Improvements in the imaging techniques have enabled the capturing of images with high resolution, thereby permitting the reconstruction of complex porous media that can be analyzed by computer simulations. The two most popular methods for numerically analyzing transport in porous media are pore network modelling (PNM) and direct numerical simulation (DNS).

This thesis describes a method to characterize the microstructure of various porous media used for fuel cell and electrolyzer. In this work, two fibrous gas diffusion layers (Toray 120C, SGL 39BA) and one powder-based sintered titanium paper were directly scanned using X-ray micro-computed tomography (μCT). The obtained raw images were binarized and cropped into three subsamples. To assess the validity of subsamples, the statistical characterization was performed using two-point correlation and chord length functions. Stochastic reconstructions were used to generate catalyst layers (CLs).

The microstructure of the subsamples are analyzed by computing pore size distribution (PSD), mercury intrusion porosimetry (MIP), and water intrusion. PNM uses pore networks extracted from the subsamples to obtain PSDs. The networks are then used for non-wetting phase intrusion in the domain using percolation algorithms. For DNS, the sphere fitting algorithm is used for the PSD calculations. The PSD information is then used for water and mercury intrusion using the cluster-based full morphology (FM) algorithm. PSDs computed from MIP data are in good agreement with the PSDs from the sphere fitting algorithm.

Dry and wet transport properties are computed using both methods and compared to experimental references. Two diameter cases provided by PNM are used for numerical calculations. Using the equivalent diameter returned closer results compared to the experimental transport properties. In PNM, relative water permeability results show different trends based on the intrusion algorithm used for the transport simulations. Employing the invasion percolation (IP) algorithm returns better results compared to the reference results. The DNS results obtained based on saturated images return acceptable results.