Physics-Informed Bayesian Modeling for Integrity Assessment of Underground Pipeline Coatings (C2026-00220)

Sreelakshmi Sreeharan, Hui Wang, Yusheng Jiang, Reece Goldsberry, Homero Castaneda

Close-interval potential surveys (CIPS) are widely used to evaluate the effectiveness of cathodic protection systems on buried or submerged pipelines yet inferring coating condition from CIPS data remains challenging due to spatial heterogeneity and measurement noise. We introduce a physics-informed Bayesian model that provides a rigorous framework for assessing underground pipeline coating integrity in the presence of spatial heterogeneity and measurement noise. We present a hybrid approach that couples a physics-based transmission-line model of cathodic protection and coating behavior with a hierarchical Bayesian inference engine. The transmission-line model describes potential distributions along the soil pipeline interface, explicitly capturing variations in soil resistivity and coating defects. Embedding this forward model within a Bayesian framework allows us to assimilate noisy close-interval potential survey (CIPS) data and infer posterior distributions over coating impedance, yielding probabilistic maps of coating health rather than single-point estimates. We apply our method to real-world CIPS datasets, with qualitative validation performed using in-line inspection (ILI) data, direct current voltage gradient (DCVG) survey results, and known rectifier locations. By quantifying parameter uncertainty, the model supports risk-based maintenance prioritization and more informed integrity management under resource and operational constraints.