Physics-and Image-Informed Reliability Prediction for Pipelines Under Competing Failure Modes

Summary & ConclusionsCorrosion increasingly threatens the reliability of pipelines, causing leak or burst failures particularly in sections where defects are clustered. This work proposes a Physics- and Image-Informed Reliability Prediction (PIIRP) framework, a novel methodology that integrates rust imaging, Finite Element Method (FEM), spatial statistics, and a competing risk model to predict pipeline failures due to leak or burst. The framework uniquely uses image-based corrosion data to initialize a physics-informed, generation-wise Non-Homogeneous Poisson Process, where FEM-derived stress fields drive pit nucleation. By unifying these components, the PIIRP framework offers a realistic way to simulate corrosion evolution and explicitly model the two competing failure modes under industry standards like ASME B31G and DNV RP-F101. The results show that the approach reduces false burst alarms by 26% and indicates that densely clustered defects can cause the occurrence of a system failure approximately eight years earlier than cases with isolated defects. These findings, aligned with forensic evidence from the Enbridge Line 6B failure and guidelines like API 1160, highlight the framework’s value for targeted inspection planning.