Incipient sensor fault diagnosis in multimode processes using conditionally independent Bayesian learning based recursive transformed component statistical analysis. (May 2019)

• Record Type: Journal Article
• Authors: Shang, Jun; Zhou, Donghua; Chen, Maoyin; Ji, Hongquan; Zhang, Hanwen
• Is Part Of: Journal of process control. Volume 77(2019)
• Page Start: 7
• Page End: 19
• Publication Date: 2019-05
• Subjects: Fault detection -- Fault isolation -- Incipient sensor fault -- Conditionally independent Bayesian learning -- Recursive transformed component statistical analysis; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2019.03.010
• Languages: English
• ISSNs: 0959-1524

Abstract

Highlights: A new multimode process monitoring method CIBL-RTCSA is proposed. CIBL is proposed for mode identification, with higher accuracy than traditional classification methods. A multiple RTCSA model with a window-switching scheme is developed for diagnosing the incipient sensor fault in multimode processes. A numerical example and simulation on a CSTH both demonstrate the effectiveness of the proposed method. Abstract: This paper considers the problem of detecting and isolating incipient sensor fault in multimode processes. A data-driven multimode process monitoring method called conditionally independent Bayesian learning based recursive transformed component statistical analysis (CIBL-RTCSA) is presented. Considering the strong assumption of conditional independence in naïve Bayes, orthogonal transformation is applied to measured variables to improve the extent of conditional independence in different operating modes. The Bayes-based mode identification is adopted for transformed data, and a multiple RTCSA model with a window-switching scheme is developed for monitoring multimode processes. With the orthogonal transformation, the accuracy of mode identification can be effectively improved compared with naïve Bayes. In addition, the fault detection and isolation performance of the proposed method outperforms traditional monitoring methods. The effectiveness of the proposed method is demonstrated by a numerical example and the simulation on a continuous stirred tank heater process.