AI-Assisted CO2 Corrosion Forecasting Method for Pipeline Integrity Monitoring System

The safe operation of pipelines is a vital part of the oil and gas industry. This paper addresses the issue of predicting CO2-induced corrosion in pipeline walls, which is critical for maintaining pipeline integrity and preventing accidents. A new AI-assisted method for forecasting the corrosive wear of pipeline walls is presented. The data preprocessing procedure, including the temporal interpolation of infrequent pipe thickness measurements, is described. Machine learning models, particularly fully connected neural networks, are trained to predict the rate of change in pipe wall thickness. Experiments demonstrate the effectiveness of this approach, with a detailed evaluation of model performance over various time intervals. The results indicate that the proposed method provides reliable predictions, maintaining the average deviation of predictions from ground-truth values within the 0.86 mm range over periods up to 11 months.