A numerical model of a solid oxide fuel cell electrode is presented. Using an already established algorithm for dropping spheres as a base, alterations are made to the algorithm to increase the realism of the model. Two changes are analyzed in detail:​ the ability to drop pore former particles, and the use of pre-agglomerated solid particles. These changes are characterized by their impact on mean pore size, tortuosity, and effective diffusivity. As pore former volume fraction is increased, so too are mean pore size and tortuosity. A higher mean pore size has a beneficial effect on effective diffusivity due to Knudsen effects, while a higher tortuosity has a detrimental effect on effective diffusivity. The impact of mean pore size and tortuosity on diffusivity generally balances and if the impact of porosity is ignored, pore former volume fraction does not greatly affect effective diffusivity. As pore former particle size is increased, mean pore size and tortuosity also increase. Similarly to before, the effects of mean pore size and tortuosity balance. However, effective diffusivity is shown to decrease slightly with an increasing pore former particle size, suggesting a change in tortuosity has greater impact on diffusivity than a change in mean pore size. For a domain constructed with pre-agglomerated particles, the tortuosity and mean pore size were both noticeably larger than when no pre-agglomerated particles are used. Effective diffusivity was only slightly higher for a domain constructed with pre-agglomerated particles than with no pre-agglomerated particles. It is also shown that the relationship of effective diffusivity with porosity for a domain constructed with pre-agglomerated particles does not fit the correlation proposed by Berson et al. [1] for low porosity structures.

A secondary goal of this work is to examine pore size measurement techniques, and present a novel technique that allows the determination of a local pore size, and therefore, a local Knudsen number. Results from the local pore size technique do not match those of the random walk method and so although the novel technique may prove to be a good starting point, it is deemed not yet suitable for use.