期刊
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
卷 168, 期 -, 页码 -出版社
ACADEMIC PRESS LTD- ELSEVIER SCIENCE LTD
DOI: 10.1016/j.ymssp.2021.108653
关键词
Prognostics and health management; Graph neural networks; Intelligent fault diagnostics and prognostics; Practical guideline; Benchmark results
资金
- Natural Science Foundation of China [52175116]
- Major Research Program of Natural Science Foundation of China [92060302]
- Shaanxi province 2020 natural science basic research plan [2020JQ-042]
- National Key Science and Technology Infrastructure Opening Project Fund for Research and Evaluation facilities for Service Safety of Major Engineering Materials, Aeronautical Science Foundation [2019ZB070001, 20200046070002]
- National Science and Tech-nology Major Project [J2019-IV-0018-0086]
- Fundamental Research Funds for the Central Universities
Deep learning methods have advanced the field of Prognostics and Health Management, but handling irregular data in non-Euclidean space remains a challenge. Research has proposed a practical guideline for utilizing graph neural networks for intelligent fault diagnostics and prognostics, and established a framework based on GNN for this purpose.
Deep learning (DL)-based methods have advanced the field of Prognostics and Health Manage-ment (PHM) in recent years, because of their powerful feature representation ability. The data in PHM are typically regular data represented in the Euclidean space. Nevertheless, there are an increasing number of applications that consider the relationships and interdependencies of data and represent the data in the form of graphs. Such kind of irregular data in non-Euclidean space pose a huge challenge to the existing DL-based methods, making some important operations (e.g., convolutions) easily applied to Euclidean space but difficult to model graph data in non-Euclidean space. Recently, graph neural networks (GNNs), as the emerging neural networks, have been utilized to model and analyze the graph data. However, there still lacks a guideline on leveraging GNNs for realizing intelligent fault diagnostics and prognostics. To fill this research gap, a practical guideline is proposed in this paper, and a novel intelligent fault diagnostics and prog-nostics framework based on GNN is established to illustrate how the proposed guideline works. In this framework, three types of graph construction methods are provided, and seven kinds of graph convolutional networks (GCNs) with four different graph pooling methods are investigated. To afford benchmark results for helping further study, a comprehensive evaluation of these models is performed on eight datasets, including six fault diagnosis datasets and two prognosis datasets. Finally, four issues related to the performance of GCNs are discussed and potential research di-rections are provided. The code library is available at: https://github.com/HazeDT/ PHMGNNBenchmark.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据