259

Syed Aijaz1, Michael Gloven2, USA.2

1TC Energy, Houston, USA. 2Pipeline-Risk, Denver, USA

Stress Corrosion Cracking (SCC) poses a significant threat to pipeline integrity, and Electromagnetic Acoustic Transducer (EMAT) tools are widely used to detect cracks as part of integrity assessments. However, EMAT inspections can be constrained by factors such as inspection frequency, tool availability, in-line inspection (ILI) tool performance, and operational challenges. In scenarios where alternative assessment methods—such as SCC Direct Assessment (SCCDA) or hydrostatic pressure testing—are not feasible, predictive modeling and machine learning offer a valuable opportunity to support SCC threat management. By leveraging data collected from pipeline digs, ILI runs, and environmental factors, these models can help prioritize future EMAT inspections and identify susceptible pipeline segments with the highest likelihood of SCC, enabling more targeted and efficient SCC threat management.

This paper presents TC Energy’s collaboration with Pipeline-Risk (PLR) to develop a classification-based machine learning model aimed at predicting SCC likelihood across TC Energy’s pipeline assets. The model integrates a wide array of data, including Magnetic Flux Leakage (MFL) inspection results, dig data (covering defects and coating), and environmental variables such as soil texture, temperature, and precipitation. The model was validated on a subset of pipeline systems and demonstrated strong predictive performance. This work highlights the value of machine learning in enabling data-driven threat assessments, offering insights into SCC-susceptible segments and supporting risk-informed decision-making.

Read/Download PaperGradeAvailable after the event