Probabilistic Assessment Procedure for Crack-Like Flaws

A statistical sampling technique was developed to perform a probabilistic assessment of crack-like flaws consistent with the procedures defined in API RP-579, Fitness-For-Service. Four random input variables were included in the assessment procedure. This included fracture toughness, flaw size, applied stress and residual welding stress. A statistical distribution was defined for each of these variables based on available data and previous studies on how best to define the variability of these input variables. A statistical sampling of these input variables was generated using computer code specifically developed for this purpose. The statistically defined sampling of input variables was evaluated using the VCE SAGE software modular for evaluating crack-like flaws. The output file from this evaluation consisted of a normalized brittle fracture term, Kr, and a normalized plastic collapse term, Lr. This output was statistically evaluated using specifically developed computer code to calculate a probability of fracture. This probability of failure is based on a limit state function defined by the fracture assessment diagram (FAD), which was statistically adjusted by wide-plate test results to more accurately reflect the actual probability of failure. This assessment procedure was developed to calculate a probability of failure in order to define the risk of fracture due to crack-like flaws consistent with the ExxonMobil risk matrix which is used to manage all risk-based decisions in the Corporation. Future work will be conducted to develop similar probabilistic assessment procedures for other forms of degradation, such as creep, high temperature hydrogen attack, thinning and local metal loss.