Pipeline Corrosion Pits Growth Prediction with Probabilistic Model and Artificial Neural Network

A scheme based on two historical in-line inspection data sets was developed to estimate the evolution of pitting depth caused by internal corrosion in pipelines. This process involved carefully aligning inspection pitting data and modeling pitting growth predictions using Bayes’ theorem. Only the most basic mathematical models were utilized; hence, the numerical solution was achieved through a carefully designed computer code. No further mathematical derivation was conducted to reach an analytic solution, thereby avoiding any approximations, which meets the preference of pipeline engineers. Furthermore, artificial neural networks were utilized to create a strong correlation between the average corrosion rate and the slope of each pipeline joint, enabling the operator to manage the pipeline with reduced inspection lengths. The scheme developed considers individual pipeline joints as the fundamental unit for assessing corrosion evolution. This joint-based approach not only allows accurate estimations but also facilitates the excavation processes necessary for effective present pipeline integrity management based on historical inspection data sets.