Online prognostics of aircraft turbine engine component's remaining useful life (RUL)

Abstract

Prognostics refers to the estimation of remaining useful life (RUL) of components of a system after a fault has been identified. Online prognostics indicates the estimation of RUL every time a new health data is provided to the user. In this paper, an artificial neural network (ANN) based approach is proposed for designing a prognostic system for aircraft turbine engine. A trained ANN is developed to estimate the health parameters such as component efficiency (η) and flow capacity (γ). The ANN was trained for a very small value of mean squared error (MSE). Then a forecasting (prediction) method is used to model the trend of estimated health parameters. The model is developed by autoregressive technique (AR) and all the data processing is done online. The proposed prognostic system also compute the distribution of the end of life (EoL) estimation of the failed component. The EoL and RUL estimation are implemented by modeling the health data using moving window and progressive window. The standard deviation (σ) of the distribution of estimated EoL indicates that progressive window performs better than the moving window with a σ reduction factor of 0.6 and 0.5 for η and γ respectively.