| As the main dynamic source of aerospace systems,aero-engine is extremely important for its safety and reliability.Accurate prediction of the remaining life of the engine can effectively avoid accidents,which is of great significance for ensuring the normal operation of the aerospace system.Since the aero-engine may be affected by a certain fault during the degradation process,the occurrence of the fault will not immediately make the system fail,but will accelerate the engine’s degradation rate,thereby shortening its remaining life.However,the current research on the prediction of the remaining life of aeroengines does not consider the impact of the fault,but ignores or mostly defines the fault as a failure,which will lead to a significant reduction in the accuracy of the remaining life prediction of the aero-engine and the level of health management.In view of the shortcomings of existing research methods,the main innovations and contents of this paper are as follows:(1)In order to describe the impact of the fault on the aero-engine degradation process,the remaining life prediction based on the Wiener process under normal conditions is studied.In order to solve the individual difference problem of the engine,the corresponding remaining life distribution is obtained using the full probability formula.At the same time,in order to solve the problem of engine online life prediction,this paper uses a combination of maximum likelihood estimation and Bayesian theory to finish the parameter estimation.Then,the experimental dataset provided by NASA is analyzed in detail,and the Moving average and Kalman filtering are used to denoise the data.Finally,it is verified through the degraded dataset that the prediction method with the individual differences and parameter updates will improve the accuracy of the remaining life prediction.(2)Aiming at the problem that the fault may exist in the degradation process,the Wiener process is used to establish a degradation model considering fault effects.Two cases where the time of the fault is a fixed point and a random variable are discussed separately,and the corresponding remaining life distribution is derived.Aiming at the problem of parameter estimation when the fault occurrence moment is definite,an SVM classification discriminant model based on particle swarm algorithm is established,and the maximum likelihood method is used to estimate the unknown parameters.Aiming at the problem of parameter estimation when the fault occurrence moment is a random variable,the EM algorithm is used to estimate the parameters of the degradation model.Through the verification analysis of the experimental dataset,compared with the prediction of the remaining life in the normal state,the prediction result of the remaining life of the engine considering fault effects is closer to the actual value.(3)Due to the occurrence of the fault,the degradation trajectory of multiple performance parameters will be changed,and there may be correlations between the parameters,so it is difficult to directly establish a multivariate degradation model.In order to conduct comparative experiments,on the basis of obtaining the single-parameter life distribution under the fault effects.At first,the multi-parameter independent residual life prediction with the fault effects is studied,and then for the problems related to multi-parameters,this paper uses the Copula function to obtain the joint life distribution under the influence of the fault.Compared with single-parameter and multi-parameter independent residual life prediction,the experimental results show that the multi-parameter correlation life prediction model considering fault effects has the highest prediction accuracy. |
