Research On Remaining Useful Life Prediction Of Aeroengine

As a core component of aircraft,the health status of the aero-engine has a vital impact on the economic efficiency,availability and safety of aircrafts.The high temperature and high pressure working environment makes the degradation of engine performance into a normal state.If engines are not maintained in time,they may even cause flight accidents.The remaining useful life(RUL)prediction of engines is considered to be one of the most effective methods to evaluate the health status of aero-engines and avoid failure,which has important research significance.This paper mainly studies the application of data-driven methods in aero-engine RUL prediction and validates them on the C-MAPSS data set exposed by NASA.The main contents include:(1)Aiming at the multi-source sensor data in the engine degradation dataset,principal component analysis(PCA)is used to extract key features and screen out information unrelated to performance degradation;the variational moving average method is used to filter the principal component data to reduce the impact of noise;the optimization method is used to minimize the error between the engine failure state and the failure threshold of the training set,and a unified engine health index(Health)is constructed Index,HI)for subsequent RUL predictions.(2)Aiming at the situation that the similarity-based method directly uses the RUL of similar engines without considering the degradation difference between engines,the echo state network(ESN)is used to predict the degradation trajectory of engine health index to improve the similarity-based method,and the corresponding RUL is obtained by comparing the degradation trajectory and the failure threshold.At the same time,in order to solve the problem of choosing ESN parameters,genetic algorithms(GA)are introduced to optimize the parameters and reduce the influence of human factors.The method is verified on the C-MAPSS data set,and the method is proved to have a high prediction accuracy by scoring function and root mean square error(RMSE).(3)Aiming at the problem that machine learning algorithms are prone to overfitting,which leads to the poor effect of the algorithm on the test dataset,the adaptive boosting(Ada Boost)algorithm is used to improve the performance of wavelet neural network(WNN)and support vector regression(SVR)models.Thus,the Ada-WNN,Ada-SVR and their ensemble prediction model are obtained.The simulation results on the C-MAPSS dataset show that the fusion model is more accurate than the single model.