve the problems of abundant sampled data and subjective selection of fault features in wind turbine bearing fault detection,a method of signal compression acquisition,automatic feature extraction and fault classification for wind turbine bearing fault is proposed.The nesterov accelerated gradient(NAG)and QR decomposition theory are used to optimize the random Gaussian observation matrix,which can realize the compression acquisition of vibration signal.The convolutional neural network(CNN)is used to extract fault features and implement fault classification.The results show that this method can ensure the accuracy of fault diagnosis and shorten the training time of CNN.(3)Considering that the sensors of wind turbines are susceptible to environmental interference,the fault diagnosis results of wind turbine bearing will be affected.In order to improve the reliability of fault diagnosis,the multi-source signal fault diagnosis method is proposed.In this method,the time-domain and frequency-domain features of bearing vibration signal,noise signal and temperature signal are extracted.Bayesian optimization algorithm is used to optimize the node structure of each hidden layer of stacked denoising autoencoders(SDAE).The feature fusion and classification are completed by optimized SDAE.Taking the vibration and fault simulation experimental platform data of rotating machinery as experimental data in the simulation,the results show that this method has higher accuracy than the single signal method,and enhances the anti-interference performance.