Probabilistic Model for Stress Corrosion Cracking of Underground Pipelines Using Bayesian Networks

Abstract

Abstract Stress corrosion cracking (SCC) continues to be a safety concern, mainly because it can remain undetected before a major pipeline failure occurs. SCC processes involve complex interactions between metallurgy, stress, external soil environment, and electrolyte chemistry beneath disbonded coatings. For these reasons, assessing SCC failure probability at any given location on a pipeline is difficult. In addition, the uncertainty in data makes the prediction of SCC challenging. The complex interactions that affect SCC failure probability can be modeled using Bayesian network models. The Bayesian network models link events by cause-consequence connections. The strengths of these connections are adjusted using expert knowledge, analytical models, and data from the field. An approach to predicting probability of High pH SCC failure using Bayesian networks was proposed in a previous publication.1 The previous paper discussed the effects of stress only. In this paper, the previously discussed model is extended to the evaluation of other factors that affect high pH SCC. The model can be used to assess the probability of failure due to SCC at different times for different pipeline segments.