Remaining Useful Life Prediction of a Turbofan Engine Using Deep Layer Recurrent Neural Networks

Abstract

Turbofan engine is a pivotal component of the aircraft. Engine components are susceptible to degradation over the life of their operation which affects the reliability and performance of an engine. In order to direct the necessary maintenance behavior, remaining useful life prediction is the key. This thesis presents a prediction framework for the Remaining Useful Life (RUL) of an aircraft engine using the whole life cycle data and deterioration parameter data based on a machine learning (ML) approach. In specific, a Deep Layer Recurrent Neural Network (DL-RNN) model is proposed to address the problem of prognostic instability based on deep learning. In addition, for the aircraft engine, a new health indicator (HI) measure is implemented based on the preprocessing of raw data. The proposed method is compared against Multilayer-Perceptron (MLP), Non-linear Auto Regressive Network with Exogenous Inputs (NARX), Cascade Forward Neural Network (CFNN) and validated through the IEEE 2008 Prognostics and Health Management (PHM) conference Challenge dataset and Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset provided by NASA results reveal a better predictive precision with respect to other ML algorithms.