Improving corrosion data modelling through an evolutionary algorithm approach

Abstract

Corrosion is a worldwide spread problem which it is estimated to have a direct cost of over 3% of the world Gross Domestic Product annually. The experts on corrosion point out that up to 25% of costs caused by corrosion annually can be saved if currently available corrosion control techniques are applied. Organic coatings are a widely studied and applied technique for extending the life span of metals exposed to various environmental conditions. The performance and behaviour of the organic coatings against corrosion can be tested with techniques such as Electrochemical Impedance Spectroscopy (EIS). Thus, information on the electrochemical behaviour of the metal–coating system, as well as the presence of pores and defects, can be extracted from EIS data analysis. EIS experimental data modelling can be achieved using several commercial software packages that utilize non-linear regression algorithms with accurate fits. However, they are highly dependent of initial values for parameters and hence present serious problem of convergence to the optimum. Evolutionary algorithms (EAs) do not show these problems and have been proven to be a robust solution in the optimization field. In previous works [1], electrochemical systems, consisting of a coated metal probe in contact with a test solution, were modeled by means of two time–constants equivalent electrical circuits, obtaining very accurate fits. In this work, more complex systems characterized by three time constants are modeled by using state of the art differential evolution approaches. Results are promising, and it is expected in the near future that EAs based optimization methods prevail in the field of electrochemical systems modelling, such as coatings or batteries.