Optimal inspection planning of corroded pipelines using BN and GA

Abstract The corrosion phenomenon in pipelines is currently a serious problem in the petroleum industry, which can lead to small leak, burst, and instable rupture of pipelines. Pipeline inspection and maintenance should be conducted to prevent failures of corroded pipelines. The inspection and maintenance processes of a corroded pipeline are generally represented as a classical decision tree, which can clearly describe the decision problem. However, a decision model based on decision trees exponentially grows with inspection and maintenance number. The optimal inspection problem becomes complicated as the inspection and maintenance number increases. Numerous complicated calculations, such as probabilities of each decision tree branch, should be accomplished, and the optimal inspection plan should be efficiently selected from all decision plans. Bayesian network (BN) and genetic algorithm (GA) are applied to develop a framework for the inspection decision making for corroded pipelines in this study. The BN is used to model the corrosion degradation process of pipelines and the probabilities of each decision tree branch and update the models according to information gained from pipeline inspection or maintenance. The GA is used to rapidly determine the optimal inspection plan from all inspection plans. A comprehensive algorithm based on BN and GA is proposed to solve the optimal inspection planning problem based on classical decision trees. The methodology is demonstrated by its application to the optimal inspection planning of a corroded pipeline. It turns out that the proposed methodology can accurately and efficiently solve the optimal inspection problem.