Steam Turbine Generator Vibration Fault Diagnosis

60%of electricity is produced by the steam turbine generator in our country,so the steam turbine generator plays an important role in power network?As a result of the steam turbine generator unit capacity increasing,for safe and reliable running of turbo-generator set is crucial,so monitoring the running state and implement fault diagnosis of steam turbine unit is very necessary?The vibration of steam turbine generator affect the steam turbine generator safe and reliable operation,even the service life of the steam turbine generator has great influence?And the vibration fault diagnosis of steam turbine generator is according to the vibration frequency,amplitude,and related information to determine the cause of vibration,the purpose of fault diagnosis is to the cause of the vibration range as narrow as possible,thus to a targeted maintenance?Bascd on the current steam turbine generator vibration fault diagnosis of large computational complexity,long time,etc.This paper presents a kernel principal component analysis and fuzzy neural network method for fault diagnosis of steam turbine generator.First using the eddy current sensor to extraction of steam turbine generator fault of the original signal,and the extraction of signal classification,using the optimal Wavelet Packet Basis noise reduction processing,then the noise signal after wavelet packet decomposition,to extract the energy of each frequency band as the feature vector.Secondly the kernel principal component analysis is used to analyse the eigenvector matrix and dimension reduction to extract in the main features of the fault characteristic value,and then after the dimension reduction of data can be divided into two types of training sample and test sample sets,each sample corresponds to a fault type,the training samples and test samples were imported into excel?Finally after dimension reduction of the training sample data as Takagi Sugeno fuzzy-neural network input and output,The matlab to establish Takagi-Sugeno type adaptive fuzzy neural network for training,will test sample data input into the trained network,get the output results?The test of the actual output and desired output,the correctness of the judgment diagnosis?The method with the least amount of data represent the maximum amount of information of the original data,the simulation performance were compared with the standard fuzzy neural network?BP neural network and RBF neural network,the results show that the diagnosis speed,fast convergence,high diagnostic efficiency etc,the simulation results proved that the method is an effective intelligent fault diagnosis methods,so the vibration fault diagnosis of steam turbine generator has a certain reference value?...