AI and Machine Learning Framework Development for Future AC Interference Induced Corrosion Monitoring and Prediction

Abstract This paper builds upon findings from the 2025 AMPP paper Field Investigation into AC Corrosion at Coating Defects Simulated on 3LPE and HBPU Coating Systems in Low Resistance Sabkha Soils, which documented corrosion depth, current density variation, coating disbondment, and corrosion product formation (e.g., goethite, magnetite) at simulated coating defects. The original work revealed the critical influence of coating type, defect geometry, and local Joule-induced heating on corrosion mechanisms, particularly the correlation between thermal effects, oxide layer formation, and increased spread resistance. Recognizing the limitations of conventional prediction methods in complex, low-resistivity environments, this paper proposes the development of a structured, AI-driven framework to enable more accurate analysis and long-term prediction of AC interference-induced corrosion. The framework leverages supervised learning models trained on variables such as iac/idc ratios, impedance shifts, Joule-induced temperature changes, and XRD-verified corrosion profiles. It also explores unsupervised techniques—clustering and principal component analysis—to identify high-risk defect patterns. Integration of FEM simulation data and live sensor feeds is examined to enhance resolution and field applicability. This evolving framework lays the foundation for a future field trial and aims to support smarter inspection planning, real-time risk assessment, and mitigation strategies across critical energy infrastructure.