MACHINE LEARNING-BASED PREDICTION OF EXTERNAL CORROSION RATES IN BURIED UNPROTECTED GAS PIPELINES
External corrosion remains a critical threat to the long-term durability of buried gas pipelines, especially under varying environmental conditions such as soil pH, moisture content, and relative humidity. This study presents a predictive framework for estimating external corrosion growth rates along the Ajaokuta�Kaduna�Kano (AKK) gas pipeline route in Nigeria. A combination of Computational Fluid Dynamics (CFD) simulations and supervised Machine Learning (ML) models was employed to capture and learn from environmental variability across five locations�Ajaokuta, Lokoja, Abaji, Kaduna, and Kano. Input features included soil properties, climate data, and corrosion history. Among the five tested ML algorithms, the XGBoost model exhibited the best predictive performance, achieving an R� score of 0.95 and a Mean Absolute Error (MAE) of 0.0065 mm/yr in Kano. Key predictors were identified as soil pH, moisture content, and relative humidity. This research delivers a novel, location-sensitive approach to external corrosion prediction and offers a reliable basis for geo-targeted maintenance planning in Nigeria�s oil and gas sector.
Authors : Waziri, A.M., Guma, T.N., Akindapo, J.O. and Orueri, D.U.
Category : Open Access Volume (Issue) : 11(2) Date Uploaded : 23rd July 2025