Research On Fault Diagnosis Of Wind Turbine Bearing Based On Vibration Analysis And Neural Network Identification

The faults of wind turbine drive systems are frequent,and rolling bearings are one of the most critical components in wind turbine transmissions.Therefore,the fault diagnosis research on the use of bearings for wind turbines is of great significance for the stable and safe operation of wind turbines.When the rolling bearings are damaged,the collected signals are mostly non-stationary and non-linear,and traditional time domain and frequency domain analysis are difficult to accurately analyze the characteristics of these signals.The characteristics of bearing fault signals used in large wind turbines are studied.Among them,the methods of fault feature extraction and fault pattern recognition from the collected original signals are mainly studied.The different types of bearings used in wind turbines are introduced.The main types of bearings and the causes of faults are listed.In terms of fault feature extraction,the EEMD algorithm is first used to decompose the original signals of different fault states collected by the experiment.EEMD is an improvement of the modal decomposition(EMD)method,and the EEMD method can reduce the problem of modal aliasing in the EMD method.For this feature,a set of simulated signals was used for verification.After decomposition,the IMF of each fault type is obtained,and the unimportant IMF component is filtered by the correlation coefficient method.Then calculate the energy value of the IMF component of each type of fault and the energy ratio of the total energy value,and use the energy ratio as the fault feature vector element to construct each type of fault feature vector.After that,the original signal is decomposed using the wavelet packet decomposition method,and the fault signal is decomposed into three layers of wavelet packets.The ratio of the energy of the decomposed frequency band to the total energy of the signal is also regarded as the fault feature vector element,and each type is constructed.Fault feature vector.After the fault feature vectors are obtained,the fault patterns are identified using the neural network using the fault feature vectors.First,use a mentor to learn neural networks: extreme learning machines and probabilistic neural networks for identification.The neural network is then used to learn the competitive neural network and the self-organizing feature mapping neural network for identification.Finally,the results are compared to study the advantages and disadvantages of each method.Finally,the methods of fault feature extraction and fault pattern recognition are compared and summarized,and a relatively reliable fault diagnosis research method is obtained.Then design and develop applications that can be applied to bearing fault diagnosis.