A Proposal for Improving Remaining Useful Life Prediction in Industrial Systems: A Deep Learning Approach

Abstract

Accurate prediction of the remaining useful life (RUL) of engineering systems provides decision-makers valuable information to apply more efficient maintenance programs maximizing the equipment usage and avoiding the increase of costs due to failures. In this area, deep learning methods have become increasingly popular because of their capability to learn complex and discriminative non-linear features that can facilitate the RUL prediction task. These network models are generally trained to minimize the mean square error (MSE) between the RUL prediction and its true value. This metric gives equal importance to the error at the beginning and at the end of a system’s useful life. However, the prediction of the RUL is more critical as a system approaches the end of its useful life. In this chapter, a performance metric for evaluating prognostic models is proposed with the objective of establishing a direct relation between RUL prediction and maintenance planning. In addition, a procedure to use this metric for training a recurrent neural network (RNN) is proposed to improve the network’s ability to learn the relationship between the raw data and the corresponding RUL, giving more importance to obtain accurate predictions as the system approaches to the end of its useful life. The procedure is applied to the NASA Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset. The satisfactory results confirm the validity of the proposal.