A digital core micro-pore area voxel grid coarsening method based on multiple convolution calculation

Abstract

The application discloses a digital core micro-pore area voxel grid coarsening method based on multiple convolution calculation, which comprises the following steps: 1, importing an original three-dimensional scanning image; 2, for the macro-pore area voxel, using a watershed algorithm in image processing to segment the macro-pore area, and extracting a pore network based on a segmentation matrix; 3, performing multiple convolution calculation on the micro-pore area voxel data, and performing coarsening marking on the micro-pore area voxel mapped by the convolution value meeting the coarsening condition; 4, generating grid cell information and connection relationship of the micro-pore area after coarsening by using the micro-pore area matrix marked by coarsening and the macro-pore area segmentation matrix; and 5, combining the pore network of the macro-pore area with the grid cell of the micro-pore area to generate a calculation grid file of a digital core hybrid model. The technical effect of the application is that the number of calculation grids of the micro-pore area is reduced under the premise of ensuring the accuracy of simulation results, and the calculation efficiency is significantly improved.

Description

Technical Field

[0001] This invention belongs to digital core analysis technology and has applications in electrochemistry, underground oil and gas extraction, geotechnical engineering and other fields, with a focus on numerical simulation technology for oil and gas fluid deposits. Background Technology

[0002] Numerical simulation based on digital cores is a cost-effective and short-cycle research method that is of great significance for determining the seepage characteristics of oil and gas reservoirs and optimizing subsequent exploitation schemes. However, for cores of shale, tight sandstone, and carbonate rocks, which have multi-scale pore structures, it is difficult to extract the complete pore structure at a single resolution due to the mutual constraints of the scanning equipment's field of view and resolution. When scanning at a lower resolution, the extracted pore structure may ignore pores below the image resolution, affecting the accuracy of the simulation. When increasing the image resolution to obtain micropore structures, the actual size of the digital core is much smaller than the voxel size represented by pixels, resulting in unrepresentative simulation results. Multi-scale numerical models based on digital cores divide the voxels of the three-dimensional scanned image into solid regions, macropore regions, and micropore regions through multi-threshold segmentation, generating multi-scale digital cores. Different scale control equations are then used to solve the macropore and micropore regions to achieve high-precision numerical simulation of multi-scale digital cores. The term "multi-scale" refers to the fact that some rocks exhibit a bimodal pore distribution, with pore sizes varying by several orders of magnitude.

[0003] Currently, there are three main types of numerical models for fluid seepage in digital cores with multi-scale pore structures: dual-pore network models, micro-continuous media models, and pore network-Darcy models. Dual-pore network models simplify both macropore and micropore regions, offering fast computation but lower simulation accuracy. Micro-continuous media models, based on the original image, label voxels in both macropore and micropore regions and obtain the pressure and velocity distribution within voxel units by solving the Darcy-Brinkman-Stoke equations and the mass conservation equation. This method offers higher simulation accuracy but lower computational efficiency, and the simulated digital core size is typically less than 300 mm. 3 Voxel; The pore network-Darcy model combines the advantages of the pore network model and the micro-continuous medium model. It simplifies the macropore region into a pore network, while treating the micropores as a continuous medium. The Hagen–Poiseuille equation and the Darcy equation control the flow of fluid in the macropore region and the micropore region, respectively. It improves the computational efficiency while ensuring the accuracy of the calculation. However, as the simulated core increases, the number of voxels in the micropore region increases exponentially. When there are tens of millions to hundreds of millions of micropore voxels, the amount of computation related to the Darcy equation becomes enormous, making numerical simulation impossible.

[0004] In summary, existing multi-scale digital core models, while ensuring computational accuracy, are often limited to simulating relatively small areas due to constraints in the number of computational grid cells and computational efficiency. Therefore, how to significantly reduce the number of computational pore grid cells and improve computational efficiency while maintaining the accuracy of the calculation results is a crucial problem that multi-scale digital core numerical models need to solve. Summary of the Invention

[0005] The purpose of this invention is to address the problem of mutual constraints between computational efficiency, computational accuracy, and computational domain size in existing numerical models. It provides a voxel mesh coarsening method for the micropore region of digital cores based on multiple convolution computation. This method can achieve multi-level voxel mesh coarsening from the center of the micropore region outward while ensuring mesh accuracy at the interface between the large-pore region and the micropore region and considering the heterogeneity of the micropore region. This greatly reduces the number of computational grid cells and improves computational efficiency.

[0006] The technical solution of this invention is a method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations, comprising the following steps:

[0007] Step 1: Import the original 3D scan image, filter and remove noise, and perform multi-point threshold segmentation to obtain the distribution information of macropore and micropore regions;

[0008] Step 2: For the macroporous voxels obtained in Step 1, the macroporous region is segmented using the watershed algorithm in image processing. Based on the segmentation map, the pore network is extracted, and the relevant geometric information of the pore throat (referring to pores and throats) is obtained, including volume, specific surface area, and inscribed circle radius.

[0009] Step 3: Perform multiple convolution calculations on the microporous voxel data obtained in Step 1, traverse the convolution matrix sequentially, and coarse-label the microporous voxels mapped by the convolution values ​​that meet the coarsening conditions.

[0010] Step 4: Based on the coarsened micropore region matrix obtained in Step 3 and the segmentation matrix obtained in Step 2, generate the coarsened mesh element information and connection relationship of the micropore region;

[0011] Step 5: Merge the pore network of the large pore area with the mesh cells of the micropore area to generate the computational mesh file of the digital core hybrid model.

[0012] Furthermore, it also includes step 6, performing numerical simulation calculations using the computational grid file of the digital core hybrid model generated in step 5.

[0013] The technical effects of this invention are:

[0014] While preserving the geometric distribution characteristics and heterogeneity of the micropore region, this invention overcomes the shortcomings of existing direct numerical simulation methods for digital cores, which suffer from numerous computational grid cells and difficulty in conducting full-core numerical simulations. It significantly reduces the number of computational grid cells in the micropore region while ensuring the accuracy of the simulation results, thus substantially improving computational efficiency. Furthermore, this invention has a wide range of applications; for any numerical simulation using image voxels as computational grid cells, this invention can be used to implement controllable multi-level coarsening of the voxel grid. Attached Figure Description

[0015] The accompanying drawings of this invention are described below:

[0016] Figure 1 This is a flowchart illustrating the steps of the invention.

[0017] Figure 2 Core images for an example;

[0018] Figure 3 This is a visualization image of the large-hole area extracted by threshold segmentation in step 1 of the embodiment.

[0019] Figure 4 This is a visualization image of the micropore region extracted by threshold segmentation in step 1 of the embodiment;

[0020] Figure 5 The pore network extracted from the macropore region in the embodiment;

[0021] Figure 6 Image showing the uncoarsed and coarsened micropore regions;

[0022] Figure 7 A comparison chart showing the number of mesh cells calculated for the uncoarsed and coarsened hybrid models;

[0023] Figure 8 This is a comparison chart of simulation results for the uncoarsed and coarsened hybrid models. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0025] like Figure 1 As shown, this invention includes the following steps:

[0026] Step 1: Import three-dimensional core images (such as micro / nano CT scan images) for filtering and noise reduction, and perform multi-threshold segmentation to obtain the distribution information of solid areas, macroporous areas and microporous areas;

[0027] Multi-threshold segmentation is the same as regular threshold segmentation, except that two thresholds are used to extract macropore voxels and micropore voxels from the image respectively; specifically:

[0028] Based on the CT value distribution of the images, thresholds a and b are set, where b > a > 0, and their relationships with each phase are as follows:

[0029] When the pixel CT value is less than a, it is a large hole;

[0030] When the pixel CT value is greater than a and less than b, it is a micropore;

[0031] When the pixel CT value is greater than b, it is a solid.

[0032] Information about macropores and micropores is extracted and stored in two voxel matrices of the same size as the original image. In the macropore matrix, the voxel value for macropores is 1, and the voxel value for the rest (referring to micropores and solids) is 0; in the micropore matrix, the voxel value for micropores is 1, and the voxel value for the rest (referring to macropores and solids) is 0.

[0033] Micropore region matrix and macropore matrix region are used to extract micropore mesh cells and macropore networks.

[0034] Step 2: For the large-pore region voxels obtained in Step 1, the large-pore region is segmented using the watershed algorithm in image processing. Based on the segmentation map, the pore network is extracted, and the relevant geometric information of the pore throat, including volume, specific surface area, and inscribed circle radius, is obtained.

[0035] The literature on watershed algorithms for image segmentation includes: Gostick, Jeff T, Versatile and efficient pore network extraction method using marker-based watershed segmentation[J]. Physical Review E, 2017, 96(2):023307.

[0036] Step 3: Perform multiple convolution calculations on the microporous region voxel matrix data obtained in Step 1, and coarsen the microporous region voxels.

[0037] Step 31: Perform the first convolution calculation on the micropore region matrix obtained in Step 1. The convolution kernel size is 2×2×2, the weight coefficient is 1, and the stride is 2 to obtain the three-dimensional matrix A1. Then, perform threshold segmentation on the A1 matrix to extract the regions that meet the coarsening conditions. Perform erosion calculation on the matrix after threshold segmentation to obtain the first-level coarsening matrix B1.

[0038] Step 32: Perform a second convolution calculation on matrix B1 obtained in step 31. Set the convolution kernel size to 2×2×2, the weight coefficient to 1, and the stride to 2 to obtain a three-dimensional matrix A2. Then, perform thresholding and erosion calculations on A2 to obtain the second-level coarsened matrix B2.

[0039] Step 33: Perform a third convolution calculation on matrix B2 obtained in step 32. Set the convolution kernel size to 2×2×2, the weight coefficient to 1, and the stride to 1 to obtain a three-dimensional matrix A3. Then, perform threshold segmentation and Euclidean distance calculation on A3 to obtain the third-level coarsened matrix B3.

[0040] The threshold segmentation algorithm in steps 31, 32, and 33 is as follows:

[0041]

[0042] In equation (1), A i (x,y,z) represents the convolution value of the identity matrix obtained from the i-th convolution calculation at the coordinate point (x,y,z), and T i To determine A i The threshold set by the condition for whether the convolution values ​​in the matrix satisfy the coarsening condition, i = 1, 2, 3.

[0043] In steps 31 and 32, the erosion algorithm shrinks the regions with non-zero values ​​in the convolution matrix inward. The erosion algorithm uses a 3×3×3 fully set matrix as its structuring element, and its erosion formula is:

[0044]

[0045] In equation (2), B j (x,y,z) represents the convolution value of the identity matrix obtained by the j-th convolution at the coordinate point (x,y,z), and T j To determine A j Whether the convolution values ​​in the matrix satisfy the threshold set by the erosion condition, j = 1, 2.

[0046] In step 33, the Euclidean distance is calculated for the voxels in matrix A3 that satisfy the coarsening condition. The calculation formula is as follows:

[0047]

[0048] In equation (3), x c y c z c Let x be the coordinates of a voxel point in the matrix that has a non-zero value. b y b z b It is the voxel with a value of zero that is closest to this non-zero voxel.

[0049] Step 34: Perform the fourth, fifth and sixth convolution calculations on the micropore region matrix obtained in Step 1. The convolution kernel sizes are 2×2×1, 2×1×2 and 1×2×2, respectively, with a weight coefficient of 1 and a stride of 1, to obtain three-dimensional matrices A4, A5 and A6.

[0050] Each convolutional matrix value maps to a region in the micropore matrix. The higher the number of convolutional layers, the larger the region mapped by the convolutional value. In B3, each convolutional value maps to an 8×8×8 pixel region in the micropore matrix.

[0051] Step 35: Using the matrix B3 obtained in step 33, retrieve each convolution value in the matrix. When it is not 0, mark the voxels in the micropore region matrix mapped by the convolution value, and modify the convolution values ​​of B3, A2, A1, A4, A5, and A6 that contain marked voxels to 0. Repeat the process of judging the convolution value of B3, marking the micropore matrix, and modifying the convolution values ​​of B3, A2, A1, A4, A5, and A6 until the B3 matrix is ​​traversed.

[0052] Labeling is accomplished by modifying the micropore matrix values: in the micropore region matrix, the micropore voxel value is 1. During the traversal of convolution values, when the first convolution value that satisfies the coarsening condition is found, its voxel value mapped in the original image is modified to k+1 (k is the number of macropores in the extracted pore network; in the segmented image obtained by the watershed algorithm, the macropore voxel value ranges from 1 to k); the second is labeled as k+2, and so on. Labeling micropore units starting from k+1 is to extract the connection relationship between macropores and micropores.

[0053] The purpose of setting the convolution value to 0 is to determine that when the convolution value is encountered again, it does not meet the coarsening condition and will not participate in the coarsening process.

[0054] Traverse matrices A2, A1, A4, A5, and A6. When the convolution value is 64 in A2, 8 in A1, and 4 in A4, A5, and A6, mark the voxels of the micropore region matrix mapped by the convolution value accordingly. Modify the convolution value of the mapped region containing the marked voxels in A2, A1, A4, A5, and A6 to 0. Repeat the process of judging, marking the micropore matrix, and modifying the convolution value until all matrices have been traversed.

[0055] The traversal method used for matrix B3 involves searching for non-zero values ​​from largest to smallest. For matrices A2, A1, A4, A5, and A6, the search is performed sequentially along the x, y, and z directions. Convolution values ​​in B3, A2, A1, A4, A5, and A6 that have been modified to 0 are not included in the subsequent coarsening process.

[0056] Traverse the coarsened micropore matrix and mark the voxels with a value of 1, that is, the micropore voxels that did not participate in the coarsening.

[0057] Step 4: Based on the coarsened micropore region matrix obtained in Step 3 and the segmentation matrix obtained in Step 2, generate the coarsened mesh element information and connection relationship of the micropore region;

[0058] The information of the mesh cells in the coarsened microporous region includes the mesh cell size, gray value, contact area between mesh cells, and center point spacing.

[0059] The gray value of the mesh cell in the coarsened microporous region is the average of the gray values ​​of all voxels contained in the original image. This value can be used to estimate material parameters such as porosity and permeability of the mesh cell in the microporous region, and to perform numerical simulation of digital cores in heterogeneous microporous regions.

[0060] Step 5: Merge the pore network of the large-pore region with the mesh cells of the micro-pore region to generate the computational mesh file of the digital core hybrid model;

[0061] The pore network is a simplified representation of the pore structure using spheres (pores) and rods (throats). See [link to relevant documentation]. Figure 5 The structure in the model. The mesh elements are continuous meshes used in finite element and finite volume method simulations.

[0062] Step 6: Perform numerical simulation calculations using the computational grid file of the digital core hybrid model generated in Step 5.

[0063] Example

[0064] Core images such as Figure 2 As shown, the digital core image has a size of 200×200×200 voxels and a resolution of 5.35 μm. The solid grayscale value is 0, the micropore grayscale value is 1, and the macropore solid grayscale value is 2. Based on this digital core image, the following steps are performed:

[0065] Step 1: Import digital core images and obtain voxel information for macropore and micropore regions through multi-point thresholding. Figure 3 and Figure 4 Visualized images of the macropore and micropore regions, respectively. Figure 3 and Figure 4 It can be seen that the macropores have good connectivity, while the micropores are formed into unconnected blocks, filling the spaces between the macropores.

[0066] Information about macropores and micropores is extracted and stored in two voxel matrices of the same size as the original image, forming macropore matrix and micropore matrix respectively.

[0067] Step 2: Extract the pore network of the large-pore region using the watershed segmentation algorithm, such as... Figure 5 As shown, with Figure 3 The comparison shows that the pore network extracted by the watershed algorithm can well characterize the pore structure.

[0068] The geometric information of the pore throat, including volume, specific surface area, and inscribed circle radius, is obtained through a watershed segmentation algorithm.

[0069] Step 3: Perform multiple convolution calculations on the microporous region voxel matrix data obtained in Step 1, and coarsen the microporous region voxels.

[0070] 1) Perform the first convolution calculation on the voxels of the micropore region in the micropore region matrix. The convolution kernel size is 2×2×2, the weight coefficient is 1, and the stride is 2 to obtain the three-dimensional matrix A1. Then, perform threshold segmentation on the A1 matrix to extract the regions that meet the coarsening conditions. Perform erosion calculation on the matrix after threshold segmentation to obtain the first-level coarsening matrix B1.

[0071] 2) Perform a second convolution calculation on the matrix B1 obtained in step 1), with the convolution kernel size set to 2×2×2, the weight coefficient to 1, and the stride to 2, to obtain a three-dimensional matrix A2. Then, perform threshold segmentation and erosion calculation on A2 to obtain the second-level coarsened matrix B2.

[0072] 3) Perform a third convolution calculation on the matrix B2 obtained in step 2), with the convolution kernel size set to 2×2×2, the weight coefficient to 1, and the stride to 1, to obtain a three-dimensional matrix A3. Then, perform threshold segmentation and Euclidean distance calculation on A3 to obtain the third-level coarsened matrix B3.

[0073] 4) Perform the fourth to sixth convolution calculations on the microporous voxels of the microporous region matrix. The convolution kernel sizes are 2×2×1, 2×1×2, and 1×2×2, respectively. The weight coefficients are all 1, and the stride is 1, to obtain the three-dimensional matrices A4, A5, and A6.

[0074] In the uncoarsened microporous region, the size of the mesh cell is 1×1×1 (that is, 1 voxel cell). B3 is used to determine the 8×8×8 (derived from the merging of 512 voxels) coarsened mesh, B2 is used to determine the 4×4×4 coarsened mesh, B1 is used to determine the 2×2×2 coarsened mesh, and A4, A5, and A6 are used to determine the 2×2×1, 2×1×2, and 1×2×2 coarsened meshes, respectively.

[0075] 5) Statistically analyze the convolution values ​​in matrix B3, traversing them from largest to smallest. When a value is not 0, mark the voxels of the micropore region matrix mapped by the convolution value, and modify the convolution values ​​of the mapped regions containing marked voxels in B3, A2, A1, A4, A5, and A6 to 0. Repeat the process of judging the convolution value of B3, marking the micropore matrix, and modifying the convolution values ​​of B3, A2, A1, A4, A5, and A6 until matrix B3 is traversed.

[0076] The matrices A2, A1, A4, A5, and A6 are traversed sequentially along the x, y, and z directions. When the convolution value is 64 in A2, 8 in A1, and 4 in A4, A5, and A6, the corresponding micropore matrix voxels mapped by the convolution values ​​are marked. If the micropore matrix mapped by the convolution values ​​in A2, A1, A4, A5, and A6 contains marked voxels, the convolution values ​​of the mapped regions in A2, A1, A4, A5, and A6 are modified to 0. The process of judging, marking micropore matrices, and modifying convolution values ​​is repeated until all matrices have been traversed.

[0077] The so-called labeling of voxels in the micropore region matrix is ​​actually the reassignment of matrix values ​​in the micropore region matrix. During the traversal of convolution values, when the first convolution value that satisfies the coarsening condition is found, the voxel values ​​mapped to it in the original image are all modified to k+1, the mapping region of the second convolution value is labeled as k+2, and so on.

[0078] The process iterates through the coarsened micropore region matrix, marking each voxel with a value of 1 (i.e., a micropore voxel that did not participate in the coarsening) sequentially. The voxel values ​​are accumulated from the values ​​marked during coarsening.

[0079] Step 4: Perform a union operation on the micropore coarsening marker matrix and the macropore watershed segmentation matrix, traverse the matrix generated by the union, and generate the mesh element information and connection relationship file of the coarsened micropore region;

[0080] The union operation is as follows: when the values ​​at the same position in both the coarsening marker matrix and the large hole watershed segmentation matrix are 0, the value at that position in the merged matrix is ​​0; otherwise, it is a non-zero value in both matrices.

[0081] The connection relationships are obtained as follows: In the original matrix after coarsening, all microporous voxels are traversed, and their values ​​in the six directions (up, down, left, right, front, back) are searched. When a value is 0, it indicates that the adjacent voxel is solid, and no information is stored. When the value is between 1 and k, it indicates that it is connected to a macropore; when it is greater than k, it indicates that the micropores are connected to each other. Furthermore, when two adjacent voxels have the same value, it means that the two voxels are located in the same coarsened cell, and their connection relationship is not stored. These connection relationships are stored in a matrix; by statistically analyzing the connection relationships, the contact area between mesh cells can be obtained.

[0082] The information of the grid cell is obtained by: the coordinates of the cell can be obtained from the position of the grid cell in the matrix; the gray level of the cell can be obtained by combining the coarsening mark with the gray level in the CT image.

[0083] Finally, two matrices are generated. One matrix stores information about the inside of the microporous mesh, such as cell position, cell size, and cell grayscale. The other matrix stores information about the inter-mesh structure, such as connection relationships and contact area between cells.

[0084] Step 5: Merge the mesh of large holes with the mesh of micro holes to generate a coarsened hybrid model data file.

[0085] Mesh merging is achieved by merging the matrices that store mesh information. Before merging, it is necessary to ensure that the number of columns in the merging matrices is the same and that the information type stored in each column is the same. Then, the matrices are merged along the column direction.

[0086] Comparison of before and after coarsening Figure 6 As shown, from Figure 6 As can be seen, the micropore region in the coarsened model exhibits mesh elements with progressively smaller sizes from the center outwards, and the smallest mesh element is located on the surface of the micropore region, with a size of 1 voxel.

[0087] Figure 6 The visualization can only show the size of the cells and their spatial relationships; the detailed data is saved as a txt text file.

[0088] Explanation: Step 4, merging, is to traverse the merged matrix to extract the connections between micropores and between micropores and macropores, as well as other information needed for simulation, such as element locations and element CT values. Step 5, merging, combines the relationships between macropores extracted using the watershed algorithm, along with other simulation-required parameters, with the final micropore-micropore and micropore-macropore relationships obtained in Step 4, along with other parameter information, to obtain a multi-scale geometric model.

[0089] Step 6: Use the generated coarsened mixed model data file to perform single-phase incompressible numerical simulation calculations to calculate the permeability of the sample.

[0090] For the single-phase incompressible numerical simulation calculation of permeability, please refer to the literature "Study on the pore structure characteristics and shale gas seepage mechanism of typical organic-rich shale in Sichuan Basin" [D], Zhang Li et al., University of Chinese Academy of Sciences (Guangzhou Institute of Geochemistry, Chinese Academy of Sciences), 2021, p145.

[0091] To verify the accuracy of this invention, the number of computational units in the coarsened hybrid model generated by this invention was compared with the number in the initial uncoarsened hybrid model. Figure 7 As shown, compared with the uncoarsened hybrid model, the mesh cells in the microporous region invented by this method are reduced by more than 80%, while compared with the total number of elements in the study area, the mesh cells in the coarsened hybrid model are reduced by 98%. Permeability simulation calculations in step 6, as shown... Figure 8As shown, under incompressible flow conditions, the coarsening of the computational grid in the microporous region has a very small impact on pressure distribution and permeability.

Claims

1. A method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations, characterized in that, Includes the following steps: Step 1: Import the original 3D scan image, filter and remove noise, and perform multi-point threshold segmentation to obtain the distribution information of macropore and micropore regions; Step 2: For the large-pore region voxels obtained in Step 1, the large-pore region is segmented using the watershed algorithm in image processing. Based on the segmentation matrix, the pore network is extracted, and the relevant geometric information of the pore throat, including volume, specific surface area, and inscribed circle radius, is obtained. Step 3: Perform multiple convolution calculations on the microporous voxel matrix obtained in Step 1. Iterate through the convolution matrix in turn and coarse out the microporous voxels mapped by the convolution values ​​that meet the coarsening conditions. The multi-convolution computation of voxel data in the micropore region includes: Step 31: Perform the first convolution calculation on the micropore region matrix obtained in Step 1. The convolution kernel size is 2×2×2, the weight coefficient is 1, and the stride is 2 to obtain the three-dimensional matrix A1. Then, perform threshold segmentation on the A1 matrix to extract the regions that meet the coarsening conditions. Perform erosion calculation on the matrix after threshold segmentation to obtain the first-level coarsening matrix B1. Step 32: Perform a second convolution calculation on matrix B1 obtained in step 31. Set the convolution kernel size to 2×2×2, the weight coefficient to 1, and the stride to 2 to obtain a three-dimensional matrix A2. Then, perform thresholding and erosion calculations on A2 to obtain the second-level coarsened matrix B2. Step 33: Perform a third convolution calculation on matrix B2 obtained in step 32. Set the convolution kernel size to 2×2×2, the weight coefficient to 1, and the stride to 1 to obtain a three-dimensional matrix A3. Then, perform threshold segmentation and Euclidean distance calculation on A3 to obtain the third-level coarsened matrix B3. Step 34: Perform the fourth, fifth and sixth convolution calculations on the micropore region matrix obtained in Step 1. The convolution kernel sizes are 2×2×1, 2×1×2 and 1×2×2, respectively. The weight coefficients are all 1 and the stride is 1, resulting in three-dimensional matrices A4, A5 and A6. Step 4: Based on the coarsened micropore region matrix obtained in Step 3 and the segmentation matrix obtained in Step 2, generate the coarsened mesh element information and connection relationship of the micropore region; Step 5: Merge the pore network of the large pore area with the mesh cells of the micropores to generate the computational mesh file of the digital core hybrid model.

2. The method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations as described in claim 1, characterized in that: It also includes step 6, performing numerical simulation calculations using the computational grid file of the digital core hybrid model generated in step 5.

3. The method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations according to claim 1 or 2, characterized in that: in In step 1, the distribution information of the macropore region and the micropore region is stored in two voxel matrices of the same size as the original image, forming the macropore region matrix and the micropore region matrix.

4. The method for coarsening voxel meshes in the micropore region of digital cores based on multiple convolution calculation as described in claim 3, characterized in that, The threshold segmentation algorithm in steps 31, 32, and 33 is as follows: In the formula, A i (x,y,z) represents the convolution value of the identity matrix obtained from the i-th convolution calculation at the coordinate point (x,y,z), and T i To determine A i The threshold set by the condition for whether the convolution values ​​in the matrix satisfy the coarsening condition, i = 1, 2, 3.

5. The method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations according to claim 4, characterized in that, The corrosion calculations in steps 31 and 32 use a 3×3×3 fully set matrix as the structural element, and the corrosion formula is as follows: In the formula, B j (x,y,z) represents the convolution value of the identity matrix obtained by the j-th convolution at the coordinate point (x,y,z), and T j To determine A j Whether the convolution values ​​in the matrix satisfy the threshold set by the erosion condition, j = 1, 2.

6. The method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations as described in claim 5, is characterized in that... In step 33, the formula for calculating the Euclidean distance of voxels in matrix A3 that satisfy the coarsening condition is as follows: In the formula, x c y c z c Let x be the coordinates of a voxel point in the matrix that has a non-zero value. s y s z s The distance from the voxel coordinate point (x c y c z c The most recent voxel point with a value of zero.

7. The method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations as described in claim 6, is characterized in that... The voxels in the microporous region were coarsened and labeled as follows: For each convolution value in matrix B3, if it is not 0, mark the voxels in the micropore region matrix mapped by the convolution value, and modify the convolution values ​​of B3, A2, A1, A4, A5, and A6 whose mapped regions contain marked voxels to 0; repeat the process of judging the convolution value of B3, marking the micropore matrix, and modifying the convolution values ​​of B3, A2, A1, A4, A5, and A6 until matrix B3 is traversed; Traverse matrices A2, A1, A4, A5, and A6 in this manner. When the convolution value is 64 in A2, 8 in A1, and 4 in A4, A5, and A6, mark the voxels of the micropore region matrix mapped by the convolution value accordingly. Modify the convolution values ​​of the mapped regions in A2, A1, A4, A5, and A6 that contain marked voxels to 0. Repeat the process of judging, marking micropore matrices, and modifying convolution values ​​until all matrices have been traversed. Traverse the coarsened and marked micropore region matrix, and mark each voxel in the micropore region matrix with a value of 1 in turn.

8. The method for coarsening voxel meshes in micropore regions of digital cores based on multiple convolution calculations according to claim 7, characterized in that, The traversal method for matrix B3 is to search for non-zero values ​​from largest to smallest, while for matrices A2, A1, A4, A5, and A6, the search is performed sequentially along the x, y, and z directions.

Images