Reliability Evaluation of Environmental Test Chambers Based on Bayesian Network

• Published in: Journal of failure analysis and prevention 2023-12, Vol.23 (6), p.2471-2488
• Main Authors: Yang, Huajian, Niu, Wangqiang, Han, Chengjing, Zhou, Xianwen, Gu, Wei
• Format: Article
• Language: English
• Subjects: Algorithms; Bayesian analysis; Characterization and Evaluation of Materials; Chemistry and Materials Science; Classical Mechanics; Corrosion and Coatings; Environmental testing; Failure; Failure modes; Failure rates; Failure times; Fault diagnosis; Fault tree analysis; Materials Science; MTBF; Original Research Article; Quality Control; Reliability; Reliability analysis; Safety and Risk; Solid Mechanics; Statistical analysis; System failures; Test chambers; Tribology
• ISSN: 1547-7029; 1728-5674; 1864-1245
• DOI: 10.1007/s11668-023-01753-1

Emphasis has been given to the reliability of a wide range of equipment in recent years. However, the reliability of environmental test chambers has received little research attention. In this paper, the Bayesian network (BN) model is proposed to evaluate the reliability of the chambers for the first time. This paper is an important supplement to reliability research on test chambers, and it proposes a novel reliability assessment approach that involves the following: (1) the maximum information coefficient is used to select components that are particularly important for system failures; (2) failure mode and effect analysis and fault tree analysis are used to initially establish the logical relationship between components; (3) max-min hill-climbing algorithm is used to learn the final BN structure; and (4) the Poisson process is introduced to calculate the failure probability of these components. Results show that BN describes the directional and quantitative dependencies between components in more detail than fault tree does. The failure rate predicted by BN is consistent with the actual data, and the determination coefficient ( R 2 ) is 0.6. At present, the mean time between failures of a single chamber is approximately 20 days. According to the reliability threshold, the cumulative failure times of the chambers can be predicted, and most of the forecast errors are within 6 days. When the chambers fail, the structural importance of the components is shown, and a reasonable fault diagnosis sequence is provided by the proposed BN model.