Research On Aeroengine Gas Path Parameter Deviation Model And Its Application

Aeroengine is regarded as the "heart" of aircraft. To enhance engine conditionmonitoring not only help to improve the transport aircraft safety, but also to reduceengine maintenance costs. There are many aeroengine condition monitoringtechnology. In all of them, the gas path parameter monitoring technology is veryimportant. This method get gas path parameter deviation through parameter deviationmodel. And then, get the engine performance status according to the trend of gas pathparameter deviation. At last, forecast gas path parameter deviation for "pre-diagnosis".Obviously, it is necessary to research on aeroengine gas path parameter deviationmodel and its application.In order to get the aeroengine gas path parameter deviation model, the deviationmodel should be broken down into three parts: the standardized model, the baselinemodel and the compensation model. Firstly, the engine condition monitoring system oforiginal engine manufacturer were analyzed, and the function expression of thebaseline model and the standardized model was obtained by experiment. And then, thecompensation model was established on Support vector machine method. Optimize theparameters of support vector machines with particle swarm optimization. Finally,actual data was used to validate the deviation model. The application shows that themodel error is small, and able to meet the actual requirements.The original value of engine gas path parameter deviation which SNR (Signal toNoise Ratio) is very low has many outlier. And the trends is difficult to observe. It,snecessary to detect outlier and smooth for the original value of deviation. First, outliersdetection algorithm is proposed on second exponential smoothing method. Then theevaluation indicator for the processing of deviation. optimization model for evaluationindicator to optimize the smoothing coefficient and significance level parameter. Andit’s solved with particle swarm optimization. Finally, the method was used to detectoutlier and smooth for the exhaust gas temperature deviation. The application showsthat the effect of this method is good.The prediction of gas path parameter deviation is essentially a time series prediction. First the time-varying fuzzy inference system theory was established. Andthen create a time-varying fuzzy neural network to learn the parameters of theinference system. The learning algorithm of the network structure was designed.Finally the Mackey-Glass chaotic time series prediction was used to prove the networkprediction accuracy.According to the demand of Air China Limited, system architectural frameworkof engine condition monitoring system was designed. The system is divided into sixlayer system. The system business function was designed. The system was divided intosix functional modules. The system architecture design was completed. The automaticoperation function module was placed in an application server. This system has beenonline application in Air China. The application results indicate that this system notonly can well and concisely complete the engine condition monitoring, but also caneffectively improve the work efficiency of airline.