Data-driven investigation to model the corrosion inhibition efficiency of Pyrimidine-Pyrazole hybrid corrosion inhibitors

Abstract

This paper proposes a quantitative structure–property relationship model (QSPR) based on machine learning (ML) for a pyrimidine-pyrazole hybrid as a corrosion inhibitor. Based on the metric values of the coefficient of determination (R2) and root mean square error (RMSE), the extreme gradient boosting (XGBoost) model was found to be the best predictive model for the N-heterocyclic dataset and its respective non-aggregated dataset. When the XGBoost model was applied to three additional pyrimidine-pyrazole hybrid derivatives, this consistency was also seen, and high corrosion inhibition efficiency (CIE) values were obtained ranging from 82.09% to 95.26%. According to the CIE trends found from the ML predictions, DFT calculations for these derivatives also reveal a strong and suitable adsorption energy trend ranging from −1.40 to −1.52 eV. Also supported by the compatibility of the energy gap trend with the CIE trend of the inhibitor molecule. This innovative method can elucidate the characteristics of potential organic corrosion inhibitors before conducting experimental research, which can speed up the preparation of fresh and strong organic corrosion inhibitors.