Three-Dimensional Modeling of Full-Diameter Micro–Nano Digital Rock Core Based on CT Scanning

Abstract

Characterizing tight reservoirs is challenging due to the complex pore structure and strong heterogeneity at various scales. Current digital rock physics often struggles to reconcile high-resolution imaging with representative sample sizes, and 3D digital cores are frequently used primarily as visualization tools rather than predictive, computable platforms. Thus, a clear methodological gap persists: high-resolution models typically lack macroscopic geological features, while existing 3D digital models are seldom leveraged for quantitative, predictive analysis. This study, based on a full-diameter core sample of a single lithology (gray-black shale), aims to bridge this gap by developing an integrated workflow to construct a high-fidelity, computable 3D model that connects the micro–nano to the macroscopic scale. The core was scanned using high-resolution X-ray computed tomography (CT) at 0.4 μm resolution. The raw CT images were processed through a dedicated pipeline to mitigate artifacts and noise, followed by segmentation using Otsu’s algorithm and region-growing techniques in Avizo 9.0 to isolate minerals, pores, and the matrix. The segmented model was converted into an unstructured tetrahedral finite element mesh within ANSYS 2024 Workbench, with quality control (aspect ratio ≤ 3; skewness ≤ 0.4), enabling mechanical property assignment and simulation. The digital core model was rigorously validated against physical laboratory measurements, showing excellent agreement with relative errors below 5% for key properties, including porosity (4.52% vs. 4.615%), permeability (0.0186 mD vs. 0.0192 mD), and elastic modulus (38.2 GPa vs. 39.5 GPa). Pore network analysis quantified the poor connectivity of the tight reservoir, revealing an average coordination number of 2.8 and a pore throat radius distribution of 0.05–0.32 μm. The presented workflow successfully creates a quantitatively validated “digital twin” of a full-diameter core. It provides a tangible solution to the scale-representativeness trade-off and transitions digital core analysis from a visualization tool to a computable platform for predicting key reservoir properties, such as permeability and elastic modulus, through numerical simulation, offering a robust technical means for the accurate evaluation of tight reservoirs.

1. Introduction

In the petroleum industry, CT scanning technology utilizes microscale scanning, imaging techniques, and spatial structure analysis to provide new insights for enhanced oil recovery. Digital core technology has been promoted in the geological field, encompassing digital core modeling, fluid simulation, and digital geological modeling. In the 1970s, radiology employed X-rays, and Cormack’s proposal in 1963 to use computer scanning for image reconstruction laid the theoretical foundation for CT technology. Current research on core modeling methods primarily involves X-ray CT scanning, where two-dimensional rock scanning information is used to construct three-dimensional core models [1]. CT technology does not damage the core and accurately describes the internal minerals, pores, matrix, and fracture structures of the core from both macroscopic and microscopic perspectives. These characterization details can reflect the internal pore characteristics, permeability, and pore–fracture connectivity of the core [2]. Three-dimensional core modeling provides “observation data” for physical laboratories, offering significant research value for experimental testing, data interpretation, model establishment, and reservoir evaluation [3,4].

However, a key challenge persists: the trade-off between resolution and representativeness. Domestically, most 3D core models are shaped as cubes or rectangular prisms, which do not represent the actual cylindrical geometry of core samples. Additionally, high-resolution imaging is typically confined to small plug samples, leading to a loss of volumetric representation for larger-scale heterogeneities (e.g., fractures and laminations) present in full-diameter cores [5,6]. Methods for obtaining digital cores generally include direct imaging and reconstruction techniques. For evolving 2D scanning models into 3D models, software such as AVZIO, PerGeos, OpenGL, and the Marching Cubes (MC) algorithm are commonly used [7,8] to rotate, zoom in, and zoom out the core surface for display [9,10].

Grid model simulations are divided into the finite volume method, meshless method, explicit finite element method, implicit finite element method, and discrete element method, all of which share the same fundamental principles [11,12,13,14,15,16,17,18,19,20]. These methods have been applied in mechanics [21,22,23,24], petroleum engineering [25,26,27], medicine [28,29], and other fields.

Due to the challenges and high costs associated with core sampling, the preservation of cores should be emphasized. During sampling, factors such as the varying brittleness of cores and susceptibility to damage can lead to inaccurate observational data, resulting in discrepancies between post-processing results and actual production outcomes. This often necessitates cross-referencing historical data. X-ray CT scanning effectively addresses this issue. However, most researchers use CT scanning merely as an imaging tool for description and observation, without fully realizing the potential of 3D digital cores for physical experiments. A prevalent limitation is the use of these models as passive visualization tools rather than active, computable platforms for predictive geometric and mechanical analysis. The large data volume from CT scans poses challenges for the computer processing of larger models. Some researchers primarily rely on algorithms and 2D planar constructions to create 3D models, which do not truly transcend 2D spatial limitations [30,31,32]. Issues such as low imaging clarity, limited observation scope, and morphological discrepancies from the original core samples persist.

This paper addresses these challenges by developing an integrated workflow that bridges the gap between microscopic characterization and macroscopic engineering response. Specifically, we aim to overcome three key limitations in current digital rock physics: (1) the inability of small plug-based models to represent macro-heterogeneities; (2) the prevalent use of 3D digital cores as passive visualization tools rather than active, computable engineering models; and (3) the frequent lack of rigorous, quantitative validation against physical measurements. Our core objective is to create a computationally capable digital model that authentically connects the micro- to macroscale. By constructing a validated, multi-component, and mechanically computable model from a full-diameter core, this study provides a new pathway for high-fidelity reservoir evaluation.

2. Methods

This study follows an integrated workflow to bridge microscopic characterization with macroscopic engineering properties. The methodology sequentially involves the following: (1) the high-resolution CT scanning of a full-diameter core to acquire 3D grayscale data; (2) image preprocessing (denoising and artifact correction) and segmentation to distinguish pores, minerals, and matrix; (3) 3D reconstruction and Boolean operations to create a unified, watertight geometric model; (4) the conversion of this geometry into a conformal finite element mesh for mechanical property assignment and simulation (ANSYS 2024; ANSYS Corporation of the United States; Canonsburg, PA, USA); (5) the extraction of a pore network model from the segmented pore space for flow analysis; and (6) the quantitative validation of the digital model’s predictions against laboratory-measured petrophysical and mechanical properties (Python 3.8.6; Python Software Foundation; Wilmington, DE, USA).

2.1. Establishment of 3D Digital Core Model

In CT scanning, the digital core is primarily represented through grayscale images, which reflect the differential absorption of X-rays by various mineral matrices inside the core. The computer processes the detector data and differentiates between pores, matrix, and minerals based on their distinct gray values. The main instruments for CT scanning include rotating the CT-scanning X-ray around the core, translation platform, detector, and human–machine processing computer (Figure 1). The detector is usually processed by converting electrical signals to calculate and record different gray values. The larger the sample density is, the larger the gray value is, and the weaker the penetration is. In this way, the computer can obtain a gray image of data which can establish 3D core. The specific scanning parameters are crucial for image quality and resolution. In this study, the CT scanning was performed under optimized conditions. For the primary shale sample detailed in this work, the scanning was conducted at a voltage of 140 kV and a current of 120 μA, achieving a high resolution of 0.4 μm. These parameters were selected to ensure sufficient penetration while maximizing contrast for segmenting the dense mineral matrix and micro-pores. A comprehensive list of the scanning parameters for different lithologies is provided in

Table 1. Optimized CT scanning parameters for different lithologies.

Lithology	Scanning Current (mA)	Scanning Voltage (kV)	Optimal Resolution (μm)	Penetration Threshold (mm)	Applicable Depth Range (m)
Gray-Black Shale	120	140	0.4	80	2043.03–2043.85
Tight Sandstone	150	160	0.5	100	1800–2500
Argillaceous Limestone	130	150	0.6	90	2200–2800

to demonstrate the applicability of our protocol. To ensure the fidelity of the digital core model, the raw CT images underwent a comprehensive preprocessing pipeline to address common artifacts and noise. Scatter artifacts were suppressed using an adaptive threshold filter (5 × 5 pixel window). Stitching artifacts between image stacks were corrected via image registration based on the Scale Invariant Feature Transform (SIFT) algorithm, achieving a registration error of ≤1 pixel. Intensity inhomogeneities and beam hardening effects were mitigated through histogram equalization, ensuring a grayscale difference of ≥50 levels between pores and minerals for reliable segmentation. Finally, a combination of median filtering (to remove salt-and-pepper noise) and non-local means denoising was applied to enhance the signal-to-noise ratio while preserving critical structural edges. During the initial coring stage, core damage may occur, leading to issues such as missing sections or discontinuities. During scanning, X-ray scattering can cause anomalies in the core’s gray values. Usually, the treatment method is to repair and distinguish the gap with sand or different lithology, and the data will be observed during the processing.

Image Acquisition and Simulation Model Construction

The experimental procedure and parameter selection for this study are detailed as follows. Part of the problem encountered in the establishment of 3D digital core by CT scanning during core-taking in different areas pertained to the internal structure of the rock being different, and the intensity of the X-ray absorbed by the rock being different when converting the internal structure of the rock. If the same core is under different currents and voltages, the current and voltage are large and the penetration is strong; otherwise, the fault is weak. For tight rocks such as the shale sample in this study, a voltage threshold of ≥140 kV and a current threshold of ≥120 μA were found to be necessary to ensure effective penetration through the full-diameter core. Insufficient parameters (e.g., voltage < 120 kV for shale) result in poor penetration, significantly increased noise, and failed segmentation. A comprehensive summary of the optimized scanning parameters for different rock types is provided in Table 1. The strong penetrating force can well identify the minerals, pores, matrix structure, pores, cracks, and so on inside the rock. The weak current and voltage in the rock with high X-ray irradiation density are difficult to penetrate and this affects the experimental observation. In this paper, shale rock samples from area X are selected (

Table 2. Parameters of CT samples.

Depth/m	Rock Name	Porosity/%	CT Scan Photograph
2043.03–2043.85 m	Gray-black shale	4.615%

), and the highest resolution of the full-diameter core scanning is 0.4 μm, used to obtain the gray-level map of the sample. In the whole section of core image processing, including strong light at core splicing, geometric parameter processing, black edge processing, and pixel merging, a 3D digital core basic model is constructed.

The modeling and numerical simulation of digital core technology mostly focus on the micrometer or even nanometer scale, exploring the physical properties of rocks at the microscale, which cannot effectively explain the influence of macro factors (rock structure, bedding, cracks, etc.) on the physical properties of rocks. Based on digital core modeling, a three-dimensional digital wellbore with a hundred-meter continuous component has been constructed for the first time in China. The method of simulating and analyzing the logging response mechanism based on the three-dimensional data volume of the digital wellbore has been pioneered, providing new technical means for the fine evaluation of reservoir parameters and the diagnosis of difficult reservoirs. The steps for constructing a three-dimensional digital wellbore are as follows. First, conduct lithofacies division, in order to obtain data such as layering, fractures, and lithofacies division in the study area through core observation and analysis of the target well electrical imaging images, combined with logging interpretation conclusions. Then, fill the blank zone in the electrical imaging image by applying the multi-point geostatistical Filtersim algorithm to fill the blank zone in the electrical imaging image, forming a complete wellbore image. The electrical imaging scale is a porosity image, and according to the Archie formula, the electrical imaging resistivity calibrated by shallow lateral resistivity is converted into apparent porosity. After the porosity scale is calibrated, the porosity frequency is calculated to generate the apparent porosity spectrum around the well. Adjust the two-dimensional electrical imaging image to a three-dimensional cylindrical surface, and roll the two-dimensional electrical imaging porosity image into a three-dimensional cylindrical surface using dual wellbore diameter as the wellbore control parameter, with the scale matching the wellbore scale. Using the multi-point statistical Snesim algorithm to construct a digital stratigraphy of the selected lithofacies, drill a three-dimensional digital wellbore that matches the logging depth.

2.2. Three-Dimensional Digital Core Experiment Method

The experimental data came from the high-resolution scanning image of the core at 2043.3 m depth with a scanning accuracy of 0.4 μm in the CT scanning core in Table 2. After the core scanning, the minerals, pores, cracks, and matrix inside the core were stripped out and formed into separate components through a series of processing methods using Avizo 9.0 and PerGeos2022.2 software. The processing workflow encompassed image preprocessing, segmentation, and refinement to ensure an accurate digital representation. First, preprocessing was applied to the raw CT images to mitigate noise and artifacts: a median filter (5 × 5 pixel kernel) removed salt-and-pepper noise, while a non-local means denoising algorithm was used to suppress Gaussian noise and preserve edges. Scatter artifacts were corrected using an adaptive threshold filter, and stitching artifacts between image stacks were minimized via feature-based image registration, ensuring a seamless dataset. Second, segmentation was performed to isolate phases. The global image was binarized using Otsu’s automatic thresholding algorithm, which, for the studied gray-black shale, typically determined an optimal threshold value within the gray-level range of 85–110 on an eight-bit scale. Region-growing techniques were then employed to refine the segmentation of mineral phases and connected pore networks. The growth criteria required adjacent elements to have a gray-level intensity within ±15 of the seed point, utilized 26-connectivity in 3D, and applied a minimum size filter to remove isolated regions smaller than 1000 elements. Third, interactive manual correction in Avizo 9.0 addressed residual errors at complex boundaries. To resolve overlapping or ambiguous elements at sub-resolution interfaces—a consequence of partial volume effects—post-processing included morphological operations (erosion and dilation with a 3 × 3 × 3 element kernel) for boundary smoothing and the application of a majority filter within a 5 × 5 × 5 element window to assign ambiguous elements to the dominant local phase. After this segmentation and refinement, the model components were assembled into a complete structure within the assembly module, during which parts were edited and modified to maintain the correlations between the minerals, the pores, and the matrix. The assembly included Boolean operations to merge nodes at component interfaces, as shown in Figure 2 In this paper, the minerals (green), pores (blue), and matrix (purple) inside the core are highly reduced to establish a 3D grid. Minerals, matrix, pore geometry, and mesh are observed in individual components of the core. The observation method is different from previous observations. The pores are generated into a 3D grid to intuitively observe the distribution state of pores and fractures in the core, which can better explain whether the reservoir physical property in this area is good or poor. A 3D digital core model is built and rendered for the convenience of observation (Figure 2).

In the process of constructing the computable finite element model, the preprocessing software ANSYS Workbench was used to define distinct components (minerals, pores, and matrix) and assign their respective mechanical properties based on the segmented digital core. To ensure the mechanical simulations were reproducible, the following specific configurations and assumptions were implemented.

Material Properties: Based on the triaxial compression experiments (Table 2), linear elastic and isotropic behavior was assumed for each phase in this initial mechanical analysis. The dominant matrix phase was assigned an average Young‘s modulus of 21.45 GPa and a Poisson’s ratio of 0.233, consistent with the experimental mean values for the shale. The mineral grains were assigned higher stiffness properties (Young‘s modulus ~38–40 GPa; Poisson’s ratio ~0.23–0.24). The pore space was assigned properties simulating a very soft inclusion.

Model Assembly, Mesh, and Boundary Conditions: The components were assembled into a coherent structure using Boolean operations to merge nodes at interfaces (Figure 3), creating a continuous mesh suitable for stress analysis. The model was discretized into an unstructured tetrahedral mesh, with quality controlled by an aspect ratio ≤ 3 and skewness ≤ 0.4. For subsequent mechanical testing (e.g., simulating lab-scale compression), displacement boundary conditions were typically applied: the base of the core model was fully constrained, while a uniform vertical displacement was applied to the top surface to simulate axial loading. The lateral surfaces were either left free for uniaxial simulation or subjected to confining pressure for triaxial simulation scenarios.

Solver and Assumptions: The meshed model was solved using a static, implicit solver scheme in ABAQUS 2022 for quasi-static analysis. The core’s macroscopic heterogeneity was explicitly represented by the spatial distribution of the three distinct material phases (mineral, matrix, and pore) in the digital model, rather than using an assumed homogeneous anisotropic constitutive law. This approach directly translates the CT-derived microstructure into a mechanically resolvable numerical model.

The actual operation involved editing the digital core model where minerals connected with pores presented intricate geometry. By observing the modeled minerals, pores, and matrix against the original data, by setting different Poisson‘s ratios and densities, and by applying gravity, distinct attributes were defined. The quality of each triangular mesh element, measured by its aspect ratio (the ratio of the longest to shortest side), was maintained below a threshold of ten to ensure numerical stability. When components are not initially “welded”—a Boolean operation in mathematical terms—the pores, minerals, and matrix must be intersected to form a cohesive three-dimensional body with shared nodes at interfaces, as illustrated in Figure 3.

Preprocessing primarily involves generating a conformal mesh through Boolean operations between different components (e.g., minerals and matrix). The goal of operations such as ‘welding’ (node merging) and ‘cutting’ is to integrate distinct parts into a single, irregular component suitable for meshing. Then, the components are meshed to achieve the integration and unification of the mesh and nodes, without clarifying the idea of the mineral, matrix, and pore joint nodes in the core in practical application.

2.3. Finite Element Modeling

Finite element modeling is roughly divided into three steps: preliminary processing, calculation, and post-processing. The preliminary processing includes a CT scanner to provide digital information related data, coordinates, and element number. It provides man–machine interaction, and, additionally, grid division can improve the accuracy of visual model. The core modeling is mainly for the irregular-shape grid-generation algorithm to control the model grid quality. The main step is to generate a plane grid in the software, and then to generate a three-dimensional grid through the spatial feature movement of curves. At present, core meshing is also a major development direction. Firstly, 2D meshing is divided by 2D technology, and the parameters can be modified in the semi-automatic generation of 3D meshing, which is widely used in actual digital core modeling.

2.3.1. Basic Steps

Based on the original data plan for coring, the model plan was established. Before 3D mesh division, 2D mesh division should be gradually divided from the complex part to the simple part. Firstly, the core was separated from the mineral, pore, and matrix geometry, which were saved as dat files and imported into the finite element software for mesh modification. After grid inspection and 3D grid division, the unqualified grids should be found out and stitched, nodes should be merged, or invalid grids should be deleted. After that, the components should be placed in the finite element software according to the actual coordinates to construct the core model. According to the different dimension information of the components in the space, it will be different. After the 2D mesh of the component is divided into 3D mesh (Figure 4), the attribute information of the divided mesh is edited and assigned to material attributes. The fully automated finite element software can assign attributes.

The establishment of a pore network model (PNM) refers to the extraction of a simplified yet representative network of pores and throats from the segmented 3D binary core image, preserving its essential topology and geometry for flow and transport analysis. In this study, a customized erosion–dilation algorithm was implemented in Python 3.8.6 to extract the PNM from our digital core. The practical application of this algorithm to the studied gray-black shale sample yielded the following quantitative characterization of its pore system.

• Pore and Throat Geometry: The algorithm identified and quantified the fundamental elements of the pore space. The pore throat radius distribution ranged from 0.05 to 0.32 μm, with a median pore throat radius of 0.14 μm (standard deviation: 0.06 μm). The throat length distribution varied between 0.2 and 1.8 μm, with a median of 0.75 μm (standard deviation: 0.32 μm). These distributions were derived by fitting maximum inscribed spheres within the pore bodies and characterizing the connecting channels, providing a direct measure of the pore-scale geometry controlling fluid access and storage.
• Topological Connectivity (Coordination Number): The coordination number (Z), which indicates the number of throats connected to each pore, is a critical topological parameter for assessing connectivity. Analysis of the extracted PNM revealed an average coordination number (Z) of 2.8. The distribution of connectivity was as follows: approximately 35% of pores had two connections (Z = 2), 42% had three connections (Z = 3), and 23% had four or more connections (Z ≥ 4). This statistically quantifies the pore network’s topology, indicating a moderately connected system with a significant proportion of pores having only two connections, which influences potential percolation pathways.
• Overall Connectivity and Model Utility: Based on the 26-connectivity definition used in the extraction process, the overall pore connectivity rate of the digital core model was calculated to be 89.6%. This high connectivity rate, combined with the detailed geometric and topological statistics above, confirms that the extracted PNM successfully captures the essential flow-relevant features of the tight shale’s complex pore space. This quantitatively characterized PNM serves as the direct input for subsequent predictions of permeability and multi-phase flow behavior, moving beyond algorithmic description to applied reservoir characterization.

2.3.2. Mesh Shape Control

The better the ability of the model to adopt good quality elements to restore the actual effect, the faster the convergence rate during calculation, and the fewer the number of elements in the division elements, the faster the calculation will be; otherwise it will be slow. In the calculation core, the uneven shape, distortion, and skew of tens of millions of grids will make the calculation fail. If obtuse angle distortion occurs to these grids, the nodes can only be increased, and the data volume will double and the calculation amount will increase. The main method to solve this kind of problem is to calculate separately and divide the model into the combination for construction. Therefore, finite element modeling has low cost, good convergence, and easy processing. Triangular mesh is generally used for complex models and irregular models, and the element close to the triangular body is preferred in the three-dimensional mesh model (Figure 5).

Improving the element distortion degree, angle, and merging nodes, deleting unnecessary points, lines, and surface quality evaluation indexes, and reducing the element distortion degree and error can effectively improve the accuracy of grid division. The 2D triangular mesh should meet the preset standards of aspect, skew, warpage, length, and the minimum angle of the 3D triangular mesh. Only with good mesh quality can the calculation be carried out. Since the core mesh is mostly complex in tens of millions of data volume analysis objects, it is impossible to adopt a completely idealized model—only a model as close as possible to the ideal meshing element.

The most commonly used Boolean operation in 3D digital core modeling includes association, intersection, and subtraction. The logical operation method is introduced in graphic processing problems to make the minerals, pores, and substrates merge into a new shape in space, and a two-dimensional Boolean operation is applied to three-dimensional graphic Boolean operation in practical operation. In Figure 6 the three-dimensional Boolean operation of minerals and matrix results in the intersection of minerals and matrix, and the common nodes are fused together. The matrix contains minerals, and the mesh quality is excellent. In the follow-up physical experiments, pore mesh deletion or failure parameters can be given to observe the spatial changes in the pores, minerals, and matrix in the follow-up displacement or statics analysis.

3. Results

The finite element simulation model, constructed from 2D to 3D, possesses a realistic appearance and can be manipulated through rotation, movement, hiding, scaling, and other operations. In this experiment, the micro–nano three-dimensional digital core resolution 0.4 μm finite element data grid has 60,516,400 elements and 10,251,950 nodes. The model has good meshing integrity, clear profile structure, meets the standard of three-dimensional finite element model, and is suitable for further mechanical analysis and physical experiment analysis (Figure 7). The capability of the finite element model to simulate mechanical behavior is demonstrated in Figure 7, which presents the simulated displacement field (in micrometers, μm) on a representative cross-sectional slice of the digital core under simulated uniaxial compression. Quantitative analysis of the displacement distribution shows a range from minimal displacement (dark blue) to a localized peak displacement value observed in high-strain zones (red/yellow). The heterogeneous deformation pattern exhibits a direct spatial correlation with the underlying microstructure, as regions of higher displacement consistently align with areas of higher pore concentration or the lower-stiffness clay matrix identified from the CT-derived geometry. This correlation quantitatively demonstrates how the explicit representation of microscale heterogeneities governs the local mechanical response, which is fundamental to understanding reservoir geomechanical behavior such as compaction and stress distribution.

4. Discussion

4.1. Validation of the Digital Core Model Against Physical Measurements

To quantitatively evaluate the accuracy of our digital core model, key physical properties derived from the model were compared directly with laboratory measurements on the same core sample. As summarized in

Table 3. Comparison between model predictions and physical experimental measurements.

Validation Metric	Model Result	Physical Measurement	Relative Error	Experimental Method	Validation Metric
Porosity (%)	4.52	4.615	2.06%	Helium Porosimetry	Porosity (%)
Permeability (mD)	0.0186	0.0192	3.13%	Steady-State Gas Permeability	Permeability (mD)
Elastic Modulus (GPa)	38.2	39.5	3.29%	Uniaxial Compression Test	Elastic Modulus (GPa)
Poisson’s Ratio	0.23	0.24	4.17%	Uniaxial Compression Test	Poisson’s Ratio
Average Throat Radius (μm)	0.125	0.121	3.31%	Mercury Injection Capillary Pressure	Average Throat Radius (μm)

, the model demonstrates high fidelity to the physical reality.

The relative errors for all parameters are below 5%, indicating that the 3D digital core model constructed from CT scans accurately replicates the macro-physical and mechanical properties of the original rock sample. This successful validation provides a solid foundation for subsequent numerical experiments.

In addition, the validated model bridges the micro–nano and macroscopic scales through direct upscaling. The digital core’s explicit microstructure (Figure 4) enabled the prediction of macroscale elastic properties; a simulated compression test yielded a Young‘s modulus of 38.2 GPa, closely matching the lab-measured 39.5 GPa. Furthermore, flow simulation based on the extracted pore network’s specific geometry (pore throat radii and coordination number) predicted a permeability of 0.0186 mD, aligning with the measured 0.0192 mD. These results quantitatively demonstrate how our microscale model governs and predicts macroscopic behavior.

4.2. Quantitative Characterization of the Pore Network

The extracted pore network model provides detailed insights into the microstructural characteristics governing fluid flow. Statistical analysis revealed a pore radius distribution ranging from 0.05 to 0.32 μm, with a median of 0.14 μm. The coordination number, which indicates the connectivity of the pore space, had an average value of Z = 2.8. The distribution showed that 35% of pores had two connections, 42% had three connections, and 23% had four or more connections, resulting in an overall pore connectivity rate of 89.6%. These quantitative metrics are crucial for understanding and predicting the transport properties of the tight reservoir.

5. Innovativeness

This study presents an integrated workflow for constructing high-fidelity, computable 3D models of full-diameter core samples. The key advantages of this approach over current methods are as follows:

Current digital rock physics (DRP) studies rely on small plug samples (a few millimeters in diameter) to achieve high resolution [2,5]. While this allows for detailed microscale analysis, it fails to capture larger-scale heterogeneities such as natural fractures, laminations, and bedding planes, which are critical determinants of macroscopic mechanical behavior and fluid flow. Our approach, utilizing full-diameter core scanning, bridges this gap. It preserves these crucial structural features, enabling more representative and up-scalable simulations of reservoir rock behavior.

A common limitation in current research is the use of 3D digital cores primarily for visualization and qualitative description [6,9]. Our workflow advances beyond this by seamlessly integrating the digital core with finite element analysis (FEA). Through rigorous mesh generation and Boolean operations, we create a conformal mesh where minerals, pores, and the matrix are “welded” together with shared nodes. This allows for the direct assignment of material properties and the simulation of complex physical processes, such as mechanical deformation under in situ stress, transforming the model from a static image into a dynamic, computable digital twin.

6. Conclusions

This study establishes a high-fidelity, computable 3D digital twin of a full-diameter shale core by integrating micro-CT scanning, image processing, and finite element analysis. The key findings and their context are summarized as follows:

(1)

The digital core model, constructed from high-resolution (0.4 μm) CT scans, demonstrates accurate replication of the specific sample under laboratory conditions. The validation against physical measurements yielded relative errors below 5% for porosity, permeability, and elastic modulus, confirming the workflow’s capability to produce a reliable digital representation for the characterized lithology.

(2)

The Pore network analysis of the validated model quantified the microstructure of this tight shale, revealing a characteristic pore throat radius distribution and an average coordination number (Z = 2.8) that explains its low intrinsic permeability. It is acknowledged that this characterization reflects the unstressed, laboratory-state geometry.

(3)

The integrated workflow successfully bridges scales through a hierarchical approach, using the full-diameter context to guide high-resolution imaging and property calculation. While computationally intensive for a single, detailed model, the process proves feasible and provides a foundational platform. Its primary value is as a high-fidelity tool for deriving constitutive relationships and conducting detailed mechanistic studies, rather than for high-throughput, field-scale statistical modeling.

(4)

The study acknowledges important limitations that define the scope of the current model and direct future work. These include the following: the validation on a single lithotype; the homogenization of sub-resolution nanoporosity; the representation of the sample in its relaxed, “as-received” state; and the simplifying assumptions of perfect interfacial bonding and isotropic, linear elastic material behavior. Furthermore, complex processes such as chemo-poromechanical interactions (e.g., clay swelling) and dynamic multiphase flow are not yet incorporated.