A comprehensive survey on applications of AI technologies to failure analysis of industrial systems

• Published in: Engineering failure analysis Vol. 148; p. 107172
• Main Authors: Bi, Siguo, Wang, Cong, Wu, Bochun, Hu, Shuyan, Huang, Wutao, Ni, Wei, Gong, Yi, Wang, Xin
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.06.2023
• Subjects: Artificial intelligence; Failure analysis; Internet-of-things; Internet-of-things; Artificial intelligence; Failure analysis
• ISSN: 1350-6307; 1873-1961
• DOI: 10.1016/j.engfailanal.2023.107172

•We recognize there yet exists a specialized survey on AI-aided failure analysis and thus investigate the existing literature.•The survey is on AI-aided reliability and failure analysis from the perspective of data, with the data structure highlighted.•We comprehensively point out a future research direction, based on the proposed frame of taxonomy. Component reliability plays a pivotal role in industrial systems, which are evolving with larger complexity and higher dimensionality of data. It is insufficient to ensure reliability and prevent failure based only on empiri- cal and parametric assumptions. Driven by huge amount of historical data, data- and statistics-based approaches aided by artificial intelligence (AI) are emerging as promising solutions. Especially, with the introduction to deep learning technology, the powerful ability of hierarchy representation is re- markably enhanced with deep cascaded layers. Furthermore, the demand for AI technology is high, and the applicability of the model in securing reliability, failure prediction and prevention in the industrial system is still nontrivial. Yet, there hardly exists such a systematic review of the AI-based approaches. In this survey, we provide a comprehensive overview of the AI- aided approaches to failure analysis in industrial systems, with sufficient or insufficient data, and imbalanced issues. We provide a concise introduction to the popular AI algorithms, classify the application scenarios of industrial systems into homogeneous or heterogeneous data-based scenarios, and review them respectively. We also summarize the resolved issues, challenges and promising directions.