Leveraging the Power of the Combination of CNN and Bi-Directional LSTM Networks for Aircraft Engine RUL Estimation
Ikram Remadna, Labib Sadek Terrissa, Ryad Zemouri, Soheyb Ayad, Noureddine Zerhouni
Abstract
The Remaining Useful Life (RUL) is a crucial metric utilized within many industrial systems and defined as the time between the current instant after detection of the degradation and the moment when the degradation reaches the failure threshold. Its accurate prediction allows for scheduling the next maintenance decision in advance that decreases costs and time of maintenance by cancelling unnecessary maintenance. Capitalizing on Deep Learning (DL)'s recent success, this paper introduces a new hybrid RUL prediction approach that combines two DL methods sequentially. The hybrid model uses Convolutional Neural Network (CNN) with Bi-directional Long Short-Term Memory (BDLSTM) networks where CNN extracts spatial features while BDLSTM extracts temporal features. Our experimental verification carried out on the NASA Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset, and the results revealed that the proposed approach is superior to other machine learning models.