Accessing Elastic Properties of Porous Solid Oxide Fuel Cell Electrodes Using 2D Image-Based Discrete Element Modeling and Deep Learning

Abstract

The mechanical properties of solid oxide fuel cells (SOFCs) can limit their mechanical stability and lifespan. Understanding the correlation between the microstructure and mechanical properties of porous electrode is essential for enhancing the performance and durability of SOFCs. Accurate prediction of mechanical properties of porous electrode can be achieved by microscale finite element modeling based on three-dimensional (3D) microstructures, which requires expensive 3D tomography techniques and massive computational resources. In this study, we proposed a cost-effective alternative approach to access the mechanical properties of porous electrodes, with the elastic properties of La_0.6Sr_0.4Co_0.2Fe_0.8O_3−δ cathode serving as a case study. Firstly, a stochastic modeling was used to reconstruct 3D microstructures from two-dimensional (2D) cross-sections as an alternative to expensive tomography. Then, the discrete element method (DEM) was used to predict the elastic properties of porous ceramics based on the discretized 3D microstructures reconstructed by stochastic modeling. Based on 2D microstructure and the elastic properties calculated by the DEM modeling of the 3D reconstructed porous microstructures, a convolutional neural network (CNN) based deep learning model was built to predict the elastic properties rapidly from 2D microstructures. The proposed combined framework can be implemented with limited computational resources and provide a basis for rapid prediction of mechanical properties and parameter estimation for multiscale modeling of SOFCs.

Graphical Abstract