Hydrogen-assisted fatigue crack growth in pipeline steels: a machine learning approach

This study presents a machine learning methodology aimed at predicting the fatigue crack growth rate and examining the impact of alloying elements on hydrogen environment embrittlement in pipeline steels. The research marks the inaugural exploration of such predictive techniques in this specific domain. A predictive model was constructed using the bagged trees method, chosen for its ability to minimize metric errors like Mean Squared Error and Root Mean Squared Error. The accuracy of the model was confirmed, demonstrating its reliability in predicting outcomes effectively. Feature importance scores reveal that hydrogen pressure is a significant environmental factor, aligning with expectations. Moreover, the analysis identifies Mn and S as primary features that exert a strong negative influence on hydrogen environment embrittlement in pipeline steels. The machine learning approach is anticipated to be valuable for comprehending the behavior of pipeline steel under hydrogenous conditions and can be applied to enhance fatigue resistance in hydrogen applications.