Anomaly detection method based on gas turbine sensor associated network
The mechanism and characteristics of four typical abnormal forms of gas turbine were analyzed,and the mapping between the different abnormal features and the sensor network characteristics were acquired.On this basis,the abnormal characteristics of sensor network were obtained,and the detection strategy of gas turbine anomaly based on sensor correlation network was proposed.The results of experiments indicated that sensor related network model based on multi source information fusion could filter out the correlation between nodes when the correlation indicator was below the threshold of 0.37,and achieve steady-state anomaly detection of gas turbine effectively,showing that there was a non-linear trend of abnormal depression greater than 12 ％ during the speed up;although the correlation between nodes under normal circumstances should be a linear trend,this method could achieve dynamic anomaly detection of gas turbine effectively.
fault diagnosissteady-state anomaly detectiongas turbinedynamic anomaly detectionsensor association network
NETL
NSTL
万方数据
维普
