Synthetic dataset of pore scale multiphase flow from direct numerical simulations

Understanding the physics of fluid displacement through the pore spaces in multiphase environments are essential for improving the safety and optimizing the performance of diverse complex subsurface engineering applications. We conduct high-fidelity two phase flow simulations at the pore-scale using direct numerical simulations on six porous media that have different permeability. We consider nine viscosity ratios and ten contact angles, resulting in a total of 540 simulations. The open-source dataset available on Zenodo includes specifics related to the flow patterns, as well as the spatial and temporal variations of pressure and velocity within the pore spaces. This comprehensive dataset can support (i). the training of machine learning algorithms and (ii). serve as a benchmark for flows predicted by computationally efficient tools including machine learning algorithms and pore network models.