Remaining Useful Life Estimation Based on a New Convolutional and Recurrent Neural Network

Abstract

Remaining useful life (RUL) estimation is an important part of prognostic health management (PHM) technology. Traditional RUL estimation methods need to define thresholds with the help of experience, and the thresholds affect the precision of the test results. In this paper, a hybrid method of convolutional and recurrent neural network (CNN-RNN) is proposed for the RUL estimation. This method can accurately predict the RUL by using a trained hybrid network without setting a threshold. The prediction accuracy of the model is further improved by processing, clustering, and classifying the data. The proposed CNN-RNN hybrid model combines CNN and RNN, it can extract the local features and capture the degradation process. In order to show the effectiveness of the proposed approach, tests on the NASA Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset of turbofan engine. The experimental results show that the proposed CNN-RNN hybrid model achieves better score values than the Multilayer Perceptron (MLP), Support Vector Regression (SVR) and Convolutional Neural Network (CNN) on FDOOI, FD003 and FD004 data sets.