Deep learning-based remaining useful life estimation of bearings using multi-scale feature extraction. (February 2019)

• Record Type: Journal Article
• Authors: Li, Xiang; Zhang, Wei; Ding, Qian
• Is Part Of: Reliability engineering & system safety. Volume 182(2019)
• Page Start: 208
• Page End: 218
• Publication Date: 2019-02
• Subjects: Remaining useful life -- Prognostics and health management -- Deep learning -- Multi-scale feature extraction -- Rolling bearing; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2018.11.011
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: A novel deep learning architecture is proposed for prognostics using multi-scale feature extraction scheme. Machine remaining useful life during operation can be efficiently predicted based on the measured vibration signal. Little prior expertise on prognostics and signal processing is required, that facilitates the industrial application. Experiments on a popular rolling bearing degradation dataset validate the effectiveness and superiority of the proposed method. Sufficient labeled training data are required in real applications. Abstract: Accurate evaluation of machine degradation during long-time operation is of great importance. With the rapid development of modern industries, physical model is becoming less capable of describing sophisticated systems, and data-driven approaches have been widely developed. This paper proposes a novel intelligent remaining useful life (RUL) prediction method based on deep learning. The time-frequency domain information is explored for prognostics, and multi-scale feature extraction is implemented using convolutional neural networks. Experiments on a popular rolling bearing dataset prepared from the PRONOSTIA platform are carried out to show the effectiveness of the proposed method, and its superiority is demonstrated by the comparisons with other approaches. In general, high accuracy on the RUL prediction is achieved, and the proposed method is promising for industrial applications.