A status-relevant blocks fusion approach for operational status monitoring. (November 2021)

• Record Type: Journal Article
• Authors: Gao, Fulin; Tan, Shuai; Shi, Hongbo; Mu, Zheng
• Is Part Of: Engineering applications of artificial intelligence. Volume 106(2021)
• Publication Date: 2021-11
• Subjects: Operational status monitoring -- Status-relevant blocks -- Conflict measurement and management -- DS evidence theory -- CNN -- LSTM -- Shield; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2021.104455
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Faced with massive data, one-off processing without distinction is not conducive to reducing data and model complexity, and many potentially valuable local details are ignored. The decision fusion of partition blocks, such as Dempster–Shafer (DS) evidence theory, opens the gate to differentiated analysis of data. However, DS evidence theory has problems that conflict factor K cannot accurately measure conflict and counter-intuitive results can be produced in response to high-conflict evidences. Therefore, a status-relevant blocks fusion approach for operational status monitoring is proposed in this research, with three main contributions: Firstly, a status-relevant modules and blocks selection strategy based on expert knowledge and wrapper model is proposed, which reduces data complexity and improves the ability to capture local effective information. Secondly, a personalized hybrid solution is designed to implement data compression, timing analysis, and models update to cope with the scale inconsistency, temporal correlation, and dynamic characteristics of equipment operating data. Finally, a conflict measurement and management strategy combining supervised and unsupervised (CMMS-SUS) is proposed to identify and revise high-conflict evidences from different blocks to eliminate the concerns of DS evidence theory. Then, Dempster's combination rule is used to fuse the low-conflict evidences without correction or the high-conflict evidences after correction. Furthermore, a realistic shield tunnel case in China is used to demonstrate the feasibility and effectiveness of this approach. Highlights: A novel partition strategy is developed for operational status monitoring. A hybrid solution is designed for data compression, timing analysis and model update. A approach named CMMS-SUS is proposed to identify and revise high-conflict evidences.