A multi-level classification model for corrosion defects in oil and gas pipelines using meta-learner ensemble (MLE) techniques

Maintaining the integrity of oil and gas pipelines is necessary for the efficient and safe transport of hydrocarbons. Corrosion defects can lead to decreased operational efficiency, leaks, a reduction in operational efficiency, and even catastrophic pipeline failures. Machine learning techniques are useful in detecting corrosion defects, ensemble methods that combine multiple classifiers offer better predictive accuracy. The aim of this work is to develop multi-level classification model an efficient ensemble technique capable of predicting the level of corrosion defects and addressing class imbalances in oil and gas pipeline data. The study uses a two-level stacking ensemble learning method that enhances corrosion defect prediction called the meta-learner ensemble (MLE). The model classifies corrosion defects into three categories: high, medium, and low. Prediction accuracy was improved by using a stacking classifier that combines multiple basic classifiers with a logistic regression meta-learner. Findings show that most corrosion defects fall within the low-risk category, with a significant number falling into the medium-to-high-risk range, highlighting the necessity for targeted maintenance. Considering the challenges of dataset imbalance, the stacking classifier achieved 94% accuracy, showing balanced performance across all risk categories. The stacking model outperformed the random forest and logistic regression models in terms of F1-scores, precision, and recall. This study demonstrates the application of stacking ensemble techniques in predicting corrosion risks and optimizing pipeline maintenance strategies. It provides vital information for improving pipeline safety and optimizing predictive maintenance practices by providing an in-depth assessment of various machine learning models, especially when real-time monitoring systems are integrated.