Research On Aeroengine Forecasting Method Based On Improved SVM

Aircraft engine maintenance costs accounted for 40% of the cost of the flight maintenance,meanwhile spare aircraft engine is high value of millions of dollars,so the aircraft engine prediction research is of great significance.The airway component is the most critical part of the aeroengine,and its gas path performance parameter has an important influence on the engine life attenuation.Therefore,this paper proposes an improved SVM algorithm for the shortcomings of the traditional support vector machine(SVM),and applies it to the trend prediction and residual life prediction of air engine performance parameters.The main research contents are as follows:1?Support Vector Machine Improved Algorithm.SVM algorithm ignores the influence of input on output,which is easy to cause large accumulation error,and each kernel function has its own defects to limit its application.To solve the above problem,this paper uses functional analysis to establish a hybrid kernel generic function SVM algorithm.In order to obtain the optimal parameters of SVM and reduce the prediction error,the improved quantum particle swarm optimization algorithm is used to optimize the parameters.Through the analysis of the thrust trend prediction in the missile launch system,the improved SVM algorithm can obtain higher precision under certain training time.2?Preconditioning of Airway Performance Parameters of Aeroengine.First,the Grubbs criterion is used to determine the anomalous points of the gas performance parameters,and the test is carried out using the Romanovschi criterion.If the two methods determine the results as abnormal points,remove them.Then,the improved SVM algorithm is used to predict the abnormal points,and the predicted value is taken as the missing value.Finally,consider the use of rough punishment smoothing algorithm noise reduction,the penalty parameters are still small part of the characteristic signal is filtered out,the empirical mode decomposition and rough punishment smoothing method combined method for noise reduction.3 ? Research on Aeroengine Performance Trend Forecasting.Based on the improved SVM algorithm,the trend prediction of single engine performance parameters and multiple gas performance parameters of aeroengine is studied.The short-term trend forecast of single air-conditioning performance parameters isanalyzed,and the superiority of improved SVM algorithm is analyzed.The forecasting results of long-term trend forecasting of multiple air-conditioning performance parameters are available.Using improved SVM to forecast the trend prediction,the partial eigenvalues are guaranteed to ensure the correct forecast of the overall trend of air-conditioning performance parameters and meet the demand of aircraft engine's remaining life forecast.4?Research on Aero-Engine Residual Life Prediction for Single and Multiple Gas Performance Parameters.The second chapter improves the SVM prediction algorithm and establishes a single gas performance parameter prediction model with the failure threshold.Although the forecasting model is simple in modeling but there is poor forecasting information,it is necessary to set the fixed failure threshold and over-forecast.The life expectancy model of multiple airway performance parameters is to consider the influence of multiple performance parameters and the corresponding flight time on the remaining life of the aeroengine,and combine with the trend prediction model and the failure decision model.The residual life prediction model of multiple airway performance parameters is used to solve the shortcoming of the prediction model of the residual performance of the single airway performance,and the prediction accuracy is also improved greatly.