Probability of rupture model for corroded pipelines

A log-logistic model is developed to predict the probability of rupture (POR) given burst of a pressurized steel pipeline at a naturally-occurring part-through wall corrosion defect. The predictors of the model are the defect depth normalized by the pipe wall thickness and defect length normalized by the square root of the product of the pipe diameter and wall thickness. A set of 140 full-scale burst tests of corroded thin-walled pipes are collected from the open literature to calibrate and validate the proposed POR model. The parameters of the POR model are evaluated using the maximum likelihood method. To facilitate deterministic defect assessments, a threshold POR of 0.35 is proposed such that defects with POR greater than or equal to 0.35 are designated as potential ruptures. The impact on the proposed POR model due to the measurement errors associated with defect sizes reported by the inspection tool is examined through sensitivity analyses.