機械学習によるSOFC微細構造変化の予測

抄録

Predicting the time evolutions of electrode microstructures of solid oxide fuel cells (SOFCs) is a challenging task as the underlying mechanisms are not fully understood and the available experimental data are scarce. The present study proposes a machine learning method for predicting microstructural changes in the SOFC electrodes. A conditional generative adversarial network with an unsupervised image-to-image translation (UNIT) architecture is incorporated to predict the reduction process of NiO-based SOFC anode. UNIT successfully predicts the microstructural change during reduction for the new microstructures and can predict temperature dependence despite the limited number of training data.