Nowadays material science involves powerful 3D imaging techniques such as X-ray computed tomography that generates high-resolution images of different structures. These methods are widely used to reveal information about the internal structure of geological cores; therefore, there is a need to develop modern approaches for quantitative analysis of the obtained images, their comparison, and classification. Topological persistence is a useful technique for characterizing the internal structure of 3D images. We show how persistent data analysis provides a useful tool for the classification of porous media structure from 3D images of hydrocarbon reservoirs obtained using computed tomography. We propose a methodology of 3D structure classification based on geometry-topology analysis via persistent homology.
Studies of flow dynamics in complex disordered media are very important in many practical areas, such as materials science, soil science, groundwater engineering, chemical engineering, and especially petroleum and gas engineering. Integral geometry methods are a useful tool for studying complex media. In this work, it is shown that, for a given porous media class, the by-layer mean flow velocity for any sample within the same class can be characterized as a function of Minkowski functionals, allowing to avoid costly natural core flood experiments or numerical simulation for screening purposes. This paper proposes the flow characterization method based on integral geometry. This method allows for obtaining the single-phase fluid flow velocity characterization models across a wide range of porous media classes for a quick estimation of the mean by-layer velocity distribution purely by extracting the geometrical measures from binary sample images. Samples from gas reservoirs are chosen as the relevant porous media examples, with regard to the growing importance of this type of reservoir caused by the global shift towards natural gas as a key energy source. Direct comparison with numerical simulation based on Stokes equation was made with commercial class software. The results demonstrate that the proposed algorithm has a relatively high degree of statistical significance and closely captures mean velocity trends. That provides a useful tool for quick, robust modelling for screening, agile calculations, and upscaling tasks.
12 November 2025 in Transaction Kola Science Centre
This study demonstrates how the application of Minkowski functionals (M0 — volume, M1 – surface area, M2 — mean integral curvature, M3 — Euler-Poincaré characteristic) along with Betti numbers can reveal relationships between different object types (phases) within a shared volume. We showcase the use of Minkowski functionals and Betti numbers for characterizing the internal structure of complex materials, exemplified by minerals from a natural geological sample (meimechite, Kontozero Complex). The significant advantage of the presented approach is its ability to provide quantitative characterization of the topological and morphological properties of the phases. The image used for analysis via algebraic topology and integral geometry was acquired using X-ray computed microtomography (X-ray μCT).
26 December 2025 in Transaction Kola Science Centre
This study presents a method for quantitative description of the internal 3D structure of a natural sample (meimechite) based on integral geometry and algebraic topology. Microtomography (with 51 µm voxel resolution) was used as the source data. The image was segmented into phases according to X-ray density levels using the multi-Otsu method. For each phase, connected components were extracted, and invariant integral-geometric characteristics were computed — Minkowski functionals M₀–M₃ and their corresponding Betti numbers b₀, b₁, b₂. A segmentation quality classifier based on the behavior of M₀–M₂ was proposed and tested for stability. A detailed topological interpretation of combined phases allowed the identified topological signatures to be related to the geological and mineralogical features of the sample.
Owing to the development of optical precision and computer technology, the number of experiments based on visual methods for processing data from track detectors is growing. The use of neural networks to identify clusters in images obtained using new-generation scanning stations will both speed up image processing in nuclear emulsions and make the track reconstruction process an order of magnitude more efficient.