Crack fault diagnosis of rotating machine in nuclear power plant based on ensemble learning. (April 2022)

• Record Type: Journal Article
• Authors: Zhong, Xianping; Ban, Heng
• Is Part Of: Annals of nuclear energy. Volume 168(2022)
• Publication Date: 2022-04
• Subjects: Crack fault -- Rotating machine -- Machine learning -- Ensemble learning -- Noise and small data; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2021.108909
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Crack fault diagnosis of rotating machine is important for nuclear power plants. Two ensemble learning strategies, Bagging and Boosting, are presented. Two ensemble learning models are created and compared with a single model. The comparison demonstrates the effectiveness of ensemble learning. Abstract: Crack faults in rotating machines can cause machine shutdown or scrapping, endangering the normal operation and safety of nuclear power plants. Intelligent diagnostic techniques based on machine learning have the potential to diagnose crack faults. However, problems such as scarcity of field fault data and high noise of plant measurements pose challenges to the application of machine learning. This study proposes an ensemble learning approach to mitigate the negative impacts of the problems. Ensemble learning is a strategy for combining multiple machine learning models into a composite model. The basic idea of ensemble learning is that even if one model makes a mistake, other models can correct it. Case studies based on bearing and gear system fault experiments show that the proposed ensemble learning models have better diagnostic results than the single model in the presence of noise and small data.