Research On Prediction Of Aircraft Engine Exhaust Gas Temperature Margin

The safety and efficiency of aircraft engines have always been the highest priority for airlines.In daily flight,the degradation of engine performance is inevitable.The degradation will cause a sharp increase in costs.The best way to improve the economic benefits is to recover the performance when the engine performance drops to a certain extent.In order to find the accurate threshold of engine performance degradation,various methods,such as deep learning methods and multiple linear regression methods,have been proposed.This paper analyzes the most important performance parameter of the aircraft engine,the engine exhaust temperature.Through different methods,a prediction model of the engine exhaust temperature is proposed.The innovations include the following parts:(1)Aiming at the time series characteristics of engine exhaust gas temperature,the auto-regressive moving average model is used to predict and analyze the engine exhaust gas temperature data.In this paper,wavelet method is used.Wavelet analysis can reveal the information in time domain and frequency domain,this property overcomes the basic shortcomings of Fourier analysis and provides a comprehensive description of the characteristics of non-stationary process.Then we use the auto-regressive moving average model to fit the trend of the data,and finally predict the development trend of the data.(2)A new prediction method of engine exhaust gas temperature,unscented particle filter,is introduced.Assuming that the development of engine exhaust gas temperature is non-linear and non-Gaussian distribution,particle filter is for this case.Particle filter is based on sequential importance sampling.It implements recursive Bayesian filtering through Monte Carlo simulation.The idea is to use a series of random samples to represent posterior probability density.The state estimation is obtained.However,the traditional particle filter method does not take into account the latest system observation value,which makes the effect of the estimation become very poor if the state value changes violently in a short time.Therefore,the unscented Kalman filter method is proposed to determine the distribution of particles in each time iteration to replace the random value of particles in a given range.In the case of great changes in state value,the sample can still get good prediction effect,which has great practical significance.Given the threshold of engine exhaust gas temperature,the probability distribution of prediction results is provided by particle filter algorithm,which is used to predict the state of nonlinear and non-Gaussian systems,and then the number of flight cycles reaching the threshold is known.