PREDICTIVE MODELING OF HYDROGEN-INDUCED CRACKING (HIC) IN X80 GRADE PIPELINE STEELS USING MACHINE LEARNING INTEGRATED WITH METALLURGICAL PARAMETERS

Abstract

Hydrogen-induced cracking (HIC) is a critical integrity threat in API X80 pipeline steels operating in sour environments because it initiates internally at hydrogen trap sites and progresses through stepwise crack linking. With increasing sour-service exposure and the emerging transition toward hydrogen-blended transport systems, there is a growing need for predictive tools that can estimate HIC susceptibility before failure occurs. This study develops an interpretable machine learning framework for predicting HIC behavior in X80 steels by integrating sour-environment variables with metallurgical parameters that control hydrogen uptake, trapping, and crack propagation. The dataset combines key descriptors of metallurgical heterogeneity (segregation index, banding index, MA fraction), inclusion population (density and maximum inclusion size), localized hardness (maximum HV), and environmental severity (pH and H₂S partial pressure). HIC response is quantified using standard indices (CLR, CTR, CSR) and a susceptibility classification scheme. Results show that HIC response is strongly nonlinear, and ensemble machine learning models outperform baseline approaches for predicting CLR and classifying susceptibility. Metallurgical pattern analysis confirms that HIC severity increases with inclusion density and inclusion size, segregation and banding intensity, MA fraction, and hardness peaks, particularly under higher H₂S partial pressure and lower pH. Explainability analysis ranks sour severity, segregation-banding structure, inclusion metrics, and hardness localization as the most influential predictors, reflecting the mechanistic pathway of hydrogen entry, trapping, crack initiation, and crack linking. The framework provides a practical basis for steel quality assurance and risk-based inspection planning by enabling early prediction of HIC risk in X80 pipeline infrastructure under sour and hydrogen-transition service conditions.