Physical property equivalent conversion method from karst geologic model to 3D printing physical model

Abstract

A physical property equivalent conversion method from a karst geologic model to a 3D printing physical model comprises the following steps that 1, overall data of an existing karst geologic model are finely analyzed, step length information is classified and arranged based on attributes of the karst geologic model, and point cloud data are exported; step 2, adopting a Deep Marking Cubes algorithm, carrying out three-dimensional reconstruction according to classification and size, and preliminarily obtaining a three-dimensional model; 3, the model is subjected to surface repairing, merging calculation, scaling, cutting, assembling and other operations, and an accurate 3D printing model is obtained; and 4, the'shell 'is filled based on the gravity factor filling model, and finally a 3D printing model of the karst geologic model is obtained. According to the karst geologic model 3D printing conversion method system based on gravity factor filling, the three-dimensional structure of oil reservoir geology is better represented, and the actual requirement of the model is better met.

Description

Technical Field

[0001] The present invention relates to the field of 3D model conversion, and in particular to a method for equivalently converting the physical properties of a large composite karst geological model into a 3D printed physical model. Background Art

[0002] Fracture-vuggy carbonate reservoirs are diverse in type, complex in spatial structure, and highly heterogeneous. Current development experience shows that oil and gas are mainly enriched in dissolution pores and caves.

[0003] The analysis of reservoir types and the evaluation of spatial combination models are of great significance for the development of such oil and gas fields. Categorizing reservoir types into three categories: fracture-cavity, cave, and fracture, and establishing an idealized reservoir seepage model can provide a reference for identifying the spatial structure of fracture-cavity carbonate reservoirs.

[0004] To support the construction of large-scale fracture-cavity physical models indoors, geological models of composite karst and fault-controlled karst need to be converted into digital models suitable for 3D printing. However, the compatibility issue between reservoir geological models and 3D printer digital models has not yet been resolved. Summary of the Invention

[0005] In order to solve the above problems in the prior art, the present invention provides a method for equivalent conversion of physical properties of a karst geological model to a 3D printed physical model.

[0006] The technical solutions of the present invention are as follows:

[0007] A method for equivalent conversion of physical properties of a karst geological model to a 3D printed physical model, characterized by comprising the following steps:

[0008] Step 1: Analyze the existing karst geological model data in detail, classify and organize the step information based on the karst geological model attributes, and export the point cloud data;

[0009] Step 2: Use the Deep Marching Cubes algorithm to perform 3D reconstruction by classification and size to obtain a preliminary 3D model;

[0010] Step 3: Perform surface repair, merging calculations, scaling, cutting, and assembly on the model to obtain a relatively accurate 3D printing model;

[0011] Step 4: Fill the “shell” based on the gravity factor filling model, and finally obtain a 3D printed model of the karst geological model.

[0012] Preferably, in step 1, the attribute types of the karst geological model include: caves, holes, fractures (cracks); the step length information of the karst geological model in the X and Y directions is concentrated and dispersed in the Z direction to form different size types.

[0013] Preferably, in step 2, the three-dimensional reconstruction method performs a combined addition calculation on models with the same attributes and different step sizes, and converts them into triangular surface models.

[0014] Preferably, in step 2, the Deep Marching Cubes network separates mesh topology from geometry, and the network consists of an encoder and a two-branch decoder, wherein the first branch of the two-branch decoder is to predict the occupancy probability of each voxel.

[0015] Further preferably, the input of the network is point clouds, which are obtained through the Encoder-Decoder structure (Occuapancy Value) and (vertex coordinates, vertex displacements), where O∈[0,1] represents the probability value of each point being on the surface. Then the probability distribution of each point for whether it is occupied is:

[0016] p n (t)=(o n ) t (1-O n ) 1-t .

[0017] Note: "Occupancy Value" is a term often used to describe the occupancy of voxels or grid cells in three-dimensional space. In computer graphics and computer-aided design, "Occupancy Value" usually indicates whether a voxel or grid cell is occupied, and the degree of occupancy. This representation is often used in three-dimensional modeling, volume rendering, and virtual environments. The value of Occupancy Value is usually binary, indicating whether there is an object occupying the location, or it may be a continuous range of values, indicating the degree of occupancy or density.

[0018] Vertex coordinates are the position of a point in three-dimensional space, typically represented by (x, y, z). In computer graphics, vertex coordinates are used to define the shape and position of a three-dimensional model. By manipulating vertex coordinates, effects such as deformation, rotation, and scaling of the model can be achieved. These coordinates are also used to define the vertices of triangular meshes, thereby constructing complex three-dimensional models. In graphics rendering, vertex coordinates are the foundation for constructing the entire scene, determining the appearance and position of the model.

[0019] "Vertex displacement" is a term in computer graphics that refers to creating deformation effects by changing the positions of vertices in a 3D model. In graphics rendering, various visual effects, such as terrain changes, can be achieved by displacing, scaling, and rotating vertices. This technique is primarily used in real-time rendering development.

[0020] Further preferably, the grid topology is implicitly defined by the states of the occupancy variables at its corners, and the second branch of the decoder predicts the vertex positions of each edge of each cell, and the combination of the implicitly defined topology and vertex positions defines the distribution on the grid.

[0021] Further preferably, the size data of the "body" in the three-dimensional reconstruction of the Deep Marching Cubes algorithm is defined. The data is the relevant data of the corresponding step length in the point cloud model of the previous step. The body models of caves, holes and fracture attributes with different step lengths are merged and added, and converted into triangular surface models to finally obtain a preliminary three-dimensional model.

[0022] Preferably, in step 3, the specific steps of repairing include repairing the erroneous triangular faces of the model, repairing the triangular faces of the hollow holes in the shell model, and repairing the reverse triangular facets.

[0023] Preferably, in step 4, the gravity factor filling model is a probability model based on a two-dimensional copula, and the gravity factor parameters directly affect the density distribution of the filling material.

[0024] Further preferably, the establishment of the joint probability distribution model of the relevant parameters based on the gravity factor includes two steps: the first step is to establish the marginal distribution function of the relevant parameters based on the gravity factor; the second step is to select the optimal Copula function to describe the correlation structure between the relevant parameters based on the gravity factor;

[0025] The AIC criterion commonly used in engineering is used to identify the optimal marginal distribution function. That is, the marginal distribution function with the minimum AIC value is considered to be the marginal distribution function that best fits the probability distribution characteristics of the original observation data. The AIC value is defined as the sum of twice the sum of the logarithms of the probability density function values ​​at the original observation data points of the variable and twice the number of probability density function distribution parameters. The expression is:

[0026]

[0027] Where: k2 is the number of related parameters in the Copula function, and for a two-dimensional Copula function, k2 = 1; (u 1i ,u 2i ), is the original observation data (x 1i ,x 2i ) is calculated by the following formula:

[0028]

[0029] Where: rank(x 1i ) and rank(x 2i ) are the measured values ​​(x 1i ) and (x 2i ) In the whole column of observation data x1={x 11 ,x 12 ,…,x 1N} and x2={x 21 ,x 22 ,…,x 2N} in the rank.

[0030] The beneficial technical effects of the present invention are as follows:

[0031] The present invention provides a method for converting physical properties equivalently from a large-scale composite karst geological model to a 3D-printed physical model. The method first meticulously analyzes the existing karst geological model's overall data, organizes step information based on the karst geological model's attribute classification, and exports point cloud data. The method then uses the Deep Marching Cubes algorithm to perform three-dimensional reconstruction by classification and size, initially obtaining a three-dimensional model. The model is then subjected to surface repair, merging calculations, scaling, cutting, and assembly operations to obtain a relatively accurate 3D-printed model. Finally, the model is filled with a "shell" based on the gravity factor, and the model is repaired to ultimately obtain a 3D-printed model of the karst geological model. The method of the present invention allows for relatively accurate 3D-printed digital models of geological models, which is of great significance for numerical simulation research on reservoir geology.

[0032] Specific:

[0033] (1) Through step-by-step processing and surface repair, merging calculation, scaling, cutting, assembly and other operations on the model, and then filling the model with gravity factors, a set of karst geological model 3D printing conversion method system based on gravity factor filling was constructed. This method system better represents the three-dimensional structure of the reservoir geology and better meets the actual needs of the model.

[0034] (2) The 3D printed digital model of the karst geological model obtained by the present invention has high accuracy and precision, and can be used to directly participate in geological body parameter modeling.

[0035] (3) Economical and saves manpower and physics.

[0036] (4) It is easy to adjust and optimize parameters, and a large amount of research can be carried out in a short period of time. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1Flowchart of a 3D printing conversion method for a karst geological model based on gravity factor filling in an embodiment of the present invention;

[0038] Figure 2 This is a karst geological model diagram of Shunbei in an embodiment of the present invention;

[0039] Figure 3 This is a diagram showing the basic unit analysis results of the Shunbei karst geological model in an embodiment of the present invention;

[0040] Figure 4 9 point cloud model images derived from the three attributes of the Shunbei karst geological model at Z=5, 25, and 50 in the embodiment of the present invention;

[0041] Figure 5 This is a diagram of the Deep Marching Cubes network structure in an embodiment of the present invention;

[0042] Figure 6 This is an STL model diagram of the three-dimensional reconstruction of the cave attributes of the Shunbei karst geological model in an embodiment of the present invention;

[0043] Figure 7 This is an STL model diagram of the three-dimensional reconstruction of the hole attributes of the Shunbei karst geological model in an embodiment of the present invention;

[0044] Figure 8 This is an STL model diagram of the Shunbei karst geological model after three-dimensional reconstruction of fracture attributes in an embodiment of the present invention;

[0045] Figure 9 This is a diagram of the triangular patch repair process of the Shunbei karst geological model in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described in detail below through specific embodiments, but this does not limit the technical solution of the present invention. All changes or equivalent substitutions based on the present invention should fall within the scope of protection of the present invention.

[0047] Example

[0048] The embodiment of the present invention provides a method for equivalent conversion of physical properties from a large composite karst geological model to a 3D printed physical model, using the karst geological model of Shunbei Oilfield as an example. Figure 1 As shown, the following steps are included:

[0049] Step 1: Analyze the existing karst geological model data in detail, classify and organize the step information based on the karst geological model attributes, and export point cloud data;

[0050] Step 2: Use the Deep Marching Cubes algorithm to perform 3D reconstruction by classification and size to obtain a preliminary 3D model;

[0051] Step 3: Perform surface repair, merging calculations, scaling, cutting, and assembly on the model to obtain a relatively accurate 3D printing model;

[0052] Step 4: Fill the “shell” based on the gravity factor filling model, and finally obtain a 3D printed model of the karst geological model.

[0053] The specific steps are as follows:

[0054] Step 1: Based on the karst geological model of Shunbei (such as Figure 2 As shown), the overall data of the model is carefully analyzed, the step information is classified and sorted based on the karst geological model attributes, and the point cloud data is exported. Through the analysis of the basic unit of the Shunbei fault-controlled karst geological model, the step lengths in the X and Y directions are relatively concentrated, and the Z direction is mainly concentrated in three value segments. The step length data in the three directions of the basic unit are mainly concentrated at 45 meters in the X direction, 46 meters in the Y direction, and the Z direction step data are relatively scattered, concentrated at three values ​​​​of 3, 24, and 48 respectively, as shown in the figure below. Figure 3 In order to make the size of the "body" more accurate, the different data in the Z direction are exported into three point cloud models according to Z=5, 25, and 50, as shown in the figure. Figure 4 shown.

[0055] Step 2: Use the Deep Marching Cubes algorithm to perform 3D reconstruction by classification and size to obtain a preliminary 3D model.

[0056] Traditional Marching Cubes methods are not differentiable and cannot be directly embedded into existing deep learning methods for end-to-end training. The Marching process of Marching Cubes only involves a single square. Even if it can be differentiated, its gradient cannot spread to other surfaces. Deep Marching Cubes is an end-to-end trainable network with a structure like Figure 5 As shown in Figure 2, it can predict explicit surface representations of arbitrary topologies. The Deep Marching Cubes network separates mesh topology from geometry. The network consists of an encoder and a two-branch decoder.

[0057] Instead of predicting a signed distance value, the first branch of the two-branch decoder predicts the occupancy probability of each voxel.

[0058] The input of the network is point clouds, which are obtained through the Encoder-Decoder structure. (Occuapancy Value) and (vertex coordinates, referred to as vertex displacements in this article). Where O∈[0,1] represents the probability value of each point being on the surface. Then the probability distribution of each point for whether it is occupied is:

[0059] p n (t)=(o n ) t (1-O n ) 1-t .

[0060] The grid topology is then implicitly (and probabilistically) defined by the states of the occupancy variables at its corners. The second branch of the decoder predicts the vertex positions for each edge of each cell. The combination of the implicitly defined topology and vertex positions defines a distribution over the grid that is differentiable and can be used for backpropagation.

[0061] Based on the above theory, the size data of the "body" in the 3D reconstruction of the Deep Marching Cubes algorithm is defined. The data is the relevant data of the corresponding step length in the point cloud model in the previous step. The volume models of caves, holes and fractures with different step lengths are merged and added, and converted into triangular surface models, and finally a preliminary 3D model is obtained, such as Figure 6-Figure 8 shown.

[0062] Step 3: Perform surface repair, merge calculation, scaling, cutting, assembly and other operations on the model to obtain a more accurate 3D printing model.

[0063] The work of repairing model triangle surface errors requires repeated inspection and repair operations until there are no errors. The specific steps are: repairing model error triangle surfaces, repairing the hollow hole triangle surfaces of the shell model, and repairing the reverse triangle surface patches. The repair results of each stage are as follows: Figure 9 Other operations are basic software operations and do not need to be explained in detail.

[0064] Step 4: Fill the “shell” and repair the model based on the gravity factor filling model, and finally obtain a 3D printed model of the karst geological model.

[0065] This method is based on the probability model of two-dimensional Copula, that is, filling the model in both horizontal and vertical directions, and converting the relevant information of the geological model into a 3D printed model as completely as possible. For the two-dimensional case, the Copula function is defined as [0,1] 2 The marginal distribution in space is a two-dimensional joint probability distribution function uniformly distributed in the interval [0,1]:

[0066] F(x1,x2)=C(F1(x1),F2(x2);θ)=C(u1,u2;θ)

[0067] Where F(x1,x2) is the joint probability distribution function of variables X1 and X2; u1 = F1(x1) and u2 = F2(x2) are the marginal distribution functions of variables X1 and X2 respectively; C is the copula function; θ is the relevant parameter of the copula function. Accordingly, if the probability density functions f1(x1) and f2(x2) of variables X1 and X2 are known, then by taking the derivative of both sides of the above equation F(x1,x2), the joint probability density function f(x1,x2) of variables X1 and X2 can be obtained as:

[0068] f(x1,x2)=f1(x1)f2(x2)c(F1(x1),F2(x2);θ)

[0069] Where, is the Copula density function. Therefore, if the marginal distribution function and Copula function of variables X1 and X2 are known, a two-dimensional joint probability distribution model of variables X1 and X2 can be established using F(x1,x2) and f(x1,x2). The correlation parameter θ of the Copula function can be directly calculated from the Kendall rank correlation coefficient τ between variables X1 and X2 using the following formula:

[0070]

[0071] Therefore, when the Kendall rank correlation coefficient τ between the variables X1 and X2 is known, the parameter θ can be obtained by the nonlinear integral equation shown in the above expression.

[0072] From the above formula, we can know that the establishment of the joint probability distribution model of the relevant parameters based on the gravity factor includes two steps: the first step is to establish the marginal distribution function of the relevant parameters based on the gravity factor; the second step is to select the optimal Copula function to describe the correlation structure between the relevant parameters based on the gravity factor. As mentioned above, the above two steps can be carried out separately and independently. The following are the identification methods of the optimal marginal distribution function and the optimal Copula function of the relevant parameters based on the gravity factor. This paper adopts the AIC criterion commonly used in engineering to identify the optimal marginal distribution function, that is, the marginal distribution function with the minimum AIC value is considered to be the marginal distribution function that best fits the probability distribution characteristics of the original observation data. The AIC value is defined as the sum of twice the sum of the logarithms of the probability density function values ​​at the original observation data points of the variable and twice the number of probability density function distribution parameters. The expression is:

[0073]

[0074] Where: k2 is the number of related parameters in the Copula function, and for a two-dimensional Copula function, k2 = 1; (u 1i ,u 2i ), is the original observation data (x 1i ,x 2i ) can be calculated by the following formula:

[0075]

[0076] Where: rank(x 1i ) and rank(x 2i ) are the measured values ​​(x 1i ) and (x 2i ) In the whole column of observation data x1={x 11 ,x 12 ,…,x 1N} and x2={x 21 ,x 22 ,…,x 2N Therefore, given a set of observational data with relevant parameters based on the gravity factor, the AIC values ​​of multiple candidate copula functions can be calculated, thereby identifying the copula function that best fits the correlation structure of the original observational data.

[0077] Based on the above theory, after measuring the data of the filling material in the horizontal and vertical directions, this paper selects Gaussian, Plackett, Frank and No.16 Copula functions as alternative Copula functions to fit the correlation structure between the related parameters based on the gravity factor. At this time, the Copula function with the smallest AIC value can be used as the probability distribution function of the density of the filling material in the horizontal and vertical directions.

[0078] The present invention provides a method for converting physical properties equivalently from a large composite karst geological model to a 3D printed physical model. The method first analyzes the overall data of the existing karst geological model, classifies and organizes step information based on the karst geological model attributes, and exports point cloud data. The method then uses the Deep Marching Cubes algorithm to perform three-dimensional reconstruction by classification and size to initially obtain a three-dimensional model. The model is then subjected to surface repair, merging calculations, scaling, cutting, and assembly operations to obtain a relatively accurate 3D printed model. Finally, the model is filled with a "shell" based on the gravity factor filling, and the model is repaired, ultimately obtaining a 3D printed model of the karst geological model. That is, the present invention constructs a set of karst geological model 3D printing conversion method systems based on gravity factor filling through step-by-step processing and surface repair, merging calculations, scaling, cutting, and assembly operations on the model, and then gravity factor filling on the model. This method system better characterizes the three-dimensional structure of reservoir geology and better meets the actual needs of the model.

[0079] In addition, the 3D printed digital model of the karst geological model obtained by the present invention has high accuracy and precision, and can be used to directly participate in geological body parameter modeling; it is economical and saves manpower and physics; it is easy to adjust and optimize parameters, and a large amount of research can be carried out in a short period of time.

[0080] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that are not conceived through creative work should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection defined in the claims.

Claims

1. A method for equivalent conversion of physical properties from a karst geological model to a 3D printed physical model, characterized in that The following steps are involved: Step 1: Analyze the existing karst geological model data in detail, classify and organize the step information based on the karst geological model attributes, and export point cloud data; Step 2: Use the Deep Marching Cubes algorithm to perform 3D reconstruction by classification and size to obtain a preliminary 3D model; Step 3: Perform surface repair, merging calculations, scaling, cutting, and assembly on the model to obtain a relatively accurate 3D printing model; Step 4: Fill the "shell" based on the gravity factor filling model, and finally obtain a 3D printed model of the karst geological model.

2. The method according to claim 1, characterized in that In step 1, the attribute types of the karst geological model include: caves, holes, and fractures (cracks); the step length information of the karst geological model in the X and Y directions is concentrated, and is dispersed in the Z direction to form different size types.

3. The method according to claim 1, characterized in that In step 2, the three-dimensional reconstruction method performs a combined addition calculation on models with the same attributes and different step sizes, and converts them into triangular surface models.

4. The method according to claim 1, characterized in that In step 2, the Deep Marching Cubes network separates mesh topology from geometry. The network consists of an encoder and a two-branch decoder, where the first branch of the two-branch decoder predicts the occupancy probability of each voxel.

5. The method according to claim 4, characterized in that The input of the network is point clouds, which are obtained through the Encoder-Decoder structure. and Where O∈[0,1] represents the probability value of each point being on the surface, then the probability distribution of each point for whether it is occupied is: p n (t)=(o n ) t (1-O n ) 1-t 。 6. The method according to claim 5, characterized in that The grid topology is implicitly defined by the states of the occupancy variables at its corners. The second branch of the decoder predicts the vertex positions of each edge for each cell. The combination of the implicitly defined topology and vertex positions defines the distribution on the grid.

7. The method according to claim 6, characterized in that The size data of the "body" in the 3D reconstruction of the Deep Marching Cubes algorithm is defined. The data is the relevant data of the corresponding step length in the point cloud model in the previous step. The body models of caves, holes, and fractures with different step lengths are merged and added, and converted into triangular surface models to finally obtain a preliminary 3D model.

8. The method according to claim 1, characterized in that In step 3, the specific steps of repair include repairing the wrong triangles of the model, repairing the triangles of the hollow holes in the shell model, and repairing the reverse triangles.

9. The method according to claim 1, characterized in that In step 4, the gravity factor filling model is a probability model based on a two-dimensional copula, and the gravity factor parameters directly affect the density distribution of the filling material.

10. The method according to claim 9, characterized in that The establishment of the joint probability distribution model of the relevant parameters based on the gravity factor includes two steps: the first step is to establish the marginal distribution function of the relevant parameters based on the gravity factor; the second step is to select the optimal Copula function to describe the correlation structure between the relevant parameters based on the gravity factor; The AIC criterion commonly used in engineering is used to identify the optimal marginal distribution function. That is, the marginal distribution function with the minimum AIC value is considered to be the marginal distribution function that best fits the probability distribution characteristics of the original observation data. The AIC value is defined as the sum of twice the sum of the logarithms of the probability density function values ​​at the original observation data points of the variable and twice the number of probability density function distribution parameters. The expression is: Where: k2 is the number of related parameters in the Copula function, and for a two-dimensional Copula function, k2 = 1; (u 1i ,u 2i ), is the original observation data (x 1i ,x 2i ) is calculated by the following formula: Where: rank(x 1i ) and rank(x 2i ) are the measured values ​​(x 1i ) and (x 2i ) In the whole column of observation data x1={x 11 ,x 12 ,…,x 1N } and x2={x 21 ,x 22 ,…,x 2N } in the rank.

Images