A methodology to clarify logical relationship among failure modes and determine system probabilities. (July 2021)

• Record Type: Journal Article
• Authors: Hu, Kun; Chen, Guohua; Abbassi, Rouzbeh; Huang, Kongxing; Zhou, Zhihang; Zeng, Tao
• Is Part Of: Journal of loss prevention in the process industries. Volume 71(2021)
• Publication Date: 2021-07
• Subjects: Logical relationship -- System probabilities -- LSE surfaces or curves -- Bayesian network; Chemical industries -- Safety measures -- Periodicals; 660.2804
• DOI: 10.1016/j.jlp.2021.104469
• Languages: English
• ISSNs: 0950-4230

Abstract

Abstract: In the vulnerability analysis, correlations among failure modes have significant effects on the estimation of failure probabilities. However, the failure modes were assumed to be independent with each other or only parts of dependencies of failure modes were considered, which might lead to inaccurate results. In the present study, a novel methodology to clarify the entire logical relationship among failure modes and determine system probabilities is developed. Firstly, based on the form-changed limit state equations (LSEs) of failure modes, the LSE surfaces or curves are plotted. Subsequently, the logical relationship among failure modes can be identified with the LSE surfaces or curves. The system consequences are further developed by the logical relationship. Bayesian network (BN) is constructed with the input of logical relationship into arcs. With BN considering logical relationship, the occurrence probabilities of failure modes are calculated and system probabilities are estimated more accurately, which are verified well with Monte Carlo simulation and analytical solution. Furthermore, the detailed compositions of occurrence probabilities of failure modes are specified by the system probabilities. The methodology is illustrated by a case study. This study can be applied to the vulnerability analysis of various hazards or disasters as long as LSEs for corresponding failure modes can be developed. Highlights: A methodology to clarify logical relationship among failure modes and determine system probabilities. Entire dependencies among failure modes are recognized in logical relationship. BN results considering logical relationship are more accurate and verified well. Detailed compositions of occurrence probabilities of failure modes are specified.