Main content area

Pore network extraction using geometrical domain decomposition

Rabbani, Arash, Mostaghimi, Peyman, Armstrong, Ryan T.
Advances in water resources 2019 v.123 pp. 70-83
algorithms, computers, digital images, image analysis, memory, micro-computed tomography, models, permeability, porous media, statistics, water resources, watersheds
Advancement of X-ray micro-computed tomography (micro-CT) imaging technologies demands more efficient numerical methods that are capable of handling huge 3D images of porous material. We present a workflow for extraction of a pore network model using domain decomposition. This method enables us to analyze huge micro-CT images with minimal computational requirements. We study the impact of domain decomposition on pore network statistics and directional permeability of several porous media. Also, an approximation of the memory usage is presented based on the size of the images and the domain decomposition ratio. The watershed segmentation algorithm is used to extract the pore network of each subdomain. Then by using the labeled matrix of the porous space, the connections between subdomains are detected. This leads to the creation of a global pore network model that is consisting of the local pore networks of each subdomain. Finally, the absolute permeability of 6 large images is calculated to demonstrate the capability of the proposed method. The largest sample we use for pore network extraction is a Fontainebleau image with 20003 voxels. It is worth noting that these simulations are performed using a personal computer and that we would envisage that much larger volumes could be analyzed by using supercomputing facilities.