Cheng, Daosheng (2009) Uncertainty analysis of large risk assessment models with applications to the railway safety and standards board safety risk model. PhD thesis, University Of Strathclyde.
Probabilistic risk analysis aims to assess the safety risk of a system so that actions can then be taken to improve safety. Uncertainty however always exists in modelling. For more informed decision making, uncertainty in the outputs of the model must be assessed through uncertainty analysis. This research focuses on parameter uncertainty of a risk model composed of fault trees and event trees. Research questions include: (1) how to model the subjective uncertainty in the basic events and the consequences; (2) how to propagate the uncertainty in the input parameters through fault trees and event trees to obtain uncertainty in the output. Structured approaches are developed to elicit the covariance matrix of the basic events and to model dependence among the consequences. To calculate the uncertainty propagation, a model is developed to mimic fault trees and event trees; an analytical solution and a simulation-based method are developed for assessing the uncertainty propagation, which are implemented independently and therefore crosscheck each other. The developments can be used for subjective uncertainty assessment of Fault-tree and Event-tree models. With the developed methods, a reasonable elicitation workload is required to model the subjective uncertainty in the input parameters; the assessments can be monitored during the elicitation process. The methods for assessing the uncertainty in the output can work efficiently for large fault trees and event trees. Two case studies have been conducted with the Safety Risk Model (SRM) developed by Rail Safety and Standard Board (RSSB), UK. In the two case studies, the developed methods are deployed and experts were confident in making the required assessments. The feasibility of the developments is validated by the case studies.
Actions (login required)