Integrated pipeline corrosion growth modeling and reliability analysis using dynamic Bayesian network and parameter learning technique

Abstract

The present study integrates the corrosion growth modeling, reliability analysis and quantification of measurement errors of in-line inspection (ILI) tools in a single dynamic Bayesian network (DBN) model for the reliability-based corrosion management of oil and gas pipelines. The Expectation-Maximization algorithm in the context of the parameter learning technique is employed to learn the parameters of the DBN model. The application of the model on simulated and real-world corrosion data demonstrates the effectiveness of the parameter learning and accuracy of the corrosion growth predicted by the DBN model. In comparison with existing growth models, the integrating and graphical features of the developed model make the process of corrosion management more intuitive and transparent to users. The employment of the parameter learning technique provides an objective and convenient approach to elicit the probabilistic information from ILI and field measurement data. The above advantages make the model more amenable to the corrosion management practice in the pipeline industry.