A combination of machine learning model and density functional theory method to predict corrosion inhibition performance of new diazine derivative compounds

Abstract

This study proposes a novel approach that combines machine learning (ML) and density functional theory (DFT) methods to construct a quantitative structure-properties relationship (QSPR) model for diazine derivatives as anti-corrosion inhibitors. A dataset is constructed by combining three existing diazine isomer datasets to represent diazine compounds. Thirty-two different ML algorithms were implemented on the dataset, and the gradient boosting regressor (GBR) model was identified as the best predictive model for diazine and each isomer dataset based on the coefficient of determination (R2) and root mean square error (RMSE) metric values. This consistency was also observed when the GBR model was implemented on four other diazine derivatives, resulting in high corrosion inhibition efficiency (CIE) values ranging from 85.02 % to 94.99 %. The DFT calculations for these derivatives also showed strong adsorption energies ranging from − 4.41 to − 6.09 eV, in line with the CIE trend obtained from the ML prediction. This novel approach can provide insights into the properties of prospective organic corrosion inhibitors prior to experimental investigations, which could accelerate the development of new and effective organic corrosion inhibitors.