Exploring the use of machine learning for interpreting electrochemical impedance spectroscopy data: evaluation of the training dataset size

Electrochemical impedance spectroscopy (EIS) interpretation is generally based on modelling the response of a corroding system by an equivalent circuit. Although effective, the approach is difficult to automate and uptake in an industrial context is limited. Machine Learning (ML) algorithms can solve complex tasks after a training process and this work explores the possibility of using ML to interpret EIS data. Two scenarios are considered: classification, i.e. identifying which equivalent circuit is associated to an EIS spectrum, and fitting, i.e. estimating the numeric values of the components of an equivalent circuit.