Enhancing Stress Corrosion Cracking Feature Identification Using EMAT Data and Field Learnings

Abstract As data-driven decision making is becoming more critical in managing the integrity of natural gas transmission pipelines, this paper examines statistics on the identification of Stress Corrosion Cracking (SCC) using field excavations. The main tool used to detect SCC is the Electromagnetic Acoustic Transducer (EMAT) In-line Inspection (ILI). This tool can detect many other less injurious stable features such as lack of fusion pipe mill anomalies. This paper provides novel data driven techniques to help operators differentiate non-injurious cracks and SCC from field learnings. These techniques include a general statistical model, an SME Based Logistic Regression Model, and a Data Driven Machine Learning Logistic Regression model for pipelines susceptible to high pH intragranular SCC, and near neutral pH trangranular SCC separately. Each technique provides a probability that an EMAT call is a true SCC anomaly. An impact assessment on each of these techniques is provided around quantitative risk assessments and risk informed decision making. These insights can be used for risk-informed decision making on Stress Corrosion Cracking Direct Assessments by incorporating the requirements around soil data and minimum Charpy V-Notch Toughness from the Mega Rules RIN-2.