Synergistic Application of Machine Learning to Microstructural Characterization on Electrode Composites of Solid Oxide Fuel Cells

Abstract

Solid oxide fuel cells (SOFCs) have been recognized as one of the powerful next-generation energy conversion systems in association of the demanding green carbon technology. The electrochemical performance is crucially dependent on the intricate microstructures of porous electrodes, either cathodes or anodes. The composite electrodes should be analyzed in the sophisticated manner by characterizing the microstructural parameters. The current work combines electron microscopy with machine learning, more specifically semantic segmentation. The semantic segmentation was synergistically combined with high volume of electron micrographs enabled by FIB-SEM which has been used for microstructural analyses in SOFCs. The Ni/YSZ anode composites were selected as a model system, by incorporating the third components, i.e., pores featured by porous electrodes. The semantic segmentation-predicted image separation was connected with the conventional linear intercept approach, leading to the automated extraction of microstructural parameters. The work reports an exemplary analysis results based on the machine learning-assisted microstructure characterization in SOFC composite materials, implying that the machine learning-assisted approach becomes an essential tool in coping with high volume of electron microscopy-generated image data.