Pore structure characterization and classification by 3D images of porous media

• Published in: Materials characterization Vol. 228; p. 115415
• Main Authors: Xu, Zhipeng, Xu, Xinkai, Li, Caoxiong, Xu, Leige, Zhang, Xinyang
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.10.2025
• Subjects: 3D images; Pore morphology; Pore structure characterization; Pore structure classification; Porous media; Porous media; Pore structure classification; Pore morphology; 3D images; Pore structure characterization
• ISSN: 1044-5803
• DOI: 10.1016/j.matchar.2025.115415

In studies of physical properties prediction for porous media, it is often desirable to first get the comprehensive characteristics and classification of pore structures. However, the comprehensive pore structure characterization requires a combination of multiple physical measurements and scanning, which is costly and time-consuming. In this paper, a workflow is developed to comprehensively characterize pore structures and rapidly classify them solely by 3D images of porous media. The pore-morphology-based mercury injection capillary pressure simulation and pore-throat system extraction are first performed in 3D images to obtain the geometric and topological parameters of pore structures. These parameters of various samples are used for the varimax orthogonal rotation based principal component analysis and the K-means++ clustering to classify pore structures. The full-morphology algorithms can comprehensively characterize pore structures without assumptions and simplifications. The workflow can function as both an independent classification tool and a provider of training datasets for machine learning models. 20 conventional porous media and 59 shale kerogen blocks are employed to assess the performance of workflow. The comprehensive pore structure characteristics are calculated and three factors with specific physical significances are extracted from the two datasets, respectively. The robustness of factor analysis and clustering algorithms is analyzed. Results show that the factor analysis outcomes remain stable, while the clustering results exhibit a certain sensitivity to perturbations in the dataset. The capillary pressure-saturation curves, pore radius distribution, and throat radius distribution of 20 porous media are analyzed. Results demonstrate that these curves exhibit the corresponding three characteristics patterns. Classification results exhibit good agreement with pore heterogeneity for 20 porous media, quantified by the newly introduced independent parameter average aspect ratio. The apparent permeability curves of 59 shale kerogen are discussed, and the workflow has the potential to be a tool for grading the permeability of shale kerogen. The developed workflow can reliably, rapidly, conveniently characterize and classify pore structures for various porous media based only on 3D images. •A workflow is developed to characterize and classify pore structure for various porous media solely by 3D images.•Comprehensive pore structure characterization is achieved by MICP simulation and pore-throat system extraction.•The integrated characterization algorithms are full-morphology without assumptions and simplifications.•The obtained pore structure characteristics are comprehensive and physically realistic.•The workflow is user-friendly serving as a standalone classifier and a provider of data source for other applications.