Data-driven corrosion behavior and prediction of EH36 steel in marine atmosphere: Integrating corrosion big data with machine learning

Marine engineering equipment and facilities face significant atmospheric corrosion challenges. Limitations in corrosion data collection and analysis techniques have hindered a comprehensive understanding of the corrosion behavior of marine steel structures under complex atmospheric environmental conditions. In this study, corrosion big data technology was integrated with machine learning methods to investigate the influence of coupled environmental factors, such as relative humidity (RH), temperature, and various pollutants, on the atmospheric corrosion behavior of EH36 steel. The curve of cumulative electric quantity detected by the corrosion big data sensor indicated that the corrosion rate of EH36 steel over a 6-month (m) exposure period initially accelerated, then decelerated, and finally stabilized. This trend is consistent with the weight loss data, confirming the reliability of corrosion big data monitoring technology. Correlation analysis identified RH and temperature as the key factors influencing corrosion. Higher RH facilitated the formation of an electrolyte film on the EH36 steel surface, accelerating corrosion. In contrast, increased temperature reduced RH, resulting in a negative correlation between temperature and corrosion rate. These findings suggest that RH is the most dominant factor affecting the EH36 steel corrosion in marine atmospheric environments. Furthermore, an extreme gradient boosting (XGB) algorithm capable of handling nonlinear relationships and interactions among atmospheric environmental parameters was constructed. The XGB model demonstrated strong predictive performance in estimating the corrosion rate of marine steel structures, contributing to the safe and reliable operation of marine engineering equipment and facilities.