Interpreting microbiologically influenced stress corrosion with machine learning and theoretical analysis

Abstract

Numerous tests have demonstrated the impact of microbiologically influenced stress corrosion (MISC) on oil and gas pipelines, but the dynamic corrosion process and its influencing variables remain unclear. In this paper, through material and environment data collection, we analyzed the important factors to MISC based on the random forest model, which were quantity of bacteria, kernel average misorientation and prior austenite grain boundary of the material. Based on this, theoretical explanation of stress and nitrate-reducing bacteria promoting stress corrosion cracking was provided. Results of this study will serve to gain further knowledge of MISC and guide future protection efforts.