Hybrid Architecture of Deep Convolutional Variational Auto-encoder for Remaining useful Life Prediction

Abstract

The remaining useful life prediction is a key element in decision-making and maintenance strategies development. Therefore, in practical situation, it is usually affected by uncertainty. The aim of this work is hence to propose a deep learning method which predicts when an in-service machine will fail to overcome the latter problem. It is based on deep convolutional variational autoencoder (CVAE). The proposed approach is validated using the C-MAPSS dataset of the aero-engine. The model’s classification performance has reached a superior accuracy compared to existing models and it is used for machine failure prediction in different time windows.