Machine learning models and computational simulation techniques for prediction of anti-corrosion properties of novel benzimidazole derivatives

Abstract

The present paper delves into the development of predictive models for the optimum prediction of inhibition efficiencies and anti-corrosion properties of newly designed benzimidazole compounds in an HCl medium. Density functional theory (DFT) was used to obtain the molecular descriptors. 17 descriptors considered as input variables were reduced to 9 after redundant variables were eliminated using the variance inflation factor (VIF). Machine learning models such as random forest regression (Rf), K-nearest neighbor (KNN), and gradient boosting (GB) were used to develop a predictive model using data from 50 benzimidazole derivatives whose inhibition efficiencies for steel alloy in HCl medium have been determined experimentally and are available in the literature. The cross-validation (CV) technique was used to evaluate the predictive capability of the models. KNN gave the best result with a coefficient of correlation (R2) of 0.654, compared to GB (0.591) and RF (0.512). Subsequently, the three models were used to predict the inhibition efficiencies of three novel benzimidazole compounds. The corrosion inhibition efficiency of three newly developed benzimidazole compounds was 91 % - 98.5 % when assessed with the best-performing model, K-nearest neighbor (KNN) with MSBAH recording the highest value of 98.5 %. Fukui indices parameters fk+k,fk−k,andf2(r) showed that both the electrophilic and nucleophilic sites are concentrated around specific atoms most of whom are heteroatoms that participate actively in electron-donor and acceptor interaction with the metal atom leading to the formation of coordinate covalent bonds. The results of electron distribution and Fukui dual descriptors support the results of the predictive model. The study also revealed that machine learning algorithms in conjunction with DFT and MD simulation is an innovative technique for the prediction of anti-corrosion properties of newly designed molecules such as corrosion inhibition efficiencies, and mechanism of inhibitor adsorption on metal/alloy surfaces exposed to an aggressive environment. This technique can be harnessed in proffering solutions to acid corrosion especially those that occur during descaling operation.