Machine learning insight into inhibition efficiency modelling based on modified graphene oxide of diaminohexane (DAH-GO) and diaminooctane (DAO-GO)

Abstract

The effective prediction of corrosion inhibition efficiency (%IE) of modified graphene oxides (GOs); diaminohexane-modified graphene oxide (DAH-GO) and diaminooctane-modified graphene oxide (DAO-GO) is vital for advanced material applications. This study employs a dual-modelling scheme to predict the %IE, for this purpose, four stand-alone machine learning (ML) models (Multivariate Regression (MVR), Gaussian Process Regression (GPR), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Neural Network (NN)), and five simple averaging (SA) ensemble paradigms (MVR-SA, GPR-SA, ANFIS-SA, NN-SA, and Decision Tree-SA (DT-SA)). Feature selection processes were carried out to develop three distinct models, leading to a comprehensive comparative analysis. The results demonstrated that the non-linear stand-alone models (GPR, ANFIS, NN) significantly outperform the linear MVR model, with the M2 model configuration yielding the highest performance across all models. Remarkably, GPR-M2 achieved perfect model tuning with zero error rates, indicating its superior predictive capabilities. Ensemble techniques further improved performance, reflecting the experimental data's complexities in %IE modelling. The hierarchical order of performance in the training phase in the testing phase is DT-SA < MVR-SA < ANFIS-SA < NN-SA < GPR-SA. The GPR-SA ensemble emerged as the most accurate technique, substantially enhancing the predictive accuracy of the ensemble models by up to 67.73% in the training phase and 50.71% in the testing phase. These findings suggest the potential of GPR-SA in boosting the performance of ensemble approaches in material science applications. The study recommended a promising future for ML in the development and application of corrosion-inhibitors.

Keywords

• Graphene
• Machine learning
• Diaminohexane
• Optimization
• DFT simulation