Investigation of corrosion Inhibition capability of ionic liquid: a machine learning approach

In this scientific inquiry, we explore the optimal integration of machine learning (ML) techniques with Density Functional Theory (DFT) calculations for predicting the corrosion inhibition efficiency (CIE) of ionic liquid compounds. The escalating demand for effective corrosion inhibitors underscores the necessity for reliable and cost-efficient prediction methodologies. Given the costly nature of experimental approaches, this study seeks to address the challenge of cost-effectiveness and efficiency in predicting corrosion inhibition properties. Utilizing a quantitative structure-property relationship (QSPR) model, our investigation centres on forecasting CIE values for three external test sets of ionic liquid compounds (IL-A, IL-B, and IL-C) as external validation. Through a comprehensive evaluation employing various metrics, the Gradient Boosting (GB) model emerges as the most accurate predictor among linear, non-linear, and ensemble models, showing excellent accuracy with a high R2 value of 0.98. Apart from that, the RMSE, MAE, and MAD values are low, namely 0.95, 0.84, and 0.94, respectively. The predicted CIE values for the three external validation ionic liquids were 88.95%, 90.82%, and 93.16%, respectively, indicating strong agreement with experimental findings. This technological advancement holds promise for anticipating the properties of new corrosion inhibitor compounds before their experimental synthesis, thereby advancing the field of corrosion inhibition research.