Attention mechanism in intelligent fault diagnosis of machinery: A review of technique and application. (August 2022)

• Record Type: Journal Article
• Authors: Lv, Haixin; Chen, Jinglong; Pan, Tongyang; Zhang, Tianci; Feng, Yong; Liu, Shen
• Is Part Of: Measurement. Volume 199(2022)
• Publication Date: 2022-08
• Subjects: Fault Diagnosis -- Attention mechanism -- Deep learning -- Fault classification; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111594
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: Attention Mechanism has become very popular in the field of mechanical fault diagnosis in recent years and has become an important technique for scholars to study and apply. The introduction of Attention Mechanism can help models achieve efficient resource allocation, improve the remote information capture capability of models, and significantly improve the performance of models for various equipment health management tasks (fault classification, life prediction, etc.) The application of Attention Mechanism in machinery has achieved fruitful research results, but there is a lack of related reviews. In order to facilitate later scholars to quickly grasp the Attention Mechanism and select the appropriate technique, this paper reviews the relevant research and applications of Attention Mechanism in Intelligent Fault Diagnosis of Machinery. Based on the methods proposed in the collected literature, this paper classifies and analyzes them from multiple perspectives to help readers grasp the development status and trends in this field. We divide the collected technologies into three categories: Recurrent-based, Convolution-based, and Self-attention-based. We describe each attention technique and its application scenarios in detail. Finally, we summarize the advantages and disadvantages of various AM techniques, and further discuss the possible future directions of attention mechanisms in the mechanistic field. The purpose of this paper is to provide a comprehensive reference for researchers and to help them find further research directions.