QUANTITATIVE SYSTEM SPECIFIC LIKELIHOOD ALGORITHMS FOR SYSTEM WIDE RISK ASSESSMENT

摘要

The System Wide Risk Assessment (S WRA) is an essential first step in the pipeline integrity management program. It is required by both Canadian and US regulators and is expected to estimate risk due to all threats, interaction of threats, and consequences. The main objective of the SWRA is to identify high risk segments so that segments with excessive risk can be mitigated. The SWRA models developed in this study employs quantified likelihood models and consequence models. A companion paper explains the consequence models. This paper presents the framework and rationale used to produce quantifiable measures of likelihood for each threat. The quantification enables sensible comparisons between threat likelihood values and also enables realistic combining of likelihood values to produce total likelihood of failure due to all threats. It also facilitates identification of key parameters that contribute to each threat. It is important to have a consistent risk framework that systematically applies to all the threats and accommodates all the different aspects and mitigative actions in each threat management process. For effective continuous improvement it is essential that the models are transparent and updatable. A consistent framework that is systematic, rigorous, transparent and updatable is utilized with explicit consideration to threat interactions. The main advantages of the likelihood models developed in this study are: 1. It is based on all evidence that is available for each threat (failure histories, observations from assessments, i.e., digs, HTs, and ILIs, and mechanistic understanding) 2. It considers all nine threat categories and relevant subcategories where causal factors are different (such as SCC and Circumferential SCC within the crack threat category) 3. It clearly considers all three types of threat interactions (Interacting coincident defects, Interacting-activating threats, and Interacting common-mode conditions) among all threat categories. It is based on subsystem specific historical failure rates for each threat, where subsystem is defined as a subset of pipelines that have different performance characteristics with respect to at least one threat. This basis enables the failure frequencies predicted to be more in line with reality and consequently improves accuracy of predictions and appropriate quantification. 4. The subsystem specific historical failure rates are then calibrated to correlate to different mechanistic characteristics so that within-pipeline-subsystem variation due to changes in parameters is represented. 5. Finally assessments or observations are used to appropriately update threat likelihood with latest knowledge from measured local observations. All of the improvements mentioned above have helped the SWRA 2013 to produce more representative results. The comprehensive set of validation exercises verify that the results are realistic.