Boosting RUL Prediction Using a Hybrid Deep CNN-BLSTM Architecture

Reliable estimation of remaining useful life (RUL) is a critical challenge in prognostics and health management (PHM), enabling the industry to better schedule future maintenance operations and reduce overhead costs and time linked to unnecessary maintenance operations. We notice that some efficient hybrid deep learning (DL) models have recently been proposed for performing RUL estimation and prediction. These novel techniques focus on combining several machine learning techniques to leverage the power of different models, especially in this paper a new hybrid method that blends convolutional neural network (CNN) and bi-directional long short-term memory (BLSTM) to extract spatial and temporal features. The experiments of our approach on the C-MAPSS dataset show the relevance of the proposed hybrid DL method, since it outperforms the results of many proposed models in the RUL prediction literature.