Remaining useful life estimation in prognostics using deep convolution neural networks

• Published in: Reliability engineering & system safety Vol. 172; pp. 1 - 11
• Main Authors: Li, Xiang, Ding, Qian, Sun, Jian-Qiao
• Format: Journal Article
• Language: English
• Published: Barking Elsevier Ltd 01.04.2018; Elsevier BV
• Subjects: Components; Convolution; Convolution neural network; Critical components; Data processing; Deep learning; Feature extraction; Information processing; Neural networks; Prognostics and health management; Reliability engineering; Remaining useful life; Sample preparation; Signal processing; Studies; Useful life; Windows (intervals); Deep learning; Remaining useful life; Convolution neural network; Prognostics and health management
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2017.11.021

•Propose a novel deep convolutional neural network-based method for remaining useful life predictions.•No prior expertise on prognostics and signal processing is required, that facilitates the application of the proposed method.•Effects of the key factors on the prognostic performance are widely investigated and the model parameters are optimized.•Experiments on a popular aero-engine degradation dataset (C-MAPSS) and comparisons with the related state-of-the-art results validate the effectiveness and superiority of the proposed method. Traditionally, system prognostics and health management (PHM) depends on sufficient prior knowledge of critical components degradation process in order to predict the remaining useful life (RUL). However, the accurate physical or expert models are not available in most cases. This paper proposes a new data-driven approach for prognostics using deep convolution neural networks (DCNN). Time window approach is employed for sample preparation in order for better feature extraction by DCNN. Raw collected data with normalization are directly used as inputs to the proposed network, and no prior expertise on prognostics and signal processing is required, that facilitates the application of the proposed method. In order to show the effectiveness of the proposed approach, experiments on the popular C-MAPSS dataset for aero-engine unit prognostics are carried out. High prognostic accuracy on the RUL estimation is achieved. The superiority of the proposed method is demonstrated by comparisons with other popular approaches and the state-of-the-art results on the same dataset. The results of this study suggest that the proposed data-driven prognostic method offers a new and promising approach.