CN114297929A - Modeling method and device for complex ore body with radial basis function surface fused with machine learning - Google Patents

Abstract

The invention provides a machine learning-fused radial basis function curved surface complex ore body modeling method and device, wherein the method comprises the following steps: resampling three-dimensional profile data and giving characteristic attributes to the ore body; training a stacking machine learning model; interpolation encryption of a sparse part with a complex section shape; extracting boundary points of the profile set and corresponding normal vectors; establishing an implicit field by a Hermite radial basis function; and visualizing the three-dimensional ore body model based on a traveling tetrahedral algorithm. The invention realizes the technical effect of establishing a continuous, reliable and smooth-surface three-dimensional ore body model for a curved surface complex ore body by utilizing the characteristic of learning and mining three-dimensional profile data by a stacking machine.

Description

Modeling method and device for complex ore body with radial basis function surface fused with machine learning

technical field

The invention relates to the technical field of three-dimensional ore body modeling, in particular to a method for modeling complex ore bodies with radial basis function surfaces integrated with machine learning.

Background technique

With the rise of digital mines and intelligent mining, the construction and expression of underground mineral resource models has become a research hotspot. The establishment of a 3D ore body model is the basis of a digital mine. An accurate and reasonable 3D ore body model can directly express the spatial form and attribute distribution of the ore body, and provide a reliable basis for professionals to predict underground resources and evaluate reserves.

The traditional 3D ore body modeling adopts the artificial explicit modeling method, and the 3D ore body model established by it has unreasonable morphological distribution and rough surface, and the model quality is not high. Although some modeling cases use implicit modeling methods, most of their researches focus on borehole data, less modeling research on profile image data, and lack of information mining of attribute characteristics of ore body profiles. In addition, artificial intelligence algorithms such as machine learning have played a great role in various fields, but there is currently a lack of research on the application of machine learning in ore body profiles. In addition, the current 3D orebody modeling methods lack the use of attribute features and geometric features of profile data, and the quality of the established 3D orebody models is limited.

SUMMARY OF THE INVENTION

In view of the lack of technology in actual modeling, it is impossible to make full use of the attribute characteristics and geometric characteristics of ore body profile data, and it is difficult to quickly establish a three-dimensional ore body model with reduced manual participation. Starting from the ore body profile data, the present invention proposes a fusion The radial basis surface ore body modeling scheme of the stacking machine learning algorithm, and the method involved in the present invention is especially suitable for ore body modeling of complex sections.

According to one aspect of the present invention, a method for modeling complex ore bodies with radial basis function surfaces fused with machine learning is provided, including:

Step 1. Perform resampling processing on the profile data, and convert it into three-dimensional discrete point data, where the three-dimensional discrete point data includes information extracted from the corresponding layer after the profile is layered by attributes;

Step 2, train the stacking machine learning model with the three-dimensional discrete point data generated after the resampling of the profile data;

Step 3. Use the trained stacking machine learning model to interpolate and encrypt the profile data to construct a new profile set;

Step 4. Extract the boundary points and the corresponding normal vectors in the new section set, and use the boundary points and the corresponding normal vectors to analyze the Hermite-type radial basis function coefficients;

Step 5. Establish the implicit field of the modeling area: determine the boundary of the modeling area, establish the 3D grid nodes in the entire modeling area, and use the resolved Hermite radial basis function to calculate the 3D grid nodes in the entire modeling area. value.

Step 6. Visualize the model: use the traveling tetrahedron algorithm to visualize the implicit field in the modeling area, and establish a three-dimensional ore body model.

Preferably, in step 1, resampling the profile data includes:

Use a rectangle to select the profile data, extract points at equal intervals within the rectangle, and obtain the x, y, and z coordinate values of the points;

Taking the section boundary line as the boundary line, the points inside the boundary line are assigned ore body attribute values, and the points outside the boundary line are assigned non-ore body attribute values;

The corresponding attribute values are added to the extracted point data to form a three-dimensional discrete point data set.

Preferably, in step 2, the three-dimensional discrete point data set is divided into a training set and a test set, firstly, the three base classifiers RF, KNN and XGBoost in the first layer of the stacking machine learning model are trained, and then the three base classifiers are trained using the three base classifiers. The output of the classifier is used as the training data for the second layer of XGBoost to train the second layer of XGBoost.

Preferably, in step 3, a plane to be interpolated in the space is selected, a rectangle is used to select a range, the selected plane range is discretized into three-dimensional discrete points, and the trained stacking model is used to predict the three-dimensional discrete points, and then the The set of points predicted as ore body attributes is converted into profile data, and the converted profile data and the original profile data are fused into a new profile data set.

Preferably, step 5 includes:

S51. Determine the boundary of the entire modeling area, and establish a bounding box that can accommodate the entire modeling area;

S52. Divide a regular grid according to the precision requirement, so that the grid nodes fill the bounding box of the entire modeling area;

S53. Use the parsed Hermite radial basis function to calculate the function value of the grid nodes in the modeling area. Nodes with a node function value less than 0 are inside the ore body model, nodes equal to 0 are at the boundary of the ore body model, and nodes greater than 0 are at the boundary of the ore body model. The nodes are outside the ore body model, so the grid nodes are divided into ore body nodes and non-ore body nodes, and an implicit field in the modeling area is established.

Preferably, in step 6, each three-dimensional grid is divided into 6 tetrahedra, the entire modeling area is divided into meshes according to this method, and the implicit field of the entire modeling area is visualized by using the marching tetrahedron algorithm.

Compared with the prior art, the beneficial effects of the present invention are:

This three-dimensional ore body modeling method, based on the stacking machine learning model, makes full use of the ore body profile data, and uses the attribute data and geometric data of the ore body in the stacking machine learning modeling at the same time, which can quickly establish a reliable, smooth surface and A 3D ore body model with high model quality. It can not only ensure the quality of the model, but also save labor.

Description of drawings

The accompanying drawings, which form a part of the specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention.

FIG. 1 is a schematic diagram of dividing a rectangular discrete point into an ore body internal point and an ore body external point according to a section boundary in an embodiment of the present invention.

2 is a schematic diagram of discrete points whose structures are x, y, z and label in the embodiment of the present invention;

FIG. 3 is a schematic flowchart of a method for modeling a complex ore body with a radial basis function surface by integrating machine learning according to an embodiment of the present invention.

FIG. 4 is a schematic diagram of performing interpolation and encryption on places with complex and sparse cross-sectional shapes in an embodiment of the present invention.

FIG. 5 is a schematic diagram of a modeling region implicit field established in an embodiment of the present invention.

FIG. 6 is a schematic diagram of a three-dimensional ore body model established in an embodiment of the present invention.

Detailed ways

Various exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the invention unless specifically stated otherwise.

Meanwhile, it should be understood that, for the convenience of description, the dimensions of various parts shown in the accompanying drawings are not drawn in an actual proportional relationship.

The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

Techniques, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the authorized description.

In all examples shown and discussed herein, any specific value should be construed as illustrative only and not as limiting. Accordingly, other examples of exemplary embodiments may have different values.

It should be noted that like numerals and letters refer to like items in the following figures, so once an item is defined in one figure, it does not require further discussion in subsequent figures.

Embodiment 1 provides a method for modeling complex ore bodies with radial basis function surfaces that integrates machine learning, which specifically includes the following steps:

Step 1: Data preprocessing stage. Resampling and normalizing the original profile of the complex curved ore body, and converting it into three-dimensional discrete point data, where the three-dimensional discrete point data includes the information of the corresponding layer extracted after stratifying the profile by attributes;

Step 2: Stacking model training phase. The stacking machine learning model is trained using the three-dimensional discrete point data generated after resampling the original profile data. The stacking machine learning model includes two layers of single machine learning models, and the first layer of machine learning models includes three bases: RF, KNN, and XGboost. Classifier, the second-layer machine learning model includes XGBoost;

Using HRBF (Hermite Radial Basis Function) to build a 3D model stage:

Step 3: Use the trained stacking machine learning model to interpolate and encrypt the original profile data, select the complex and sparse part of the original profile data, extract the plane located in the middle of any two profiles, convert the plane into three-dimensional discrete points, and use The stacking machine learning model makes predictions, and fuses the predicted profile data (that is, the virtual profile) with the original profile data to construct a new profile set;

Step 4: Extract the boundary points and the normal vectors corresponding to the boundary points in the new section set, the normal vectors corresponding to the section boundary points all point to the inside of the section, and use the boundary points and the corresponding normal vectors to analyze the Hermite-type radial basis function coefficients;

Step 5: Establish the implicit field of the modeling area: Determine the boundary of the modeling area, divide the 3D grid according to the accuracy requirements, the grid length is the modeling accuracy, establish the 3D grid nodes in the entire modeling area, and use the analyzed The Hermite radial basis function calculates the node values of the 3D grid in the entire modeling area, so as to establish the implicit field of the modeling area;

Step 6: Visualize the model: Use the Marching Tetrahedrons algorithm to visualize the implicit field in the modeling area to establish a three-dimensional ore body model.

In step S1, the profile data is resampled, and the information is extracted hierarchically according to the attributes as follows:

First, perform statistical analysis on all section sizes, select the rectangle that can frame the section with the largest shape in the same plane, and frame each section according to the size of this rectangle. Then, all rectangles are discretized, and the section boundary is extracted. According to the section boundary, the rectangular discrete points are divided into ore body internal points and ore body external points, as shown in Figure 1.

Add ore body attribute values to each three-dimensional discrete point according to the division, and form discrete points with the structure of x, y, z and label, as shown in Figure 2.

In step S2, the specific process steps of using the three-dimensional discrete point data to train the stacking machine learning model are as follows:

S21. Divide the original data set (discrete three-dimensional point data) into a training set (accounting for 80%) and a test set (accounting for 20%);

S22. Divide the training set into 5 parts. During each training, one part is taken as the test set without repetition, and the other 4 parts are used as the training set. According to this classification method, 5 classifiers can be obtained;

S23. For each classification method in S22, use the corresponding test set predicted by the 5 classifiers, and vertically splicing each prediction result (ie, classification probability) in turn, as the feature of the meta-classifier; The test set divided in S21 is predicted, and another feature of the meta-classifier is obtained based on the prediction result;

S24. Two features provided by each classifier can be obtained from S23. The three base classifiers have a total of 6 features. The 6 features are used as the training features of the meta-classifier, and the corresponding real category is used as the training label. The meta-classification The test set of the device is the real label corresponding to the prediction result of the test set divided by S21 in S23;

S25, using the training set and the test set obtained in S24 to train the meta-classifier to obtain the final stacking machine learning model.

In step S3, use the trained stacking machine learning model to interpolate and encrypt the profile data, which specifically includes:

As shown in Figure 4, interpolation and encryption are performed on the areas with complex and sparse cross-section shapes. First, select a place where the profile data is complex and sparse. In the selected area, take a plane in the middle of the two original profiles as the plane to be interpolated, and select the plane to be interpolated according to the rectangle at the same position in step S1. Convert the rectangle after the box selection operation into 3D discrete points, use the trained stacking machine learning model to predict the 3D discrete points, extract the 3D discrete points whose predicted attribute is the ore body, and convert them into new profile data. Combine the new profile data with the original profile data to form a new profile set.

In step S4, establishing the implicit field of the modeling region is realized through the process shown in Figure 5, and the specific steps are:

S41, determine the scope of the modeling area, and establish a three-dimensional grid with a maximum enclosing rule;

S42. Determine the length, width and height of the unit 3D grid according to the modeling accuracy requirements, and the length, width and height correspond to the horizontal modeling accuracy and the vertical modeling accuracy respectively; use the unit 3D grid to fill the entire modeling area, and Extract and store the node coordinate values of all 3D meshes;

S43. Extract the boundary point coordinates of the profile set and the normal vector corresponding to the boundary points at equal intervals, and use them to analyze the Hermite type radial basis function coefficients. The formula is as follows:

分别为第i和第j个空间离散点，i、j为离散点个数；α_i，β_i分别为待求解的隐函数系数，具体可以由以下公式求解：Among them, ψ(x) is the radial basis function of the cubic distance;

are the ith and jth discrete points in space, respectively, i and j are the number of discrete points; α _i , β _i are the implicit function coefficients to be solved, which can be solved by the following formula:

为梯度算子，H为Hessian算子，可以由以下公式计算：Among them, n _j is the normal vector corresponding to the boundary point, expressed as n _x , n _y , n _z ,

is the gradient operator, and H is the Hessian operator, which can be calculated by the following formula:

Among them, H _ij is the partial differential value on x _i and x _j .

S44. Input the coordinates of the three-dimensional grid nodes extracted in S42 into the parsed Hermite radial basis function, and obtain the function value of each three-dimensional grid node. According to the node function value less than 0, the node within the ore body model is equal to According to the rule that nodes with 0 are at the boundary of the ore body model and nodes greater than 0 are outside the ore body model, the grid nodes are divided into ore body nodes and non-ore body nodes, and an implicit field in the modeling area is established.

In step S5, using the traveling tetrahedron algorithm to visualize the regional implicit field established in step S4, the steps of establishing a three-dimensional ore body model are as follows:

The traveling tetrahedron algorithm is used to tetrahedral the 3D grid of all cells in the modeling area, and then each face of the tetrahedron is triangulated and rendered according to the function value of the tetrahedral node to visualize the implicit field of the entire modeling area. As an explicit model, a three-dimensional ore body model is finally established, as shown in Figure 6.

The present invention provides a method for modeling complex ore bodies on radial basis surfaces that integrates machine learning, and has the following advantages:

1) The 3D ore body model is established based on the 3D section, which integrates expert experience modeling, which is different from most modeling methods that use original drilling data;

2) Build a 3D ore body model based on the stacking machine learning algorithm and the Hermite radial basis function. On the one hand, the stacking machine learning algorithm can be used to mine the attribute features of the 3D profile, fully mine the modeling data source, and solve the lack of the Hermite radial basis function. On the other hand, the Hermite radial basis function can be used to automatically and quickly establish a three-dimensional ore body model;

The invention provides a radial basis surface complex ore body modeling method integrating machine learning. According to the information provided by the three-dimensional section, the Hermite-type radial basis function is used as the core, and the stacking machine learning algorithm is used to mine the attribute features of the three-dimensional section. More constraints are provided for establishing implicit fields for Hermite-type radial basis functions, and a 3D ore body model with smooth surface is quickly established by using Hermite-type radial basis functions. The invention combines multiple algorithms, uses different algorithms to have different effects on the modeling data, makes full use of the attribute features and geometric features of the upper three-dimensional profile data, and quickly establishes a reliable model, which provides a new idea for three-dimensional ore body modeling. .

According to the second aspect of the present application, the present invention also provides a complex ore body modeling device with a radial basis function surface fused with machine learning, and the complex ore body modeling device with a radial basis function surface fused with machine learning includes a memory, a processor and a radial basis function ore body modeling program fused with machine learning that is stored in the memory and can be run on the processor, and the radial basis function ore body modeling program fused with machine learning is executed by the processor. The steps of any one of the described method for modeling complex ore bodies with radial basis function surfaces fused with machine learning.

A complex ore body modeling device with radial basis function surface integrating machine learning can run in computing devices such as desktop computers, notebooks, PDAs, and cloud servers. For the device for modeling complex ore bodies with radial basis function surfaces incorporating machine learning, the operable devices may include, but are not limited to, a processor and a memory.

Those skilled in the art can understand that the above example is only an example of an apparatus for modeling a complex ore body with a radial basis function surface fused with machine learning, and does not constitute a model for a complex ore body with a radial basis function surface fused with machine learning. The definition of the modeling device may include more or less parts, or a combination of some parts, or different parts, such as the one that integrates machine learning with the radial basis function surface complex ore body modeling device also It can include input and output devices, network access devices, buses, etc. The so-called processor may be a central processing unit (Central-Processing-Unit, CPU), or other general-purpose processors, digital signal processors (Digital-Signal-Processor, DSP), application-specific integrated circuits (Application-Specific-Integrated). -Circuit, ASIC), Field-Programmable-Gate-Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc., the processor is the control center of the radial basis function surface complex ore body modeling device incorporating machine learning, Various interfaces and lines are used to connect the various parts of the entire operational device. The memory can be used to store the computer program and/or module, and the processor implements the one by running or executing the computer program and/or module stored in the memory and calling the data stored in the memory. Various functions of a complex orebody modeling device for radial basis function surfaces that incorporate machine learning. The memory may mainly include a storage program area and a storage data area, the storage may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart-Media- Card, SMC), Secure-Digital (SD) card, Flash-Card (Flash-Card), at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and scope of the present invention shall be included in the scope of the present invention. within the scope of protection.

It should also be noted that the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device comprising a series of elements includes not only those elements, but also Other elements not expressly listed, or which are inherent to such a process, method, article of manufacture, or apparatus are also included. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article of manufacture or apparatus that includes the element.

Claims (8)

1. A modeling method for a complex ore body with a radial basis function curved surface, which is integrated with machine learning, comprises the following steps:

resampling the original profile data, and converting the processed data into three-dimensional discrete point data;

training a stacking machine learning model by using the three-dimensional discrete point data;

carrying out interpolation encryption on the original profile data by using a trained stacking machine learning model to construct a new profile set;

extracting boundary points and normal vectors corresponding to the boundary points from the profile set, and analyzing a Hermite radial basis function by using the boundary points and the normal vectors;

determining a modeling area boundary, establishing three-dimensional grid nodes in the whole modeling area, calculating the three-dimensional grid node values by utilizing an analyzed Hermite radial basis function, and establishing a modeling area implicit field based on the three-dimensional grid node values;

and carrying out visual operation on the implicit field of the modeling area by utilizing a traveling tetrahedron algorithm to establish a three-dimensional ore body model.

2. The method for modeling the complex ore body with the radial basis function curved surface fused with machine learning according to claim 1, wherein the resampling processing is carried out on the section data, and the method comprises the following steps:

selecting profile data by using a rectangular frame, extracting data points at equal intervals in the rectangle, and acquiring coordinate values of the data points in the x, y and z directions;

using the section boundary line as a boundary line, giving an ore body attribute value to a point inside the boundary line, and giving a non-ore body attribute value to a point outside the boundary line;

and giving the data points of the ore body attribute value and the non-ore body attribute value to form a three-dimensional discrete data set.

3. The method for modeling the radial basis function curved surface complex ore body fused with machine learning according to claim 1, wherein the step of training the stacking machine learning model by using the three-dimensional discrete point data comprises the following steps:

training three base classifiers of RF, KNN and XGboost in a first layer in a stacking machine learning model by using the three-dimensional discrete point data;

and (4) taking the output of the three base classifiers as training data of a second layer XGboost in the stacking machine learning model, and training the second layer XGboost to obtain the stacking machine learning model.

4. The method for modeling the complex ore body with the radial basis function curved surface by fusing machine learning according to claim 1, wherein the method for interpolating and encrypting the original profile data by using the trained stacking machine learning model to construct a new profile set comprises the following steps:

taking the middle plane of any two original sections as a plane to be interpolated in a section area where original section data is complex in shape and sparse, and converting the plane to be interpolated into new three-dimensional discrete point data;

predicting the new three-dimensional discrete point data by using a trained stacking machine learning model, and extracting the prediction attribute of the prediction result as the three-dimensional discrete point of the ore body;

and converting the three-dimensional discrete points of the ore body into new profile data, and fusing the new profile data and the original profile data to construct a new profile set.

5. The method for modeling a complex ore body with a curved surface based on radial basis functions fused with machine learning according to claim 1, wherein the method for modeling the complex ore body with the curved surface based on radial basis functions fused with machine learning comprises the steps of determining the boundary of a modeling area, establishing three-dimensional grid nodes in the whole modeling area, calculating the node values of the three-dimensional grid nodes by using the resolved radial basis functions of Hermite type, and establishing an implicit field of the modeling area based on the node values of the three-dimensional grid nodes, and comprises the following steps:

determining the boundary of the whole modeling range, and establishing a bounding box capable of containing the whole modeling area;

dividing a regular grid according to the precision requirement, and filling the bounding boxes of the whole modeling area with grid nodes;

calculating a function value of a grid node of the modeling area by using the analyzed Hermite radial basis function;

dividing the grid nodes into ore body nodes and non-ore body nodes based on the function values;

and establishing a modeling area implicit field based on the divided ore body nodes and the non-ore body nodes.

6. The method according to claim 5, wherein the criterion for dividing the mesh nodes into the mineral nodes and the non-mineral nodes based on the function values comprises:

the node with the function value smaller than 0 is positioned in the ore body model;

the node with the function value of 0 is positioned on the boundary of the ore body model;

nodes with function values greater than 0 are outside the ore body model.

7. The method for modeling the complex ore body with the radial basis function curved surface fused with machine learning according to claim 1, wherein the hidden field of the modeling area is visualized by using a traveling tetrahedron algorithm, and the building of the three-dimensional ore body model comprises the following steps:

dividing each three-dimensional grid into 6 tetrahedrons, dividing the grid of the whole modeling area according to the method, and visualizing the hidden field of the whole modeling area by utilizing a traveling tetrahedron algorithm, thereby forming a three-dimensional ore body model.

8. A fusion machine-learned radial basis function surface complex ore body modeling apparatus comprising a memory, a processor, and a fusion machine-learned radial basis function ore body modeling program stored on the memory and executable on the processor, the fusion machine-learned radial basis function ore body modeling program when executed by the processor implementing the steps of the fusion machine-learned radial basis function surface complex ore body modeling method of any one of claims 1 to 7.

Images