The Research For Exhaust Temperature Anomaly Detection In Gas Turbine

With advances in technology, as gas turbine is compact, quick start, smoothoperation, high thermal efficiency, it has become one of the core power equipmentin the21st century. Anomaly detection and fault diagnosis technology is animportant means to guarantee the economy and security of gas turbine. Poorconditions of combustion operation, failure-prone, and once the damage seriousconsequences, thus enhancing the anomaly detection and fault diagnosis of thecombustion chamber has a very important significance.Anomaly detection is implemented by monitoring the changes the turbine exitgas turbine exhaust temperature. In the gas turbine exhaust temperature distribution,the temperature of each measuring point theoretically should be uniform, butbecause of the initial installation errors, the combustion chamber of each case isdifferent, the temperature of each point there are some inherent differences. Thisdifference reflects the inherent characteristics of the device, making the gas turbineexhaust temperature distribution have some frequent mode (ie, intrinsic mode), justlike a fingerprint device. As long as no structural changes, the frequent mode is thedevice’s identity label which will not change. Contents of this paper is to extractfrequent mode which don’t affect working conditions and environmental changes.The mode enables the detection of the gas turbine exhaust temperature anomalyevolution.Based on the principle of gas turbine, a model is established to reflect thedistribution of the gas turbine exhaust temperature reflecting inherent characteristics.Through simulation analysis, the law of variable conditions and failure to exhausttemperature distribution of the typical effects was achieved. Theoretically it provedthe existence of frequent mode exhaust temperature distribution. Since theestablishment of the exhaust temperature distribution theoretically frequent modehas some limitations, a multiple linear regression model among exhaust temperatureand average temperature, air inlet temperature, gas turbine generator outlet pres surewas established by useing a gas turbine actual data. Based on this model, thefrequent mode of the gas turbine exhaust temperature distribution was extracted andprove. Anomaly detection algorithm based on distance discrimination, improvementof quality control chart and the modified Bias were studied. All of them can monitorthe gas turbine exhaust temperature anomalies in the evolution of a timely warning.Advantages and disadvantages of the three detection methods were analyzed.