Deep Learning algorithm, based on convolutional neural networks, for equivalent electrical circuit recommendation for Electrochemical Impedance Spectroscopy

[EN] Electrochemical Impedance Spectroscopy (EIS) is a technique widely used in the field of electrochemistry, due to its ability to probe the dynamics of electrochemical systems. Commonly, EIS spectra are analysed using Equivalent Electrical Circuits (EECs). The EEC selection is not trivial. For th...

Descripción completa

Detalles Bibliográficos
Autores: Sáez-Pardo, Fermín|||0009-0000-0759-6781, Giner-Sanz, Juan José|||0000-0003-0441-6102, Pérez-Herranz, Valentín|||0000-0002-4010-0888
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de València (UPV)
Repositorio:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglés
OAI Identifier:oai:riunet.upv.es:10251/231703
Acceso en línea:https://riunet.upv.es/handle/10251/231703
Access Level:acceso abierto
Palabra clave:Artificial vision
Convolutional neural networks
Deep learning
Electrical equivalent circuits
Electrochemical impedance spectroscopy
Descripción
Sumario:[EN] Electrochemical Impedance Spectroscopy (EIS) is a technique widely used in the field of electrochemistry, due to its ability to probe the dynamics of electrochemical systems. Commonly, EIS spectra are analysed using Equivalent Electrical Circuits (EECs). The EEC selection is not trivial. For this reason, Digby D. Macdonald proposed the Electrochemical Genome Project, which would consist of a large database of EIS spectra and an Artificial Intelligence able to recommend EECs given an experimental EIS spectrum. In this work, we developed a Deep Learning algorithm, based on Convolutional Neural Networks (CNNs), for EEC recommendation for EIS spectra. To achieve this, we optimized the CNN in 3 sequential stages: first, the convolutional architecture was optimized; then, the Initial Learn Rate was selected; and finally, the dense network architecture was optimized. At the end of this process, we obtained a CNN model with a maximum test accuracy of 61.11 %. The obtained results show that CNNs are good candidates for EEC recommendation tools for EIS spectra.