Research On Vibration Warning And Fault Diagnosis Method Of A 600MW Turbine Generator Unit

Turbine generator unit,fan,feed water pump and other rotating machinery is the key equipment of thermal power plant,which occupies an important position in the production of electric power industry in our country.It is of great practical significance and social and economic value to monitor its normal working condition and explore and study the technical methods of fault diagnosis.Based on the engineering practice of thermal power plants,this paper carries out resea rch on the three aspects of shafting vibration warning,fault feature extraction and fault state identification of steam turbine generator sets.The specific research contents are as follows:Firstly,jump vibration alarm and machine of coal-fired power plants of the present rules,the steam turbine generator unit operation parameters on the influence of the rotor shaft system vibration,the combination of expert experience to determine the vibration impact parameter,use the data analysis method such as da ta cleaning,data integration to build the healthy operation of data sets.The mapping model between the operating parameters of the unit and the vibration of the bearing bush is established by using the BP neural network which is widely used at present.According to the numerical comparison between the predicted value of the vibration of the model and the measured value of the vibration,when the difference between the two exceeds the given threshold,an alarm signal is sent out,and the diagnosis module is triggered by the alarm signal.The analysis of the measured data of a600 MW unit shows that the proposed method can satisfy the engineering application well.On this basis,in order to make accurate diagnosis of the alarm signals sent by the unit,and considering the nonlinear and non-stationary characteristics of rotor vibration signals,Singular Spectrum Decomposition(SSD)i s applied to the fault feature extaction of the rotor system.In view of the poor stability of the method which takes the energy ratio as the stopping condition of iteration,the mutual information criterion is analyzed to improve the stopping condition of iteration,and the improved SSD method is obtained.Combined with the excellent time-frequency resolution of Teager energy operator demodulation and the characteristics of signal feature tracking,a new time-frequency analysis method is proposed to extract the time-frequency characteristics of rotor fault signals.Finally,the effectiveness of the proposed method is verified by the simulation signals and the engineering measured signals.The proposed method can accurately extract the rotor fault time-frequency characteristics,and the diagnosis effect is better than the traditional time-frequency analysis method Hilbert-Huang Transform(HHT).Finally,a new method for rotor fault diagnosis based on modified SSD,Refined Composite Multiscale Dispersion Entropy(RCMDE)and Support Vector Machine(SVM)is proposed.This method the basic idea is: first,to improve the rotor vibration signal of SSD decomposed several Singular Spectral Component(SSC),will be to refactor get SSC component signals,and then uses the RCMDE reconstruction algorithm for mining the characteristics of the components in the different time scales and construct entropy feature vector,after input entropy feature vector into the SVM classifier was trained and tested,so as to realize the rotor fault state recognition.The analysis of measured rotor fault signals shows that the proposed method can accurately identify rotor fault types,and has higher identification accuracy compared with Empirical Mode Decomposition(EMD),Variational Mode Decomposition(VMD)and RCMDE methods.