Three-dimensional modeling method for geological section

Abstract

The invention relates to the technical field of geological image processing, and particularly discloses a three-dimensional modeling method for a geological section, which is used for solving the problem of how to accurately identify and hide irrelevant geological areas and improve the modeling efficiency in geological three-dimensional modeling. Comprising the steps of data acquisition and preprocessing, geologic model constraint, implicit modeling and machine learning optimization, three-dimensional modeling and graphic rendering optimization, model correction and optimization and dynamic interaction and analysis, and four analysis formulas based on hidden requirements are provided; according to the method, by setting the hiding demand judgment formula for the coal seam, the rock stratum, the mined area and the inactive fault, the geological area irrelevant to target research is screened, the geological three-dimensional modeling process is optimized, the accuracy and efficiency of the model are improved, interference of irrelevant data is reduced, and the method has good application prospects. It is ensured that the mineral resource evaluation result better conforms to the actual geological condition, and the precision and credibility of mineral resource prediction are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of geological image processing, and more particularly to a three-dimensional modeling method for geological sections. Background Art

[0002] In existing technologies, 3D lithologic modeling relies on obtaining surface models from a 2D geological information database, and is completed through human-computer interaction interpretation based on geophysical inversion and rock property statistics. 3D fracture modeling, as the primary component of 3D lithologic modeling, uses fracture boundaries on geological maps and sections as geological elements, and human-computer interaction interpretation based on various geophysical information gradient zones as geophysical elements. In recent years, machine learning and deep learning algorithms have been widely used in 3D lithologic modeling, primarily including unsupervised clustering algorithms and supervised classification algorithms. While unsupervised clustering algorithms can improve modeling efficiency, they do not consider existing geological constraints. Supervised classification algorithms offer faster computational speeds and can complete geophysical interpretations within existing geological constraints, making them highly advantageous for 3D implicit lithologic modeling. However, when there is an imbalance in rock lithologic categories, failure to adjust the training samples and algorithms of the supervised classification algorithm can affect its prediction accuracy. To this end, researchers in this field have proposed a cost-sensitive positive sample unlabeled learning algorithm based on the "bagging method" to address the extreme imbalance problem of having only a small number of positive samples in the quantitative prediction of mineral resources. This algorithm is primarily used for the quantitative prediction of gold mineral resources. Compared to gold mines, coal mines have regionally stable distribution, with little thickness variation but a wide range of extension. Coal-bearing strata are multi-layered, with frequent thickness variations, layered sedimentary characteristics, and rapidly changing lithologic differences. As a result, existing machine learning models also face problems during training, such as imbalanced lithologic categories, large differences in the proportion of data samples for coal seams or interbedded waste layers, and missing coal seam drillhole data in local areas. Furthermore, coal-bearing strata have local areas with gentle overall dips, dense fault zones, and the combined effects of faults and folds. This raises concerns about changes in coal seam thickness, coal seam fragmentation, and gas enrichment caused by fault dislocation. Therefore, improving on publicly available machine learning methods to address the current problems with 3D modeling of geological profiles of coal seams using machine learning is a current challenge. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a three-dimensional modeling method for geological profiles. By setting hidden demand judgment formulas for coal seams, rock strata, mined areas and inactive faults, geological areas irrelevant to the target research are accurately screened, thereby optimizing the geological three-dimensional modeling process, improving the accuracy and efficiency of the model, reducing the interference of irrelevant data, ensuring that the mineral resource assessment results are more in line with actual geological conditions, and effectively improving the accuracy and credibility of mineral resource predictions.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A three-dimensional modeling method for geological sections comprises the following steps:

[0006] Step 1: Data acquisition and preprocessing: Combine the geological information data source library and geophysical data, use machine learning methods to extract geological structural constraints, and perform human-computer interactive interpretation by combining two-dimensional geological information and geophysical information gradient zones;

[0007] Step 2: Constraining the geological model: Using the geological structure data in the 2D geological database, a linear model is used to constrain the constructed 3D geological model. Lithologic modeling is performed. Combining stratigraphic data and physical properties, a surface model is used to establish a planar geological contact surface based on the lithologic data. The fracture surfaces are constrained using the fracture boundaries on the geological profile, and fractures that do not affect the structure are removed during the visualization phase.

[0008] Step 3: Implicit modeling and machine learning optimization: Use an unsupervised clustering algorithm to initially stratify and classify geological entities, automatically monitor the spatial distribution of lithology categories, and use a supervised classification algorithm combined with labeled drillhole data to further improve the classification model.

[0009] Step 4: 3D modeling and graphics rendering optimization: Voxel technology is used to render the underground lithology. The transition and contact relationships between different underground rock layers are represented through volume data. Different lithology layers are visualized through color mapping and transparency settings. An algorithm based on viewpoint selection and geometric visibility is used to delete identified areas that need to be hidden during the 3D image generation process.

[0010] Step 5: Model calibration and optimization: Adjust the model's 3D lithologic layers and fracture surfaces based on actual geological profile data to ensure they conform to geological conditions. Based on geophysical inversion results, calibrate the model's rock property data.

[0011] Step 6: Dynamic interaction and analysis: Using modern virtual reality technology, dynamic interaction is performed on the three-dimensional geological model. Through user interactive control, geological profiles at different depths can be viewed in real time and ore body distribution information can be analyzed.

[0012] As a further solution of the present invention, the process of data acquisition and preprocessing in step 1 includes:

[0013] Step 21, data acquisition: acquiring geological profile data, geological maps and geological structure maps, groundwater and soil information, and production exploration data from a geological information data source database. The geological profile data includes but is not limited to borehole data, stratum thickness, lithology information, and mineral composition. The geological maps and geological structure maps include fault location, fold, and lithology distribution data. The groundwater and soil information includes but is not limited to hydrogeology and groundwater flow. The production exploration data includes but is not limited to mining information and ore body analysis information. Gravity data, magnetic data, electromagnetic data, and seismic data are acquired from a geophysical database. The seismic data includes but is not limited to lithology and fault analysis information obtained through seismic wave velocity analysis.

[0014] Step 22, data storage and integration: Establish a unified data repository, use geographic information systems to store and manage spatial data, integrate data from different sources, ensure spatial consistency between data, and convert them into a unified format (the unified format requires normalization and standardization of the acquired data to facilitate data processing and analysis).

[0015] As a further solution of the present invention, in step 22, it is necessary to unify the formats of data from different sources into data in numerical format, and to perform k-means clustering algorithm on the data in the same format for standardization, and then use the minimum-maximum normalization method to normalize the standardized data.

[0016] As a further solution of the present invention, the specific implementation process of step 2 for geological model constraint is as follows:

[0017] Step 31, feature extraction and structural constraints: Extract the bottom layer thickness, lithology information, the location, strike and dip of geological faults, mineral composition and spatial distribution based on geological maps and profiles; extract data related to gravity anomalies and underground density distribution, data related to magnetic anomalies and rock magnetism, data related to seismic wave velocity and lithology, geological structure, and electrical conductivity based on gravity, magnetic and seismic wave data; combine geological profiles and geophysical data, and constrain the spatial distribution of different lithology areas based on the physical characteristics of different rock layers, combined with seismic wave and gravity anomaly data; infer the geometry of the fault based on the strike and dip of the fault and combined with seismic reflection wave data; and construct the geometry of the stratum contact surface based on the mineral composition, lithology and fault information and seismic wave reflection data;

[0018] Step 32, human-computer interactive interpretation: Use Web GIS to provide geological experts with an intuitive data viewing and editing interface, allowing them to annotate, edit, and confirm unclear areas of the data. The machine learning model is optimized based on expert feedback, and the data source and model of the machine learning model are updated based on the newly added annotations during the human-computer interaction process.

[0019] Step 33, data integration and output: Merge geological data and geophysical data, integrate all constraints and form the final input data based on the machine learning model and expert feedback.

[0020] As a further solution of the present invention, in step 3, the execution process of implicit modeling and machine learning optimization includes:

[0021] Step 41, preliminary hierarchical classification using an unsupervised algorithm: Use a K-means clustering algorithm to divide the data points into K clusters, where the center of each cluster is the average value of the data points. Input geological profile data, lithologic data, and geophysical data into the clustering algorithm, which automatically discovers the hierarchical structure of the geological body and outputs a cluster identifier for each geological body point for preliminary hierarchical classification.

[0022] Step 42: further optimize the model using a supervised classification algorithm: combine the labeled borehole data, select features that affect lithologic classification, use the labeled borehole data to train a random forest model so that the model learns the lithologic classification pattern, use cross-validation to evaluate model performance, adjust the model's hyperparameters based on accuracy and recall, optimize the classification results, and output the classification results of the geological body based on the probability values ​​output by the classification model;

[0023] Step 43, combining the unsupervised classification and supervised classification results: combining the preliminary results of the unsupervised classification with the classification results of the supervised learning, and using a weighted average method to obtain the classification results of each geological body.

[0024] In practical geological modeling, the quality and depth of coal seams are crucial for mineral resource assessment. These factors need to be considered to determine whether to hide low-quality and excessively deep coal seams, ensuring that the 3D model only includes those seams that meet research requirements. In ore cluster modeling, the primary research focus is on coal seams, while unrelated rock formations can interfere with modeling results. By explicitly hiding these unrelated rock formations, research objectives can be effectively focused, model complexity can be reduced, and data processing efficiency and subsequent analysis accuracy can be improved. Mined areas within a mining area should be hidden during modeling to avoid misjudgment or interference with unmined areas. Mining depth and a cut-off threshold are used to determine which areas have been maturely mined and can be removed from the 3D model. Inactive faults can be misleading during mining area modeling, especially when they do not affect the distribution and assessment of mineral resources. Therefore, inactive faults that do not affect ore predictions need to be hidden. Manual screening and manipulation of these data is inefficient and limited by the expertise and knowledge of the screeners, resulting in low uncertainty and accuracy.

[0025] As a further solution of the present invention, in step 4, deleting the identified portion to be hidden during the 3D image generation process is achieved by:

[0026] Step 11, calculate the normal vector: obtain the unit vector of the obtained vector by the cross product of the two vectors formed by the key point of the stratigraphic contour on each geological section and its two nearest adjacent points;

[0027] Step 12, generate shadows: Based on the above key points, determine whether there is a shadow by the angle between the light source and the normal vector of the key point. If the angle is 0, there is no shadow, and if the angle is not zero, there is a shadow. The maximum value between the unit vector from the light source to the key point and 0 is the light intensity of the key point.

[0028] Step 13, removing the hidden part: for the key point, if the distance from the point to the surrounding set area is less than a preset threshold, the set area is regarded as the geological area to be hidden, and the geological area to be hidden is removed;

[0029] Step 14, generate a rendering: combine the shadow intensity with the information that needs to be hidden in step 13 to generate the final visual geological profile rendering.

[0030] As a further solution of the present invention, in step 13, the demands that need to be hidden include but are not limited to the hidden demands of non-target coal seams, the hidden demands of rock strata related to non-coal seam research, the hidden demands of mined areas, and the hidden demands of inactive faults. The judgment of each hidden demand is made by numerical judgment calculated by the hidden demand judgment formula of non-target coal seams, the hidden demand judgment formula of rock strata related to non-coal seam research, the hidden demand judgment formula of mined areas, and the hidden demand judgment formula of inactive faults.

[0031] As a further solution of the present invention, in the hidden demand judgment formula of non-target coal seams, based on the quality and depth of the coal seams, the depth difference of the coal seams is measured using Gaussian distribution, and the difference between the coal seam depth and the maximum depth is processed by weighted activation function to analyze the mineability of the coal seams. The hidden demand judgment formula of non-target coal seams is:

[0032]

[0033] Where: P is the key point of the geological section, H c (P) is the hidden demand analysis value of the non-target coal seam at point P, Q P is the coal seam quality at point P, Q min is the minimum quality threshold of coal seam, D P is the coal seam depth at point P, D maxis the maximum mining depth of the coal seam, α1 is the coal seam quality correction coefficient, and σ1 is the standard deviation of the sensitivity of the control coal seam depth;

[0034] H c (P) exceeds the preset threshold, the coal seam is considered as an unimportant coal seam, and the important coal seam is deleted using the profile method.

[0035] As a further solution of the present invention, in the formula for determining the need to hide non-coal seam research related rock layers, whether to hide these non-coal seam research related rock layers is determined based on the type, thickness and location of the rock layer, whether the rock layer is irrelevant to the coal seam research is determined by a threshold, and the hiding judgment is further dynamically adjusted according to the thickness of the rock layer and the distance between the rock layer and the research area. The formula for determining the need to hide non-coal seam research related rock layers is:

[0036] In (T R ≠k c )∧(T P >T max ) when H nc (P) = 1;

[0037] In the case of not meeting (T R ≠k c )∧(T P >T max )hour,

[0038] Where: T R is the lithology analysis value, k c is the representative value of the coal seam, obtained by preset, T P is the thickness of the rock layer, T max is the maximum thickness of the rock layer, H nc (P) is the hidden demand analysis value of the relevant rock formations in non-coal seam research, D P,max is the maximum depth of the coal seam at point P, D pro is the distance between the rock layer and the study area, R min is the preset minimum distance between the rock layer and the study area, α2 is the influencing control factor of the coal seam depth at point P, σ2 is the influencing control factor of the distance between the rock layer and the study area, and sigmoid(·) is the sigma function;

[0039] H nc If (P) is greater than the preset threshold, the area is regarded as a non-coal seam study area and is removed through the regional segmentation algorithm.

[0040] As a further solution of the present invention, the hidden demand judgment formula in the mined area is based on the depth of the mined area, the mining progress and the relative position to the current research area. The hidden demand judgment formula of the mined area is:

[0041]

[0042] Where: H Mining (P) is the hidden demand judgment analysis value of the mined area at point P, L expl (P) is the mining progress of the coal seam where point P is located, L expl,min is the minimum display threshold of coal seam mining progress, α3 is the correction coefficient of coal seam mining progress, D expl (P) is the distance between the mined area of ​​the coal seam where point P is located and the study area, β2 is the display effect correction coefficient of the distance between the mined area and the study area, and σ3 is the display effect control parameter of the distance between the mined area and the study area;

[0043] H Mining If (P) is greater than a preset threshold, the area is considered as a mining area and masking technology is used to mask the mining area.

[0044] As a further solution of the present invention, the formula for determining the hiding requirement of inactive faults determines whether to hide these faults based on the activity, location, and impact of the faults. The formula for determining the hiding requirement of inactive faults is:

[0045]

[0046] Where: H fault (P) is the hidden demand judgment analysis value of the inactive fault, F fa (P) is the fault activity analysis value, which is obtained by analyzing the seismic activity and slip degree, α4 is the display influence coefficient of coal seam activity, f fa,min is the minimum value of fault activity analysis, D TCA (P) is the distance between the fault area where point P is located and the study area, D TCA,max is the maximum allowed display distance between the fault area and the study area, β3 is the adjustment factor for the distance effect of the fault area;

[0047] H fault If (P) is greater than the preset threshold, the area is considered as an inactive fault and the segmented rendering technology is used to process the inactive fault with a transparent effect.

[0048] As a further solution of the present invention, in step 5, model correction and optimization includes the following steps:

[0049] Step 51, correcting the lithologic layers and fault surfaces in the model: Based on the geological profile data, the 3D lithologic layers and fault surfaces are adjusted. The boundaries, thickness, and spatial distribution of the lithologic layers in the 3D model are adjusted. In particular, for the contact surfaces between different lithologies, interpolation is used to fill in the data-vacant areas to ensure the continuity of the lithologic layers throughout the entire region. Based on the changes in lithologic properties in the actual profile, the various lithologic layers in the 3D model are adjusted. In particular, the mutation areas are carefully adjusted to ensure that the lithologic layers conform to the geological structural constraints in 3D space. The strike, dip, and displacement of the fault are obtained using the geological profile data and the position information of the fault lines in the geological map. The fracture surfaces in the 3D model are adjusted based on the position and attributes of the actual fracture surfaces to ensure that they conform to the actual structural characteristics. For known active faults, the influence of their displacement is emphasized. The fracture surfaces are corrected using geometric constraints (such as plane equations) to ensure that the morphology of the fracture surfaces is consistent with the actual profile data.

[0050] Step 52, calibrating the rock property data based on the geophysical inversion results: using a seismic velocity model derived from the reflection wave data to estimate underground lithology and porosity, inverting the density distribution of underground materials using geophysical measurement data, and using underground electrical parameters derived from electromagnetic wave detection results, using the seismic velocity model and geophysical inversion results to perform regression analysis, comparing the inversion data with the original rock property data to obtain an optimized rock property model;

[0051] Step 53, verifying the corrected 3D geological model: comparing the corrected 3D geological model with the actual geological profile, strictly checking the accuracy of lithology distribution and fracture surface location, calculating the errors of the model before and after correction, including but not limited to lithology distribution error, fracture surface location error, and physical property data error, and evaluating them using standard error.

[0052] To solve the technical problems raised by the background technology, the present invention proposes a three-dimensional modeling method for geological profiles. The technical effects include: the present invention integrates three-dimensional implicit modeling, machine learning optimization and geological constraint correction, combines geological information databases and geophysical data, uses machine learning methods to extract geological structural constraints, constructs a three-dimensional structural model, and realizes rock type and fracture surface modeling through the surface model of two-dimensional geological information and fracture boundary constraints. It uses an unsupervised clustering algorithm to perform preliminary classification of geological bodies, and combines it with a supervised classification algorithm to further optimize the model using drilling data, thereby overcoming the challenge of unbalanced rock type categories and improving the accuracy of modeling. In terms of efficiency, during the modeling process, the algorithm extracts lithologic distribution, ore body characteristics and fault information from known drilling data and geological profiles to ensure that the structure of the model is consistent with the actual geological situation. Geophysical inversion and rock property statistics are combined with human-computer interactive interpretation to effectively correct the deviations in the lithologic model. By establishing three-dimensional lithologic layers and fracture surfaces, non-target coal seams, non-coal seams, mined areas and inactive faults are hidden based on preset formulas. The data that needs to be hidden is automatically judged and processed, reducing the complexity of manual intervention and data processing, clearly displaying the underground structure and resource distribution of the mining area, and providing a reliable basis for the quantitative prediction of mineral resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flow chart of a three-dimensional modeling method for geological sections proposed by the present invention;

[0054] Figure 2 This is a flowchart specifically implementing step 4 of the method proposed in the present invention;

[0055] Figure 3 This is a flowchart of a specific implementation of step 1 of the method proposed in the present invention;

[0056] Figure 4 This is a flowchart of the specific implementation of step 2 of the method proposed in the present invention;

[0057] Figure 5 This is a flowchart of the specific implementation of step three of the method proposed in the present invention;

[0058] Figure 6 The algorithm control panel interface of the application platform of the method proposed in the present invention;

[0059] Figure 7 This is a screenshot of the data processing interface of the application platform of the method proposed in the present invention;

[0060] Figure 8 This is a classification result diagram of the application platform of the method proposed in the present invention;

[0061] Figure 9This is a diagram of the geological section three-dimensional model control interface of the application platform of the method proposed in the present invention. DETAILED DESCRIPTION

[0062] The following is a clear and complete description of the technical solutions of the present invention, in conjunction with the accompanying drawings. Obviously, the technical solutions described are only part of the present invention, not all of it. All other technical solutions derived by persons of ordinary skill in the art based on the technical solutions of the present invention without inventive effort fall within the scope of protection of the present invention.

[0063] like Figures 1 to 5 As shown, the present invention proposes a three-dimensional modeling method for geological sections, comprising the following steps:

[0064] Step 1: Data acquisition and preprocessing: Combine the geological information data source library and geophysical data, use machine learning methods to extract geological structural constraints, and perform human-computer interactive interpretation by combining two-dimensional geological information and geophysical information gradient zones;

[0065] Step 2: Constraining the geological model: Using the geological structure data in the 2D geological database, a linear model is used to constrain the constructed 3D geological model. Lithologic modeling is performed. Combining stratigraphic data and physical properties, a surface model is used to establish a planar geological contact surface based on the lithologic data. The fracture surfaces are constrained using the fracture boundaries on the geological profile, and fractures that do not affect the structure are removed during the visualization phase.

[0066] Step 3: Implicit modeling and machine learning optimization: Use an unsupervised clustering algorithm to initially stratify and classify geological entities, automatically monitor the spatial distribution of lithology categories, and use a supervised classification algorithm combined with labeled drillhole data to further improve the classification model.

[0067] Step 4: 3D modeling and graphics rendering optimization: Voxel technology is used to render the underground lithology. The transition and contact relationships between different underground rock layers are represented through volume data. Different lithology layers are visualized through color mapping and transparency settings. An algorithm based on viewpoint selection and geometric visibility is used to delete identified areas that need to be hidden during the 3D image generation process.

[0068] Step 5: Model calibration and optimization: Adjust the model's 3D lithologic layers and fracture surfaces based on actual geological profile data to ensure they conform to geological conditions. Based on geophysical inversion results, calibrate the model's rock property data.

[0069] Step 6: Dynamic interaction and analysis: Using modern virtual reality technology, dynamic interaction is performed on the three-dimensional geological model. Through user interactive control, geological profiles at different depths can be viewed in real time and ore body distribution information can be analyzed.

[0070] It should be noted that the data acquisition and preprocessing process in step 1 includes:

[0071] Step 21, data acquisition: acquiring geological profile data, geological maps and geological structure maps, groundwater and soil information, and production exploration data from a geological information data source database. The geological profile data includes but is not limited to borehole data, stratum thickness, lithology information, and mineral composition. The geological maps and geological structure maps include fault location, fold, and lithology distribution data. The groundwater and soil information includes but is not limited to hydrogeology and groundwater flow. The production exploration data includes but is not limited to mining information and ore body analysis information. Gravity data, magnetic data, electromagnetic data, and seismic data are acquired from a geophysical database. The seismic data includes but is not limited to lithology and fault analysis information obtained through seismic wave velocity analysis.

[0072] Step 22, data storage and integration: establish a unified data repository, use geographic information system to store and manage spatial data, integrate data from different sources, ensure spatial consistency between data, convert them into a unified format, and perform data preprocessing (in step 22, it is necessary to unify the format of data from different sources into numerical format data, and standardize the data in the same format using k-means clustering algorithm, and then use the minimum-maximum normalization method to normalize the standardized data).

[0073] It should be noted that the specific implementation process of the geological model constraint in step 2 is as follows:

[0074] Step 31, feature extraction and structural constraints: Extract the bottom layer thickness, lithology information, the location, strike and dip of geological faults, mineral composition and spatial distribution based on geological maps and profiles; extract data related to gravity anomalies and underground density distribution, data related to magnetic anomalies and rock magnetism, data related to seismic wave velocity and lithology, geological structure, and electrical conductivity based on gravity, magnetic and seismic wave data; combine geological profiles and geophysical data, and constrain the spatial distribution of different lithology areas based on the physical characteristics of different rock layers, combined with seismic wave and gravity anomaly data; infer the geometry of the fault based on the strike and dip of the fault and combined with seismic reflection wave data; and construct the geometry of the stratum contact surface based on the mineral composition, lithology and fault information and seismic wave reflection data;

[0075] Step 32, human-computer interactive interpretation: Use Web GIS to provide geological experts with an intuitive data viewing and editing interface, allowing them to annotate, edit, and confirm unclear areas of the data. The machine learning model is optimized based on expert feedback, and the data source and model of the machine learning model are updated based on the newly added annotations during the human-computer interaction process.

[0076] Step 33, data integration and output: Merge geological data and geophysical data, integrate all constraints and form the final input data based on the machine learning model and expert feedback.

[0077] The data acquisition and preprocessing process in step 1 obtains a variety of data from geological information databases and geophysical databases, including geological profiles, geological maps, geological structure maps, groundwater and soil information, production exploration data, and gravity, magnetic, electromagnetic and seismic data, to provide accurate input for subsequent modeling. By establishing a unified data repository and using a geographic information system to manage and integrate data from different sources, the spatial consistency and accuracy of the data are ensured, providing high-quality data support for three-dimensional geological modeling.

[0078] It should be noted that in step 3, the execution process of implicit modeling and machine learning optimization includes:

[0079] Step 41, preliminary hierarchical classification using an unsupervised algorithm: Use a K-means clustering algorithm to divide the data points into K clusters, where the center of each cluster is the average value of the data points. Input geological profile data, lithologic data, and geophysical data into the clustering algorithm, which automatically discovers the hierarchical structure of the geological body and outputs a cluster identifier for each geological body point for preliminary hierarchical classification.

[0080] Step 42: further optimize the model using a supervised classification algorithm: combine the labeled borehole data, select features that affect lithologic classification, use the labeled borehole data to train a random forest model so that the model learns the lithologic classification pattern, use cross-validation to evaluate model performance, adjust the model's hyperparameters based on accuracy and recall, optimize the classification results, and output the classification results of the geological body based on the probability values ​​output by the classification model;

[0081] Step 43, combining the unsupervised classification and supervised classification results: combining the preliminary results of the unsupervised classification with the classification results of the supervised learning, and using a weighted average method to obtain the classification results of each geological body.

[0082] In step three, the geological bodies are stratified and classified by combining the unsupervised K-means clustering algorithm and the supervised random forest classification algorithm. The K-means clustering algorithm automatically identifies the preliminary hierarchical structure of the geological bodies and preliminarily classifies the geological data based on the clustering results. Subsequently, the random forest model is trained using labeled drill hole data to further optimize the classification. The hyperparameters are adjusted through cross-validation to improve the accuracy and robustness of the model. Finally, the weighted average method is used to fuse the results of the two to achieve more accurate classification of the geological bodies and provide more accurate data support for three-dimensional modeling.

[0083] In practical geological modeling, setting specific judgment formulas and thresholds to address the hidden requirements of different areas, such as coal seams, rock strata, mined areas, and inactive faults, can effectively improve model accuracy and efficiency. First, a comprehensive assessment of coal seam quality and depth is performed to hide low-quality or excessively deep coal seams, ensuring that the model only includes those seams that meet research requirements and avoid interference from irrelevant data. Second, during the modeling of mineral clusters, rock strata not relevant to coal seam research are explicitly hidden. By screening relevant data and focusing on the coal seam research objectives, unnecessary data processing is reduced, improving modeling efficiency and the accuracy of subsequent analysis. Third, mining depths and cutting thresholds are set to hide mined areas to avoid misjudgment and interference with unmined ore bodies. Finally, for inactive faults, manual screening and intervention are used to remove those that do not affect mineral resource distribution and assessment, further improving modeling accuracy and rationality. These operations effectively optimize data processing, model simplification, and accurate predictions during the modeling process, providing more reliable support for mineral resource assessment.

[0084] In practical geological modeling, the quality and depth of coal seams are crucial for mineral resource assessment. These factors need to be considered to determine whether to hide low-quality and excessively deep coal seams, ensuring that the 3D model only includes those seams that meet research requirements. In ore cluster modeling, the primary research focus is on coal seams, while unrelated rock formations can interfere with modeling results. By explicitly hiding these unrelated rock formations, research objectives can be effectively focused, model complexity can be reduced, and data processing efficiency and subsequent analysis accuracy can be improved. Mined areas within a mining area should be hidden during modeling to avoid misjudgment or interference with unmined areas. Mining depth and a cut-off threshold are used to determine which areas have been maturely mined and can be removed from the 3D model. Inactive faults can be misleading during mining area modeling, especially when they do not affect the distribution and assessment of mineral resources. Therefore, inactive faults that do not affect ore predictions need to be hidden. Manual screening and manipulation of these data is inefficient and limited by the expertise and knowledge of the screeners, resulting in low uncertainty and accuracy.

[0085] It should be noted that in step 4, the deletion of the identified parts to be hidden during the 3D image generation process is achieved in the following way:

[0086] Step 11, calculate the normal vector: the unit vector of the vector obtained by the cross product of the two vectors formed by the key point of the stratum contour on each geological section and its two nearest adjacent points is the normal vector of the key point;

[0087] Step 12, generate shadows: Based on the above key points, determine whether there is a shadow by the angle between the light source and the normal vector of the key point. If the angle is 0, there is no shadow, and if the angle is not zero, there is a shadow. The maximum value between the unit vector from the light source to the key point and 0 is the light intensity of the key point.

[0088] Step 13, removing the hidden part: for the key point, if the distance from the point to the surrounding set area is less than a preset threshold, the set area is regarded as the geological area to be hidden, and the geological area to be hidden is removed;

[0089] Step 14, generate a rendering: combine the shadow intensity with the information that needs to be hidden in step 13 to generate the final visual geological profile rendering.

[0090] It should be noted that in step 13, the demands that need to be hidden include but are not limited to the hidden demands of non-target coal seams, the hidden demands of rock strata related to non-coal seam research, the hidden demands of mined areas, and the hidden demands of inactive faults. The judgment of each hidden demand is made by the numerical judgment calculated by the hidden demand judgment formula of non-target coal seams, the hidden demand judgment formula of rock strata related to non-coal seam research, the hidden demand judgment formula of mined areas, and the hidden demand judgment formula of inactive faults.

[0091] It should be noted that in the hidden demand judgment formula for non-target coal seams, based on the quality and depth of the coal seams, the Gaussian distribution is used to measure the depth difference of the coal seams. The difference between the coal seam depth and the maximum depth is processed by the weighted activation function to analyze the mineability of the coal seams. The hidden demand judgment formula for non-target coal seams is:

[0092]

[0093] Where: P is the key point of the geological section, H c (P) is the hidden demand analysis value of the non-target coal seam at point P, Q P is the coal seam quality at point P, Q min is the minimum quality threshold of coal seam, D P is the coal seam depth at point P, D max is the maximum mining depth of the coal seam, α1 is the dimensional balance coefficient, and σ1 is the standard deviation of the sensitivity of the control coal seam depth;

[0094] H c (P) exceeds the preset threshold, the coal seam is considered as an unimportant coal seam, and the important coal seam is deleted using the profile method.

[0095] In this formula, The difference between coal seam quality and the minimum quality threshold is processed by Gaussian activation function, ensuring that coal seams below the minimum quality threshold are considered unmineable and thus hidden. Gaussian distribution is used to measure the difference between the depth of the coal seam and the maximum mineable depth. Coal seams deeper than the maximum mineable depth are considered unmineable and need to be hidden.

[0096] Example 1

[0097] Combined with specific examples, the specific execution process of the above formula is explained:

[0098] After extracting the multi-source data of drilling records and profiles, they are converted into a unified numerical format and stored in the database, ensuring that each point has two fields: coal seam quality and coal seam depth. There are five data points in the acquired data, and their original features are coal seam quality and coal seam depth. The coal seam quality of point P1 is 60 (coal seam quality score), and the coal seam depth is 500 meters. The coal seam quality of point P2 is 45, and the coal seam depth is 700 meters. The coal seam quality of point P3 is 55, and the coal seam depth is 550 meters. The coal seam quality of point P4 is 50, and the coal seam depth is 650 meters. The coal seam quality of point P5 is 65, and the coal seam depth is 480 meters.K-means clustering is used for preliminary classification. The data of these five points are divided into two categories according to the similarity of coal seam quality and coal seam depth, where k is 2. Cluster A contains data points P1, P3, and P5, and cluster B contains data points P2 and P4. Local standardization (Z-score) is performed within each cluster. For cluster A (P1, P3, and P5), the mean and standard deviation of coal seam quality are calculated to be 60 and 4.08 respectively, and the mean and standard deviation of coal seam depth are calculated to be 510 and 29.44 respectively. Therefore, the coal seam quality of data points P1, P3, and P5 is standardized by Z-score. The standardized values ​​of the coal seam quality are 0, -1.225, and 1.225 respectively. The Z-score standardized values ​​of the coal seam depth are -0.34, 1.36, and -1.02 respectively. Similarly, the mean and standard deviation of the data are calculated for the three points of cluster B. The mean and standard deviation of the coal seam quality are 47.5 and 2.5 respectively. The mean and standard deviation of the coal seam depth are 675 and 25 respectively. The Z-score standardized values ​​of the coal seam quality of data points P2 and P4 are -1 and 1 respectively. The Z-score standardized values ​​of the coal seam depth are 1 and -1 respectively. After standardization within each cluster, all The standardized values ​​of the data are global minimum and maximum normalized, and each feature is mapped to [0,1]. For coal seam quality, the data summary P1, P3, P5, P2, and P4 are 0, -1.225, 1.225, -1, and 1, respectively. The global minimum value is -1.225, and the global maximum value is 1.225. After processing using the global minimum-maximum normalization formula, P1, P3, P5, P2, and P4 are 0.5, 0, 1, 0.0918, and 0.9082, respectively. For coal seam depth, the data summary P1, P3, P5, P2, and P4 are -0.34 and 1.36, respectively. , -1.02, 1, -1, the global minimum is -1.02, the global maximum is 1.36, after the global minimum-maximum normalization formula, P1, P3, P5, P2, P4 are 0.2857, 1, 0, 0.8487, 0.0084 respectively, then the non-target coal seam hidden demand judgment formula is used, under the premise that the minimum quality threshold and the maximum mining depth are set to 0.5 and 0.5 respectively, when α1 and σ1 are taken as 0.2 and 0.2 respectively, the data processing for points P1, P3, P5, P2, P4 are as follows (decimals are rounded to four decimal places):.

[0099] (1) For P1:

[0100]

[0101] Among them H c (P) = 0.5 * 0.563 = 0.2815;

[0102] (2) For P3:

[0103]

[0104] Among them H c (P) = 0.525 * 0.044 = 0.0231;

[0105] (3) For P5:

[0106]

[0107] Among them H c (P) = 0.475 * 0.044 = 0.0209;

[0108] (4) For P2:

[0109]

[0110] Among them H c (P) = 0.52*0.221 = 0.1148;

[0111] (5) For P4:

[0112]

[0113] Among them H c (P)=0.4797*0.049=0.0235;

[0114] Finally, when the preset hidden threshold is 0.05 (the actual value is adjusted based on experience and verification data), according to H c Formula judgment of (P):

[0115] P1:H c (P) = 0.2815 > 0.05, hidden;

[0116] P2:H c (P) = 0.1148 > 0.05, hidden;

[0117] P3:H c (P) = 0.0235 ≤ 0.05, display;

[0118] P4:H c (P) = 0.0.0235 ≤ 0.05, display;

[0119] P5:H c (P) = 0.0209 ≤ 0.05, displayed;

[0120] The normalized post-processed data obtained above and the H of each key point c (P) The data and the final "show" and "hide" marks are stored in the GIS database and the 3D modeling system. During the rendering process, full rendering or hiding processing is selected based on the display status of the key points corresponding to each area.

[0121] It should be noted that in the formula for determining the need to hide non-coal seam research related rock layers, whether to hide these non-coal seam research related rock layers is determined based on the type, thickness, and location of the rock layer. A threshold is used to determine whether the rock layer is irrelevant to coal seam research, and the hiding judgment is further dynamically adjusted based on the rock layer thickness and the distance between the rock layer and the research area. The formula for determining the need to hide non-coal seam research related rock layers is:

[0122] In (T R ≠k c )∧(T P >T max ) when H nc (P) = 1;

[0123] In the case of not meeting (T R ≠k c )∧(T P >T max )hour,

[0124] Where: T R is the lithology analysis value, k c is the representative value of the coal seam, obtained by preset, T P is the thickness of the rock layer, T max is the maximum thickness of the rock layer, H nc (P) is the hidden demand analysis value of the relevant rock formations in non-coal seam research, D P,max is the maximum depth of the coal seam at point P, D pro is the distance between the rock layer and the study area, R min is the preset minimum distance between the rock layer and the study area, α2 is the dimensional balance coefficient, σ2 is the influencing control factor of the distance between the rock layer and the study area, and sigmoid(·) is the sigma function;

[0125] H nc If (P) is greater than the preset threshold, the area is regarded as a non-coal seam study area and is removed through the regional segmentation algorithm.

[0126] (T R ≠k c )∧(T P >T max), it means that the lithology type is not a coal seam, and the thickness of the rock layer exceeds the maximum threshold, then the rock layer is judged to be a non-coal seam research-related rock layer, and the hiding demand value is set to 1, indicating that the rock layer needs to be hidden; the hiding demand value is adjusted by the sigmoid function and the Gaussian function. The sigmoid function corrects the hiding demand based on the depth of the rock layer and the maximum depth value to ensure that the hiding demand will only be enhanced when the depth of the rock layer is greater than its maximum depth. At the same time, the Gaussian function takes into account the distance between the rock layer and the research area. The rock layer that is closer will be weighted according to the square difference of the distance to ensure that the rock layer that is farther away will not be mistakenly judged as a research-related rock layer.

[0127] Example 2

[0128] To further illustrate the processing process, the following example is given:

[0129] The data information after standardization and normalization of the acquired data points is shown in Table 1 below:

[0130] Table 1 Get data details table

[0131]

[0132]

[0133] The default parameters are set to: k c =0.50, T max =0.70,α2=0.20,R min =0.30;

[0134] For point P6:

[0135] Check conditions: (T R ≠k c )∧(T P >T max ), P6 point T R =0.30 and k c =0.50 are not equal, and T P =0.80 greater than T max =0.70, the conditions are met, and H is defined directly nc (P) = 1;

[0136] For point P7:

[0137] Check conditions: (T R ≠k c )∧(T P >T max ), P7 point T R =0.50 and k c =0.50 is equal, the condition is not met, enter the next branch, use sigmoid and Gaussian function to calculate,

[0138] For point P8:

[0139] Check conditions: (T R ≠k c )∧(T P >T max ), P8 point T R =0.4 and k c =0.50 is not equal, but T P =0.60 greater than T max =0.70, the condition is not fully met, enter the next branch, use sigmoid and Gaussian functions to calculate,

[0140] Perform hidden judgment and area segmentation:

[0141] When the preset hidden threshold is 0.5, nc (P) is greater than 0.5, and the area is determined to be a non-coal seam research area. The area is hidden through the regional segmentation algorithm. The judgment of each point is as follows:

[0142] For P6: H nc (P)=1>0.5, it is considered as a non-coal seam research related rock layer and needs to be hidden;

[0143] For P7: H nc (P) = 0.307 < 0.5, which is considered as the area related to coal seam research, showing;

[0144] For P8: H nc (P) = 0.0166 < 0.5, which is considered to be an area related to coal seam research and is displayed.

[0145] It should be noted that the hidden demand judgment formula in the mined area is based on the depth of the mined area, the mining progress, and the relative position to the current research area. The hidden demand judgment formula in the mined area is:

[0146]

[0147] Where: H Mining (P) is the hidden demand judgment analysis value of the mined area at point P, L expl (P) is the mining progress of the coal seam where point P is located, L expl,min is the minimum display threshold of coal seam mining progress, α3 is the dimensional balance coefficient, D expl(P) is the distance between the mined area of ​​the coal seam where point P is located and the study area, β2 is the display effect correction coefficient of the distance between the mined area and the study area, and σ3 is the display effect control parameter of the distance between the mined area and the study area;

[0148] H Mining If (P) is greater than a preset threshold, the area is considered as a mining area and masking technology is used to mask the mining area.

[0149] It is an S-shaped curve used to evaluate whether the area needs to be hidden according to the mining progress. As the mining progress increases, the value of the hiding demand gradually increases. When the mining progress is above the minimum threshold, the value tends to 1, indicating that the area is considered mined and hidden. is a Gaussian distribution function that determines whether hidden areas are necessary based on distance. Areas farther from the study area have less impact, while closer, mined areas require a stronger hiding requirement. This formula effectively incorporates mining progress and the relative position of the mined area and the study area to ensure that only data relevant to the study is retained. By controlling thresholds and correction coefficients, it flexibly hides mined areas, preventing them from unnecessarily impacting ore body analysis and modeling results, thereby improving the accuracy and effectiveness of mineral resource assessments.

[0150] Example 3

[0151] In order to further illustrate the specific role of the analysis formula proposed in the present invention, the example data of P9, P10, and P11 after format unification, local standardization, and global normalization are used for illustration. The example data is shown in Table 2:

[0152] Table 2 Instance data details of points P9, P10, and P11

[0153]

[0154] At the same time, the preset parameters are: α3=10, L expl,min =0.4, β2=5, σ3=0.2;

[0155] There are two parts in the distribution calculation formula. The first is Analysis and calculation of three parts:

[0156] (1)P9:

[0157] (2)P10:

[0158] (3)P11:

[0159] Then calculate the corresponding Part, specifically:

[0160] (1)P9:

[0161] (2)P10:

[0162] (3)P11:

[0163] Then get H Mining (P) are as follows:

[0164] (1)P9:H Mining (P)=0.931*4.25*10 -18 =3.96*10 -18 ; (2) P10: H Mining (P)=0.9986*4.54*10 -5 =4.54*10 -5 ;

[0165] (3)P11:H Mining (P)=0.9995*0.535=0.5347;

[0166] Assume the preset threshold is 0.5, H Mining If (P) is greater than 0.5, the area where the point is located is considered to be a mined area and needs to be masked using masking technology. Then:

[0167] P9:H Mining (P)=3.96*10 -18 , below 0.5, not hidden;

[0168] P10:H Mining (P)=4.54*10 -5 , below 0.5, not hidden;

[0169] P11:H Mining (P)=0.5347, greater than 0.5, hidden.

[0170] It should be noted that the formula for determining the hidden requirements of inactive faults is based on the activity, location, and impact of the faults to determine whether to hide these faults. The formula for determining the hidden requirements of inactive faults is:

[0171]

[0172] Where: H fault (P) is the hidden demand judgment analysis value of the inactive fault, F fa(P) is the fault activity analysis value, which is obtained by analyzing the seismic activity and slip degree, α4 is the dimensional balance coefficient, f fa,min is the minimum value of fault activity analysis, D TCA (P) is the distance between the fault area where point P is located and the study area, D TCA,max is the maximum allowed display distance between the fault area and the study area, β3 is the adjustment factor for the distance effect of the fault area;

[0173] H fault If (P) is greater than the preset threshold, the area is considered as an inactive fault and the segmented rendering technology is used to process the inactive fault with a transparent effect.

[0174] The formula for determining whether to hide inactive faults is used to determine whether these faults should be hidden during modeling based on their activity, location, and impact on mineral resource distribution. This formula comprehensively evaluates factors such as fault activity and impact range to determine whether to hide a fault, thereby improving the accuracy of mineral resource assessment and preventing irrelevant faults from interfering with model results. The formula accurately determines whether to hide inactive faults in geological modeling by comprehensively evaluating the fault's activity, distance from the study area, and its impact on mineral resources. This formula utilizes fault activity analysis values, location distance, and impact factors, using advanced mathematical methods such as Gaussian distribution and sigmoid functions to dynamically adjust the impact of fault display. Specifically, by combining fault activity with location, the formula automatically adjusts the visibility of inactive faults in the 3D model. This effectively prevents misleading mineral resource distribution and assessment caused by irrelevant faults, reduces noise interference in the model, and improves modeling accuracy. At the same time, the formula precisely controls the display range of faults, ensuring that only those that have a substantial impact on mineral resource assessment are retained. Ultimately, this refined fault concealment process improves the efficiency and accuracy of the mineral resource assessment process, providing a scientific basis for practical geological exploration and mineral resource development. In particular, under complex geological conditions, resource distribution predictions can be optimized by removing unnecessary inactive faults.

[0175] Example 4

[0176] To further illustrate the analysis process of the above formula, the analysis process of the formula is explained in detail by combining the data of P12, P13, and P14 after unified data format, standardization, and normalization. The point data of P12, P13, and P14 are shown in Table 3:

[0177] Table 3 Details of point data of P12, P13 and P14

[0178]

[0179] At the same time, the preset parameters are: α4=5, f fa,min =0.3,β3=4,D TCA,max =1;

[0180] First, calculate the formula part:

[0181] (1)P12:

[0182] (2)P13:

[0183] (3)P14:

[0184] Second, calculate part:

[0185] (1)P12:

[0186] (2)P13:

[0187] (3)P14:

[0188] Again, H of P12, P13, and P14 fault (P) are 0.6711*0.931=0.624, 0.0681*0.55=0.0375, 0.1712*0.802=0.1373 respectively;

[0189] Final judgment and subsequent processing: When the preset hidden threshold is 0.5, H fault If (P) is greater than 0.5, the area is considered to be an inactive fault and is processed with a transparent effect using segmented rendering technology:

[0190] For P12:H fault (P) = 0.624 > 0.5, this area is an inactive fault and is hidden using transparent processing;

[0191] For P13:H fault (P) = 0.0375 < 0.5, the area is not an inactive fault, shown;

[0192] For P14:H fault (P) = 0.1373 < 0.5, indicating that this region is not an inactive fault.

[0193] Actual effect:

[0194] By applying the technical methods specifically described in Examples 1 to 4 above, the generated three-dimensional model improves the assessment accuracy of coal seams by more than 20%, improves data processing efficiency by 30%, and provides more accurate guidance for subsequent mineral resource prediction and development planning.

[0195] It should be noted that in step five, model calibration and optimization includes the following steps:

[0196] Step 51, correcting the lithologic layers and fault surfaces in the model: Based on the geological profile data, the 3D lithologic layers and fault surfaces are adjusted. The boundaries, thickness, and spatial distribution of the lithologic layers in the 3D model are adjusted. In particular, for the contact surfaces between different lithologies, interpolation is used to fill in the data-vacant areas to ensure the continuity of the lithologic layers throughout the entire region. Based on the changes in lithologic properties in the actual profile, the various lithologic layers in the 3D model are adjusted. In particular, the mutation areas are carefully adjusted to ensure that the lithologic layers conform to the geological structural constraints in 3D space. The strike, dip, and displacement of the fault are obtained using the geological profile data and the position information of the fault lines in the geological map. The fracture surfaces in the 3D model are adjusted based on the position and attributes of the actual fracture surfaces to ensure that they conform to the actual structural characteristics. For known active faults, the influence of their displacement is emphasized. The fracture surfaces are corrected using geometric constraints (such as plane equations) to ensure that the morphology of the fracture surfaces is consistent with the actual profile data.

[0197] Step 52, calibrating the rock property data based on the geophysical inversion results: using a seismic velocity model derived from the reflection wave data to estimate underground lithology and porosity, inverting the density distribution of underground materials using geophysical measurement data, and using underground electrical parameters derived from electromagnetic wave detection results, using the seismic velocity model and geophysical inversion results to perform regression analysis, comparing the inversion data with the original rock property data to obtain an optimized rock property model;

[0198] Step 53, verifying the corrected 3D geological model: comparing the corrected 3D geological model with the actual geological profile, strictly checking the accuracy of lithology distribution and fracture surface location, calculating the errors of the model before and after correction, including but not limited to lithology distribution error, fracture surface location error, and physical property data error, and evaluating them using standard error.

[0199] Step 5 combines actual geological profile data with geological map information to precisely adjust the lithologic layers and fault surfaces in the 3D model, ensuring that the model's geological structure is highly consistent with the actual situation. Interpolation is used to fill data gaps, adjust lithologic layer boundaries and thicknesses, and correct the strike, dip, and displacement of fault surfaces. This optimizes the model's continuity and accuracy, particularly in the treatment of sudden changes and active faults, ensuring that the model truly reflects geological changes. This process improves the model's accuracy and reliability, effectively preventing the impact of model errors on mineral resource assessment, and providing a more scientific and reliable foundation for further mining exploration and mining planning.

[0200] In summary, the present invention integrates 3D implicit modeling, machine learning optimization, and geological constraint correction. By combining a geological information database with geophysical data, machine learning methods are used to extract geological structural constraints and construct a 3D structural model. Lithology and fracture surface modeling is then achieved using a surface model of 2D geological information and fracture boundary constraints. An unsupervised clustering algorithm is used to preliminarily classify geological bodies. Combined with a supervised classification algorithm, the model is further optimized using borehole data. This overcomes the challenge of lithology category imbalance and improves modeling accuracy and efficiency. During the modeling process, the algorithm extracts lithology distribution, ore body characteristics, and fault information from known borehole data and geological profiles to ensure that the model structure is consistent with the actual geological conditions. Geophysical inversion and petrophysical property statistics are combined with human-computer interactive interpretation to effectively correct deviations in the lithology model. By establishing 3D lithologic layers and fracture surfaces, non-target coal seams, non-coal seams, mined areas, and inactive faults are hidden based on preset formulas. The data to be hidden is automatically determined and processed, reducing manual intervention and the complexity of data processing. This clearly displays the underground structure and resource distribution of the mining area, providing a reliable basis for quantitative prediction of mineral resources.

[0201] like Figures 6 to 8 As shown, it is a partial screenshot of each interface of the platform where the method proposed in the present invention is applied. Figure 6 The screenshot of the algorithm control interface is shown in the figure. The left side shows the parameter setting box of the algorithm applied by the method. Figure 6 The right side shows the model training progress and classification results. Figure 7 By displaying the data source and data preprocessing in the data processing, the information of data source collection and the line chart of data preview are displayed, which is convenient for staff to view in time. Figure 8 The analysis of lithology, geological age, rock layer distribution and faults shows further expansion functions based on the application of this method, which facilitates a more detailed understanding of the specific conditions of each geological section and helps analyze the geological conditions of the study area.

[0202] exist Figure 9The interface for controlling the model on a platform using the three-dimensional modeling method for geological profiles proposed in the present invention is shown. Specifically, controls for model rotation, scaling, and perspective movement are displayed. The model can be rotated in three dimensions, the perspective can be moved, and the model can be scaled to an appropriate size. Front, side, and top views of the model can also be displayed, presenting richer geological profile information in two dimensions. In the display column for the stratum settings, stratum transparency can be adjusted and stratum labels can be displayed. Layer display control is performed through the specific execution steps of step 13 proposed in the method of the present invention. The thresholds for non-target coal seams, non-relevant rock seams, mined areas, and inactive faults are displayed based on the four formulas proposed in the present invention. This allows for the establishment of three-dimensional lithologic layers and fracture surfaces, and the automatic determination and processing of data that needs to be hidden based on preset formulas to hide non-target coal seams, non-coal seams, mined areas, and inactive faults. Detailed stratum information, including stratum depth, stratum dip, and stratum thickness, is also displayed at the bottom of the three-dimensional geological profile model. The specific three-dimensional geological profile model is modeled based on actual acquired data. Figure 9 It only shows the interface for layer control based on the values ​​calculated by the four formulas involved in the three-dimensional modeling method for geological sections proposed by the present invention in the model control interface.

[0203] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0204] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A three-dimensional modeling method for geological sections, characterized in that: The steps include: Step 1: Data acquisition and preprocessing: Combine the geological information data source library and geophysical data, use machine learning methods to extract geological structural constraints, and combine two-dimensional geological information and geophysical information gradient zones for human-computer interactive interpretation; Step 2: Constraining the geological model: Using the geological structure data in the 2D geological database, a linear model is used to constrain the constructed 3D geological model. Lithologic modeling is performed. Combining stratigraphic data and physical properties, a surface model is used to establish a planar geological contact surface based on the lithologic data. The fracture surfaces are constrained using the fracture boundaries on the geological profile, and fractures that do not affect the structure are removed during the visualization phase. Step 3: Implicit modeling and machine learning optimization: Use an unsupervised clustering algorithm to initially stratify and classify geological entities, automatically monitor the spatial distribution of lithology categories, and use a supervised classification algorithm combined with labeled drillhole data to further improve the classification model. Step 4: 3D modeling and graphics rendering optimization: Voxel technology is used to render the underground lithology. The transition and contact relationships between different underground rock layers are represented through volume data. Different lithology layers are visualized through color mapping and transparency settings. An algorithm based on viewpoint selection and geometric visibility is used to delete identified areas that need to be hidden during the 3D image generation process. Step 5: Model calibration and optimization: Adjust the model's 3D lithologic layers and fracture surfaces based on actual geological profile data to ensure they conform to geological conditions. Based on geophysical inversion results, calibrate the model's rock property data. Step 6: Dynamic interaction and analysis: Using virtual reality technology, dynamic interaction is performed on the 3D geological model. Through user interactive control, geological profiles at different depths can be viewed in real time and ore body distribution information can be analyzed.

2. A three-dimensional modeling method for geological sections according to claim 1, characterized in that: In step 4, the parts that need to be hidden are deleted during the 3D image generation process by: Step 11, calculate the normal vector: obtain the unit vector of the obtained vector by the cross product of the two vectors formed by the key point of the stratigraphic contour on each geological section and its two nearest adjacent points; Step 12, generate shadows: Based on the above key points, determine whether there is a shadow by the angle between the light source and the normal vector of the key point. If the angle is 0, there is no shadow, and if the angle is not zero, there is a shadow. The maximum value between the unit vector from the light source to the key point and 0 is the light intensity of the key point. Step 13, removing the hidden part: for the key point, if the distance from the point to the surrounding set area is less than a preset threshold, the set area is regarded as the geological area to be hidden, and the geological area to be hidden is removed; Step 14, generating a rendering: combining the shadow intensity with the information that needs to be hidden in step 13, generating a visual rendering of the geological profile.

3. A three-dimensional modeling method for geological sections according to claim 2, characterized in that: In step 13, the demands that need to be hidden include but are not limited to the hidden demands of non-target coal seams, the hidden demands of rock strata related to non-coal seam research, the hidden demands of mined areas, and the hidden demands of inactive faults. The judgment of each hidden demand is made by the numerical judgment calculated by the hidden demand judgment formula of non-target coal seams, the hidden demand judgment formula of rock strata related to non-coal seam research, the hidden demand judgment formula of mined areas, and the hidden demand judgment formula of inactive faults.

4. A three-dimensional modeling method for geological sections according to claim 3, characterized in that: In the hidden demand judgment formula for non-target coal seams, based on the quality and depth of the coal seams, Gaussian distribution is used to measure the depth difference of the coal seams. The difference between the coal seam depth and the maximum depth is processed by the weighted activation function to analyze the mineability of the coal seams. The hidden demand judgment formula for non-target coal seams is: Where: P is the key point of the geological section, H c (P) is the hidden demand analysis value of the non-target coal seam at point P, Q P is the coal seam quality at point P, Q min is the minimum quality threshold of coal seam, D P is the coal seam depth at point P, D max is the maximum mining depth of the coal seam, α1 is the coal seam quality correction coefficient, and σ1 is the standard deviation of the sensitivity of the control coal seam depth; H c (P) exceeds the preset threshold, the coal seam is considered as an unimportant coal seam, and the unimportant coal seam is deleted using the profile method.

5. A three-dimensional modeling method for geological sections according to claim 4, characterized in that: In the formula for determining the need to hide non-coal seam research related rock layers, whether to hide these non-coal seam research related rock layers is determined based on the type, thickness, and location of the rock layer. A threshold is used to determine whether the rock layer is irrelevant to coal seam research. The hiding judgment is further dynamically adjusted based on the thickness of the rock layer and the distance between the rock layer and the research area. The formula for determining the need to hide non-coal seam research related rock layers is: In (T R ≠k c )∧(T P >T max ) when H nc (P) = 1; In the case of not meeting (T R ≠k c )∧(T P >T max )hour, Where: T R is the lithology analysis value, k c is the representative value of the coal seam, obtained by preset, T P is the thickness of the rock layer, T max is the maximum thickness of the rock layer, H nc (P) is the hidden demand analysis value of the relevant rock formations in non-coal seam research, D P,max is the maximum depth of the coal seam at point P, D pro is the distance between the rock layer and the study area, R min is the preset minimum distance between the rock layer and the study area, α2 is the influencing control factor of the coal seam depth at point P, σ2 is the influencing control factor of the distance between the rock layer and the study area, and sigmoid(·) is the sigma function; H nc If (P) is greater than the preset threshold, the area is regarded as a non-coal seam study area and is removed through the regional segmentation algorithm.

6. A three-dimensional modeling method for geological sections according to claim 4, characterized in that: The hidden demand judgment formula in the mined area is based on the depth of the mined area, the mining progress and the relative position to the current research area. The hidden demand judgment formula in the mined area is: Where: H Mining (P) is the hidden demand judgment analysis value of the mined area at point P, L expl (P) is the mining progress of the coal seam where point P is located, L expl,min is the minimum display threshold of coal seam mining progress, α3 is the correction coefficient of coal seam mining progress, D expl (P) is the distance between the mined area of ​​the coal seam where point P is located and the study area, β2 is the display effect correction coefficient of the distance between the mined area and the study area, and σ3 is the display effect control parameter of the distance between the mined area and the study area; H Mining If (P) is greater than a preset threshold, the area is considered as a mining area and masking technology is used to mask the mining area.

7. A three-dimensional modeling method for geological sections according to claim 4, characterized in that: The formula for determining the hidden requirements of inactive faults is based on the activity, location, and impact of the faults to determine whether to hide these faults. The formula for determining the hidden requirements of inactive faults is: Where: H fault (P) is the hidden demand judgment analysis value of the inactive fault, F fa (P) is the fault activity analysis value, which is obtained by analyzing the seismic activity and slip degree, α4 is the display influence coefficient of the fault activity, and f fa,min is the minimum value of fault activity analysis, D TCA (P) is the distance between the fault area where point P is located and the study area, D TCA,max is the maximum allowed display distance between the fault area and the study area, β3 is the adjustment factor for the distance effect of the fault area; H fault If (P) is greater than the preset threshold, the area is considered as an inactive fault and the segmented rendering technology is used to process the inactive fault with a transparent effect.

8. A three-dimensional modeling method for geological sections according to claim 1, characterized in that: The data acquisition and preprocessing process in step 1 includes: Step 21, data acquisition: acquiring geological profile data, geological maps and geological structure maps, groundwater and soil information, and production exploration data from the geological information data. The geological profile data includes but is not limited to borehole data, stratum thickness, lithology information, and mineral composition. The geological maps and geological structure maps include fault location, fold, and lithology distribution data. The groundwater and soil information includes but is not limited to hydrogeology and groundwater flow. The production exploration data includes but is not limited to mining information and ore body analysis information. Gravity data, magnetic data, electromagnetic data, and seismic data are acquired from the geophysical database. The seismic data includes but is not limited to lithology and fault analysis information analyzed by seismic wave velocity. Step 22, data storage and integration: Establish a unified data repository, use geographic information systems to store and manage spatial data, integrate data from different sources, ensure spatial consistency between data, and convert them into a unified format.

9. The three-dimensional modeling method for geological sections according to claim 1, characterized in that: The specific implementation process of the geological model constraint in step 2 is as follows: Step 31, feature extraction and structural constraints: Extract the bottom layer thickness, lithology information, the location, strike and dip of geological faults, mineral composition and spatial distribution based on geological maps and profiles; extract data related to gravity anomalies and underground density distribution, data related to magnetic anomalies and rock magnetism, data related to seismic wave velocity and lithology, geological structure, and electrical conductivity based on gravity, magnetic and seismic wave data; combine geological profiles and geophysical data, and constrain the spatial distribution of different lithology areas based on the physical characteristics of different rock layers, combined with seismic wave and gravity anomaly data; infer the geometry of the fault based on the strike and dip of the fault and combined with seismic reflection wave data; and construct the geometry of the stratum contact surface based on the mineral composition, lithology and fault information and seismic wave reflection data; Step 32, human-computer interactive interpretation: Use Web GIS to provide geological experts with an intuitive data viewing and editing interface, allowing them to annotate, edit, and confirm unclear areas of the data. The machine learning model is optimized based on expert feedback, and the data source and model of the machine learning model are updated based on the newly added annotations during the human-computer interaction process. Step 33, data integration and output: Merge geological data and geophysical data, integrate all constraints and form the final input data based on the machine learning model and expert feedback.

10. The three-dimensional modeling method for geological sections according to claim 1, characterized in that: In step 3, the execution process for implicit modeling and machine learning optimization includes: Step 41, preliminary hierarchical classification using an unsupervised algorithm: Use a K-means clustering algorithm to divide the data points into K clusters, where the center of each cluster is the average value of the data points. Input geological profile data, lithologic data, and geophysical data into the clustering algorithm, which automatically discovers the hierarchical structure of the geological body and outputs a cluster identifier for each geological body point for preliminary hierarchical classification. Step 42: further optimize the model using a supervised classification algorithm: combine the labeled borehole data, select features that affect lithologic classification, use the labeled borehole data to train a random forest model so that the model learns the lithologic classification pattern, use cross-validation to evaluate model performance, adjust the model's hyperparameters based on accuracy and recall, optimize the classification results, and output the classification results of the geological body based on the probability values ​​output by the classification model; Step 43, combining the unsupervised classification and supervised classification results: combining the preliminary results of the unsupervised classification with the classification results of the supervised learning, and using a weighted average method to obtain the classification results of each geological body.

Images