Research Progress on Data-Driven Industrial Fault Diagnosis Methods

• Published in: Sensors (Basel, Switzerland) Vol. 25; no. 9; p. 2952
• Main Authors: Lei, Liang, Li, Weibin, Zhang, Shiwei, Wu, Changyuan, Yu, Hongxiang
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 07.05.2025; MDPI
• Subjects: Algorithms; Artificial intelligence; Big Data; Business operations; Cloud computing; Customer relationship management; Data analysis; Data collection; Data mining; Data science; Datasets; deep learning; Enterprise resource planning; Fault diagnosis; Fault location (Engineering); industrial big data; Industrial equipment; Industrial production; Information technology; Internet; Knowledge management; large language models; Maintenance and repair; Manufacturing; Methods; Production planning; Review; Sensors; Software; Supply chain management; China; deep learning; fault diagnosis; large language models; industrial big data
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s25092952

With the advent of Industry 5.0, fault diagnosis is playing an increasingly important role in routine equipment maintenance and condition monitoring. From the perspective of industrial big data, this paper systematically reviews the current mainstream industrial fault diagnosis methods. The content covers the main sources of industrial big data, commonly used datasets, and the construction of related platforms. In conjunction with the development of multi-source heterogeneous data, the paper explores the evolutionary path of fault diagnosis methods. Subsequently, it provides an in-depth analysis of data-driven fault diagnosis techniques in industrial applications, with particular emphasis on the pivotal role of deep learning algorithms in fault diagnosis. Next, it discusses the applications and development of large models in the field of fault diagnosis, focusing on their potential to enhance diagnostic intelligence and generalization under big data environments. Finally, the paper looks ahead to the future development of data-driven fault diagnosis methods, pointing out that data quality, interpretability of deep learning, and edge-based large models are important research directions that urgently require breakthroughs.