Correlation between molecular features and electrochemical properties using an artificial neural network. (15th December 2016)

• Record Type: Journal Article
• Authors: Chen, Fiona Fang; Breedon, Michael; White, Paul; Chu, Clement; Mallick, Dwaipayan; Thomas, Sebastian; Sapper, Erik; Cole, Ivan
• Is Part Of: Materials & design. Volume 112(2016)
• Page Start: 410
• Page End: 418
• Publication Date: 2016-12-15
• Subjects: Electrochemical property -- Molecular structure -- Corrosion inhibitor -- Artificial neural network -- Molecular modelling; Materials -- Periodicals; Engineering design -- Periodicals; Matériaux -- Périodiques; Conception technique -- Périodiques; Electronic journals; 620.11
• DOI: 10.1016/j.matdes.2016.09.084
• Languages: English
• ISSNs: 0264-1275

Abstract

Abstract: The increasing demand for environmentally-friendly and non-toxic coating systems from the aerospace and heavy industry sectors is driving innovation in corrosion inhibitor design and functional coating development. A fundamental understanding of how molecular structure and functionality influences the electrochemical responses of inhibited coatings is crucial for the design of effective functional coatings to replace stalwart, yet highly toxic industrial solutions. In this paper, an artificial neural network approach is presented to quantitatively study the relationship between the structural/molecular features of inhibitor compounds and their experimentally measured electrochemical properties. The presented method is applied to correlate molecular features of corrosion inhibitors with experimentally obtained corrosion potential (Ecorr), corrosion current (Icorr) and anodic/cathodic Tafel slopes. The neural network model, trained through an automatic optimization process, was able to predict the electrochemical performance for a given inhibitor molecule candidate. We will demonstrate how it can be utilised to assess the impact of molecular structure on the final effectiveness of the candidate corrosion inhibitor molecule. The presented neural network learning method could be applied to other areas in materials science for accelerating general materials discovery and functional coating design. Graphical abstract: Highlights: A combined experimental and modelling approach to elucidate key molecular properties of corrosion inhibiting molecules. Electrochemical properties are correlated with molecular features using a neural network model for inhibitor design. Robust predictions of electrochemical properties are achieved via an automatically trained network from measurements. Impact of molecular features on the effectiveness of corrosion inhibitor on an aluminium alloy is assessed and ranked.