Machine Learning-Driven Discovery and Design of Corrosion Inhibitors

Corrosion inhibitors remain one of the most practical approaches for reducing metal degradation in acidic, chloride-rich, marine, and industrial environments. However, conventional inhibitor discovery is constrained by slow experimental screening, narrow chemical exploration, inconsistent reporting of electrochemical conditions, and the growing need for environmentally acceptable molecules. This analytical mini-review examines how machine learning (ML) improves corrosion inhibitor discovery and design. The central argument is that ML does not improve an inhibitor molecule by changing its chemistry directly; rather, it improves the discovery pathway through molecular representation, prediction, ranking, interpretation, uncertainty-aware selection, and experimental feedback. Recent studies show that quantitative structure-property relationship models, support vector regression, kernel ridge regression, random forest, gradient boosting, k-nearest neighbor methods, and molecular graph models can estimate inhibition efficiency, screen candidate libraries, identify influential descriptors, and guide electrochemical validation. The review also emphasizes limitations that are especially important for conference-level interpretation: small datasets, heterogeneous experimental protocols, weak applicability-domain reporting, overfitting, and the risk of confusing statistical feature importance with mechanistic proof. The strongest future contribution of ML is expected to come from closed-loop workflows that combine curated corrosion datasets, chemically meaningful descriptors, uncertainty-aware models, electrochemical validation, and sustainability-oriented molecular design.

کلیدواژه ها:

machine learning ، corrosion inhibitor ، inhibition efficiency