Sharon Ellman, Arjen Mascini , Tom Bultreys 

Pore-scale modelling is an important tool to improve our understanding of multiphase flow in porous media. Slow fluid invasion is commonly modelled using quasi-static pore network models (PNM). These models simulate the invasion in a network of simplified pores and throats by invading network elements in order of the quasi-static “invasion” capillary pressure needed for the invading fluid to enter. Despite a multitude of studies that address the predictiveness of PNM, it remains unclear what the leading causes of errors in these methods are, particularly during imbibition. To address this, we developed a novel method to directly validate the invasion capillary pressure models that underlie quasi-static PNM for the first time. The new method compares these models to local capillary pressures measured during in-situ flow experiments visualized with 4D μCT. We applied this to two different open-source PNM extractions from a μCT dataset of a glass beads pack that underwent slow imbibition. This methodology is limited by the temporal resolution of the data, hence we tested assumptions regarding displacement sequences when individual displacements could not be resolved. To constrain the uncertainty on the input parameters, we used local contact angles measured from the μCT images. In the PNMs we investigated, the model-predicted invasion-Pc values were on average greater than the direct measurements made using curvatures from the μCT images. Important sources of mismatch were the difficulty to accurately describe the pore space of a beadpack as a network of pores and throats, as well as the relatively low temporal resolution of the μCT dataset. The method presented here can be used to direct the development of improved pore network models.