Simplified Bayesian Network Model for Stress Corrosion Cracking

Abstract Stress corrosion cracking (SCC) occurs due to specific combination of environmental, loading, and materials factors. Modeling of SCC requires considering a wide range of environmental and mechanical considerations, which results in models that are complex and difficult to use. The present study presents the simplification of a Bayesian SCC model that was previously validated with historical SCC rates in the field. Inputs of the model have been grouped into three categories (exposure mitigation and resistance) and results (probability of SCC occurrence and SCC rates) have been grouped into 27 combinations allowing for a partial validation of near neural and high pH stress corrosion cracking (SCC) model. This study provides a simple set of 27 reliable and simple SCC crack growth rate distributions as a function of pipeline exposure/environment, SCC mitigation/control, and crack growth resistance generated by the original SCC model. Additionally, the SCC rates can be easily updated using in-line inspection (ILI) data, allowing operators to use reduced uncertainty in predicted crack sizes. The combination of these two methods allows operators to quickly obtain expected crack sizes from knowledge of key input conditions, and update predictions with data where it is available without the need to use a complex Bayesian network model.