Research On Oil Instability Fault Of Steam Turbine Generator Set Early Warning Method

Oil instability fault is one of the common faults of turbine generator set, belongs to the category of low frequency vibrations. With the high-parameter and large capacity unit increase, corresponding to increased costs, due to the oil instability fault has sudden, in the event, the amplitude will reach a high value in seconds, not only will affect the normal operation of the unit, and can cause severe damage to parts of the unit, to the country’s huge economic losses. In order to improve the state of the turbinecondition monitoring and fault diagnose levels, in order to ensure safe and economical operation of the unit. So, for early warning of imminent oil instability fault.This paper through to the oil instability fault mechanism analysis and related parameters influence on the oil instability fault mechanism analysis.And fault characteristics of oil instability fault analysis, FMEA and FTA analysis, and according to the fault tree and FMEA table of the bottom events of the cause of the problem put forward the corresponding treatment measures, and finally put forward the oil instability fault fault diagnosis process.introducesThrough wavelet analysis, time series analysis, artificial neural network principle methods. First, the statistical properties of the raw vibration data analysis, including the stability analysis, correlation analysis of time series; Then noise by wavelet analysis of data trends before and after de-noising; Then ARMA time series model to forecast, analyze estimate after de-noising effect, indicating that after de-noising predict significantly better than the direct effect of the original sequence to predict; Artificial neural network is used mainly BP neural networks and genetic algorithms BP neural network to predict the vibration data, compared to predict the effect of the two methods of analysis, through genetic algorithm optimized BP neural network to predict the effect of significantly improved; Early warning mechanism study oil instability faultof turbine statistics primarily through a large number of oil instability fault case, the oil instability fault mechanism analysis, vibration signal characteristic analysis, the cause and the corresponding measures and fault warning system in oil instability fault, and the establishment of oil-whirl failure warning system.Time series analysis and neural network forecasting method apply to turbine oil instability fault can be done early warning to prevent the oil-whirl failure occurs, reduce the failure rate of the equipment, improve the accuracy of fault diagnosis, as soon as possible propose appropriate maintenance decisions.