A fuzzy logic based algorithm for defining and extracting pore network structure from tomography images of rocks

• Published in: Advances in water resources Vol. 119; pp. 197 - 209
• Main Authors: Moaddel, A., Müter, D., Gooya, R., Sørensen, H.O., Stipp, S.L.S.
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.09.2018; Elsevier Science Ltd
• Subjects: Algorithms; computed tomography; Fuzzy logic; geometry; Image processing; Image segmentation; Information entropy; Mathematical models; Medical imaging; Percolation; Pore network model; Pores; Porosity; Porous media; Properties; Properties (attributes); Quality assessment; Rock; Rocks; Tomography; water resources; Watersheds; X ray imagery; Fuzzy logic; Porous media; Tomography; Pore network model; Percolation; Information entropy
• ISSN: 0309-1708; 1872-9657
• DOI: 10.1016/j.advwatres.2018.07.011

•The definition of a pore is vague and arbitrary. To make sense of pore network models and pore related quantities (pore size distribution, pore coordination number, throat size distribution), a quantifiable definition of a pore is needed.•We applied fuzzy logic to tackle the issue of quantifying the pore as an entity in a porous medium and tested it on a variety of rock types, using tomography images.•By using a plot of information content of pore coordination number distribution, it is possible to cluster different rocks. X-ray tomography has opened a window into the invisible world of pore structure in rocks. Using tomography images as input, pore network models try to use a simple geometric representation to acquire a detailed description of the pore volume. Algorithms developed so far to extract the pore network from tomography images are primarily focused on calculating petrophysical properties of the rock sample. The present study is concerned with extracting a pore network that enables us to elicit information about structural properties and features of the rock sample. However, there is no quantifiable definition of what a pore is and deciding how to divide the pore volume into discrete pores is vague and arbitrary. Fuzzy logic is an extension of classical logic that can accommodate situations where the nature of problem itself is vague. We have developed an algorithm to extract the pore structure and quantify the pore properties from tomography images of rocks. In the algorithm a fuzzy logic based inference engine is used to assess the quality of pores extracted by watershed segmentation and modify them if needed. We have also designed a fingerprinting plot based on the information content of the network, to quantify performance of the algorithm. Results show that the algorithm is able to link the overall structure of the porous space to the final network and produce reliable results for rocks with a variety of pore structures.