Remaining Useful Life Prediction And Health Status Assessment Of Aero-engine

The safety, reliability and economy of aero-engines, as the core components of aircrafts, are always the focus of attention for manufacturers, repair factories and others relevant departments of aircraft. It is very important to predict the residual useful life of an aero-engine for its safety and economic benefits. The gas path is the key part of an aero-engine. Its failure rate is higher and is monitored mainly. The performance parameters of the gas-path part have more types and they are fluctuated largely and illegible trend. This paper use data mining and artificial intelligence algorithms to build a new methodology to the residual life predicting and health status evaluation and predicting model for the gas-path part. These models have considered the complex affection of time on wing and multi-parameters of gas-path. The main researches of this paper are as follows:This paper chooses a suitable data smoothing method based on analyzing the characteristic of gas-path parameters. In the case of the data characteristic is large fluctuation and illegible trend, Savitzky–Golay filter method and Exponential Smoothing method are applied on the data and the result shows that Savitzky–Golay filter method is better and more apposite for the pretreatment of gas-path parameters.The different consequences of prediction caused by the different kernel are compared in this paper. The suitable kernel is selected for the Support Vector Machine according to the performance parameter. After the decision of the kernel the optimal parameters of the Support Vector Machine are decided by applying Genetic Algorithms. The result shows that it can get a more accurate prediction to use the optimal parameters.The Support Vector Regression is employed to predict the parameters trend. The failure decision function considered the complex affection of time on wing and multi-parameters of gas-path is built by Support Vector Classification. The residual life predicting method is the combination of the predicting of the parameters and the failure decision function. A practical case is applied to illustrate the accuracy of this method.The self-adaption health status evaluation method is built with the Hidden Markov Model based on whole life data of aero-engine group. This method regards the multi-parameters of gas-path as the observation vector of Gaussian Mixture Hidden Markov Model to simulate the whole life degradation process of aero-engine. And a new health status predicting method is built with the combination of parameters trend predicting method.