In-silico Model for Predicting the Corrosion Inhibition Efficiency of Steel Inhibitors

摘要

Quantitative Structure-Activity Relationships (QSAR) based models have been widely used for predicting corrosion inhibition performance of metals. However, one of the major limitations in these studies is that the authors have restricted themselves to use only a single class of molecules having similar molecular structure. In this study, a computational end-to-end framework was developed to investigate the properties of organic corrosion inhibitors which are responsible for inhibition of steel in acidic solution. The framework consists of modules like data preprocessing, descriptor selection and model building. A robust predictive model for multiple class of corrosion inhibitors was developed using advanced machine learning algorithm such as gradient boosting machine (GBM), random forest, support vector machines (SVM) etc. The descriptors were selected using novel integrated ensemble technique. The model based on GBM algorithm was able to predict the corrosion inhibition efficiency of inhibitors with significantly higher accuracy.