A systematic review of data-driven approaches to fault diagnosis and early warning

• Published in: Journal of intelligent manufacturing Vol. 34; no. 8; pp. 3277 - 3304
• Main Authors: Jieyang, Peng, Kimmig, Andreas, Dongkun, Wang, Niu, Zhibin, Zhi, Fan, Jiahai, Wang, Liu, Xiufeng, Ovtcharova, Jivka
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.12.2023; Springer Nature B.V
• Subjects: Advanced manufacturing technologies; Bibliographic literature; Bibliographies; Big Data; Business and Management; Control; Data visualization; Fault diagnosis; Machines; Manufacturing; Manufacturing industry; Mechanical engineering; Mechatronics; Processes; Production; Robotics; Deep learning; Industry 4.0; Prognostics and health management; Industrial big data; Data visualization; Industrial Internet of Things
• ISSN: 0956-5515; 1572-8145
• DOI: 10.1007/s10845-022-02020-0

As an important stage of life cycle management, machinery PHM (prognostics and health management), an emerging subject in mechanical engineering, has seen a huge amount of research. Here the authors present a comprehensive overview that details previous and current efforts in PHM from an industrial big data perspective. The authors first analyze the historical development of industrial big data and its distinction from big data of other domains and summarize the sources, types, and processing modes of industrial big data. Then, the authors provide an overview of common representation and fusion (data pre-processing) methods of industrial big data. Next, the authors comprehensively review common PHM methods in the data-driven context, focusing on the application of deep learning. Finally, two industrial cases from our previous studies are included in this paper to demonstrate how the PHM technique may facilitate the manufacturing industry. Furthermore, a visual bibliography is developed for displaying current results of PHM in an appropriate theme. The bibliography is open source at “ https://mango-hund.github.io/ ”. The authors believe that future research endeavors will require an understanding of this previous work, and our efforts in this paper will make it possible to customize and integrate PHM systems quickly for a variety of applications.