A data-driven approach for health status assessment and remaining useful life prediction of aero-engine

Abstract

Artificial Intelligence (AI) algorithms are widely used to improve the health monitoring tools exploited in aeronautics and this result in an increase in reliability and flight safety. Furthermore, better management of maintenance costs is another positive consequence of the application of AI-based health monitoring tools. In this paper, two AI-based health monitoring tools are developed to predict the Remaining Useful Life (RUL) of a fleet of simple turbojet engines VIPER 632 – 43 subject to a compressor degradation process. The AI algorithms used in this work are the Nonlinear Autoregressive Network with Exogenous Inputs (NARX) and the Long Short-Term Memory (LSTM) neural networks, which are two types of Artificial Neural Network (ANN) particularly suitable for time-series forecasting. The data about engine operation in degraded condition necessary to develop the just cited tools are obtained by performing a series of simulations in transient condition in which a degraded state at the compressor is implemented. The Matlab&Simulink software, equipped with the T-MATS Simulink library developed by NASA is used to develop a virtual model of the VIPER 632 – 43 engine and to perform the simulation. Furthermore, an adequate mathematical model is used to simulate the trend of the degradation level during time.