NEXT-GEN CORROSION ANALYTICS: INTELLIGENT FORECASTING FOR OIL PIPELINE INTEGRITY

Abstract

Effective monitoring and precise forecasting of internal CO₂ corrosion are paramount for safeguarding the structural integrity of oil pipelines. This research employs an array of advanced machine learning and deep learning methodologies to predict corrosion progression, leveraging a comprehensive dataset encapsulating critical environmental and operational factors. Various regression techniques, including Support Vector Regression (SVR), Decision Trees, Random Forest, K-Nearest Neighbours (KNN), and Polynomial Regression, were meticulously evaluated. Among linear models, the Random Forest Regressor exhibited superior predictive accuracy with an R² of 0.9936 and an MSE of 0.0080. Notably, third-degree polynomial regression emerged as the most effective nonlinear approach (R² = 0.9988). Regularized regression models underscored Ridge Regression as the optimal choice, achieving an R² of 0.9989. Within deep learning paradigms, the Feed-Forward Neural Network (FNN) and ResNet models demonstrated remarkable efficacy, attaining a R² score of 0.9928. Furthermore, ensemble learning, particularly a hybridized CNN-CNN Forecaster model, exhibited commendable predictive precision. The study emphasizes the necessity of model selection for accurate corrosion forecasting and facilitating proactive maintenance.