Prediction of Corrosion Rate for Carbon Steel Using Regression Model with Commercial LPR Sensor Data

Abstract

In this study, a model was proposed to predict the corrosion rate (Mils per Year, MPY) of carbon steel in a 3.5% NaCl solution, with the objective of comparing the effectiveness of a commercial LPR sensor against traditional electrochemical methods, using potentiostat-based LPR techniques. The primary factors considered in the experiments were temperature, flow velocity, and pH, tested through a full factorial design to identify the most influential variables. Statistical analysis showed that temperature and flow velocity had a significant effect on corrosion rate, with their interaction having the most substantial impact. In contrast, pH had no statistically significant influence within the tested conditions, likely due to the dominant effects of temperature and flow velocity in the high-salinity environment. The MPY data were validated through Tafel plots, immersion coupon tests, and other electrochemical techniques to confirm the reliability of the measurements. A regression model trained on 54 MPY data points demonstrated high accuracy, achieving a coefficient of determination (R2) of 0.9733. The model also provided reliable predictions for factor combinations excluded from the training dataset. Additionally, scenario-based evaluations highlighted the model’s performance under simulated operating conditions, while revealing challenges related to sensor contamination during long-term use. These findings emphasize the potential of commercial LPR sensors as effective tools for real-time corrosion monitoring and demonstrate the utility of the regression model in marine environments.