Visualizing and studying pore-scale flow dynamics inside a natural porous media can be a tedious and costly task. Micromodels have been widely used in for this purpose, but often the modeled geological object’s representation is not sufficiently realistic. Thus we investigated numerical approaches to design microchips realistically mimicking rock structure. Our objective is to provide a tool for on-demand generation of realistic porous rock structure with adjustable porosity.
Our new aggregation model of microstructure is derived to combine images of grains, obtained from high resolution images of rock, and simulate porous networks with grains of variable compactness.
A demonstration simulation package is provided within Plug Im!, a signal-and-image-processing platform from IFPEN (Visit Plug Im! website: www.plugim.fr ).
Our method can generate realistic 2D system with either high or low porosity values. Different grain shapes and sizes can be chosen from a library that can be enriched or customized.
Such virtual microstructure can be used to study digitally the effect of the microstructure on fluid flow properties. It could also generate geologically realistic training images for other machine learning models, for instance in a perspective of automated interpretation of core scans or thin sections.