BIM-GIS three-dimensional twinborn geological modeling method and system based on cognitive enhancement

Abstract

The embodiment of the invention provides a BIM-GIS three-dimensional twin geological modeling method and system based on cognitive enhancement. The method is applied to the technical field of geological modeling, and comprises the following steps: acquiring different data sources to carry out geological data acquisition, and carrying out preprocessing to obtain preprocessed geological data; converting the geologic body into a voxel three-dimensional model according to the preprocessed geologic data; performing multi-level optimization processing on the voxel three-dimensional model, constructing an optimized model, and performing three-dimensional display; and performing mineral resource space distribution prediction according to the voxel three-dimensional model, predicting the position and the number of mineral resources in combination with historical exploration data and a convolutional neural network, and performing underground water flow simulation analysis in combination with an underground water flow model. In this way, by combining the BIM and GIS technologies, the three-dimensional geological data can be efficiently and accurately managed and displayed, and the method can be widely applied to the fields of geological exploration, mining development, urban construction and the like.

Description

Technical Field

[0001] The present disclosure relates to the field of geological modeling technology, and in particular to a BIM-GIS three-dimensional twin geological modeling method and system based on cognitive enhancement. Background Art

[0002] In traditional geological data management systems, most work relies on 2D graphics, tables, and manual annotation to represent geological information. However, these methods lack intuitiveness, are highly complex, and rely heavily on manual labor. This not only reduces the real-time and accuracy of data, but also significantly increases the risk of human error.

[0003] To address this issue, 3D geological data management technology has emerged. 3D modeling technology can present geological data in the form of 3D models, helping geological prospectors gain a more intuitive and accurate understanding of underground geological structures and resource distribution. To address the inherent challenges of 3D geological modeling, geological data management systems based on BIM (Building Information Modeling) and GIS (Geographic Information System) technologies are gaining adoption. While these technologies offer advantages in many areas, existing technologies still have some shortcomings, primarily including complex modeling processes, insufficient modeling accuracy, and limited visualization.

[0004] Therefore, there is an urgent need for a design solution that can effectively solve the technical problems existing in the application of BIM and GIS technologies in geological modeling. Summary of the Invention

[0005] The present disclosure provides a BIM-GIS three-dimensional twin geological modeling method and system based on cognitive enhancement, which combines building information modeling (BIM) and geographic information system (GIS) technology to realize three-dimensional geological data display, and at least solves the technical problems of complex process and insufficient accuracy of three-dimensional modeling methods.

[0006] According to a first aspect of the present disclosure, a BIM-GIS three-dimensional twin geological modeling method based on cognitive enhancement is provided, comprising the following steps: Collect geological data based on different data sources and perform preprocessing to obtain preprocessed geological data; Converting the geological body into a voxel three-dimensional model based on the preprocessed geological data; Performing multi-level optimization processing on the voxel three-dimensional model to construct an optimized model, and performing three-dimensional display through a three-dimensional visualization engine; The spatial distribution of mineral resources is predicted based on the voxel three-dimensional model, and the location and quantity of mineral resources are predicted by combining historical exploration data and convolutional neural networks. The groundwater flow model is then used to perform groundwater flow simulation analysis.

[0007] According to the above aspects and any possible implementation, a further implementation is provided, wherein the process of acquiring geological data from different data sources and performing preprocessing to obtain preprocessed geological data is as follows: Collect geological data from multiple data sources including drilling data, seismic exploration data, and remote sensing impact data in the mining area. Build a unified geological data integration platform by collecting and integrating data from different exploration methods. Gaussian filter kernel function is used for automated data cleaning to identify and remove abnormal data and noise, and improved Kriging interpolation method is used to automatically fill in missing data to obtain processed data; The processed data in different formats are converted into a unified format to obtain processed geological data.

[0008] According to the above aspects and any possible implementation, a further implementation is provided, wherein the improved Kriging interpolation method uses a PSO method to optimize the Kriging difference parameters, specifically: The kriging difference parameters to be optimized are set as one-dimensional particles, and the position and velocity of each particle are initialized. The kriging difference parameters to be optimized are the nugget value C0, the partial sill value C, and the range a. In the optimization process, the sum of the difference between the actual variation function and the theoretical variation function of the interpolation point is selected as the fitness function in the particle swarm method; An improved particle swarm update method is used to compare the position selections between particles to obtain the optimal position, and then compare the particle with the global optimal position, retain the optimal position, and thus update the particle speed and position; Perform iterative updates, and when the termination condition is met, stop updating and output the optimized Kriging difference parameters, where the termination condition includes reaching the maximum number of iterations or obtaining the minimum fitness function value.

[0009] According to the above aspects and any possible implementation, an implementation is further provided, wherein the improved particle swarm update method realizes Kriging interpolation optimization through dynamic parameter adjustment rules, specifically including: ; ; ; ; ; in, is the inertia weight, which is adjusted according to the nonlinear attenuation rule. , is the acceleration factor, r1, r2 are random factors between [0,1], v j(k), x j (k) and v j (k+1), x j (k+1) represents the speed and position of the jth particle at time k and time k+1 respectively, p j (k) is the individual optimal position, g(k) is the global optimal position, k is the current number of iterations, is the maximum value of inertia weight, is the minimum inertia weight, K max is the maximum number of iterations.

[0010] According to the above aspects and any possible implementation, a further implementation is provided, wherein the process of converting the geological body into a voxel three-dimensional model according to the pre-processed geological data is as follows: Based on the pre-processed geological data, and by performing voxel calculation preparation on the geological body, determining the geological data and network accuracy parameters required for modeling; Selecting a geological body to be modeled in three-dimensional space, and determining a geological direction based on structural characteristics or dip angle information of the geological body; Performing triangular meshing on the geological body and constructing a bounding box, and simultaneously retrieving collective properties of the bounding box and the geological body; Extracting a diagonal vector from the bounding box and splitting it into x, y, and z components, determining a center point of the bounding box based on the x, y, and z components obtained by the splitting, and creating a central voxel at the center point; Expanding the central voxel in each coordinate direction, generating a plurality of adaptive points by cross-referencing surrounding components, and combining the central voxel with the adaptive points to complete voxelization of the bounding box and obtain a voxelized bounding box; Determine whether the geological volume and the voxelized bounding box are similar. If so, determine whether the voxel size is in the optimal range. If not, perform a collision test and re-evaluate the voxel size and make adjustments. If it is within the optimal range, voxelization is directly performed on the geological body to generate the final three-dimensional voxel model and complete the voxel modeling; if it is not within the optimal range, collision tests are continued or the diagonal vector split is returned and re-performed until the appropriate voxel size is obtained.

[0011] According to the above aspects and any possible implementation, an implementation is further provided, wherein the process of predicting the spatial distribution of mineral resources based on the voxel three-dimensional model and combining historical exploration data and a convolutional neural network to predict the location and quantity of mineral resources is as follows: Extracting mineralization alteration characteristics from the voxel three-dimensional model, wherein the mineralization alteration characteristics include pyrite content, magnetic susceptibility gradient, and spatial distribution continuity index; Constructing a 3D convolutional neural network, processing the mineralization and alteration characteristic input values ​​into the neural network, and outputting a mineral resource probability distribution map, wherein the input of the 3D convolutional neural network is a 64×64×64 voxel block, the neural network includes a convolutional layer connected to a global average pooling layer and a fully connected layer, and the loss function of the neural network introduces a geological continuity constraint term; Based on the probability distribution map of mineral resources, the resource volume is calculated using a linear regression model, specifically: ; in, is the target variable, representing the estimated mineral resource based on the weighted fusion of drill core grade and 3D voxel density. It is the geological background benchmark value, reflecting the background resource volume determined by regional mineralization background statistics. is the regression coefficient, which is used to quantify the contribution weight of each mineral-controlling factor to resource distribution. It is a multi-source geological indicator including independent variables including geophysical parameters, geochemical parameters and structural parameters. is the geological uncertainty error term, which obeys the normal distribution , In order to optimize the number of ore-controlling factors, the stepwise regression method was used to screen and determine them.

[0012] According to the above aspects and any possible implementation, a further implementation is provided, wherein the process of performing multi-level optimization processing on the generated voxel three-dimensional model to construct the optimized model is as follows: A data smoothing method is used to eliminate irregular shapes in the voxel 3D model, and the voxel 3D model is compared with the field survey data to verify the accuracy of the voxel 3D model and perform error correction to obtain an optimized model.

[0013] According to a second aspect of the present disclosure, an intelligent three-dimensional geological data modeling system based on BIM-GIS fusion is provided, which is used to implement the intelligent three-dimensional geological data modeling method based on BIM-GIS fusion as described in the first aspect. The system includes: a data acquisition and processing module, a three-dimensional modeling module, a visualization module, and a data analysis module; The data acquisition and processing module is used to acquire geological data based on different data sources and perform preprocessing to obtain preprocessed geological data; The three-dimensional modeling module is used to convert the geological body into a voxel three-dimensional model according to the pre-processed geological data; The visualization module is used to perform multi-level optimization processing on the generated voxel three-dimensional model, construct an optimized model, and display the optimized model in three dimensions through a three-dimensional visualization engine; The data analysis module is used to predict the spatial distribution of mineral resources based on the voxel three-dimensional model, and to predict the location and quantity of mineral resources in combination with historical exploration data and convolutional neural networks, and to perform groundwater flow simulation analysis in combination with a groundwater flow model.

[0014] According to the above aspects and any possible implementation, there is further provided an implementation, wherein the visualization module includes a stereoscopic view display module, a cross-sectional view display module, and a bottom perspective view display module; The stereoscopic view display module is used to display the overall structure of the three-dimensional geological body using the voxel three-dimensional model; The cross-sectional view display module is used to provide the user with a geological model cut view of a specified cross section and to display geological hierarchy and ore body distribution information; The bottom perspective view display module is used to display the stratum distribution of the geological body and distinguish different geological layers and mineral distributions through mapping of different colors and textures.

[0015] According to the above aspects and any possible implementation, there is further provided an implementation, wherein the data analysis module includes a spatial relationship analysis module, a mineral resource prediction module, and a groundwater flow simulation module; The spatial relationship analysis module is used to perform spatial analysis on three-dimensional geological data, construct spatial relationships between different geological bodies, and analyze the spatial distribution characteristics of underground resources; The mineral resource prediction module is used to predict the distribution of underground mineral resources based on historical exploration data and convolutional neural network algorithms, and provide potential locations and quantities of mineral resources; The groundwater flow simulation module is used to simulate the flow path of groundwater and the distribution of water resources in combination with hydrogeological data.

[0016] Compared with the prior art, the present invention has the following technical effects: (1) The present invention can perform accurate three-dimensional modeling: through the modeling method combining boundaries and voxels, the present invention can accurately reflect the shape and characteristics of complex underground geological bodies, overcoming the problem of insufficient accuracy of traditional geometric modeling methods. (2) The present invention achieves deep integration of BIM and GIS: Through the deep combination of BIM and GIS technologies, the present invention can not only realize spatial analysis and visualization of geological data, but also improve the efficiency and accuracy of data management, and provide accurate decision-making support for mineral resource development and groundwater flow analysis.

[0017] (3) The present invention is capable of intelligent data analysis and prediction: By integrating technologies such as data mining and machine learning, the present invention can realize automatic data analysis and prediction, reduce manual intervention, improve the intelligence level of the system, and further enhance the efficiency of geological exploration and resource management.

[0018] (4) The present invention effectively improves visualization effects and user experience: through the support of multiple visualization methods, the system can intuitively display three-dimensional geological models, provide clearer and more detailed views of underground structures, and help users make more scientific decisions.

[0019] It should be understood that the contents described in the Summary of the Invention section are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are provided for a better understanding of the present disclosure and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, among which: Figure 1 A schematic diagram of a BIM-GIS three-dimensional twin geological modeling method based on cognitive enhancement according to an embodiment of the present disclosure is shown; Figure 2 A flowchart of a modeling method of a BIM-GIS three-dimensional twin geological modeling method based on cognitive enhancement according to an embodiment of the present disclosure is shown; Figure 3 A schematic flow chart of a voxelization method for a BIM-GIS three-dimensional twin geological modeling method based on cognitive enhancement according to an embodiment of the present disclosure is shown; Figure 4 A schematic diagram of the structure of an intelligent three-dimensional geological data modeling system based on BIM-GIS fusion according to an embodiment of the present disclosure is shown; Figure 5 A schematic diagram of the workflow of a three-dimensional modeling module of an intelligent three-dimensional geological data modeling system based on BIM-GIS fusion according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0021] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present disclosure.

[0022] The management and visualization of geological data have always been a challenge in geological exploration and construction planning. Traditional geological data management methods rely on two-dimensional maps and tables, which are unable to effectively display complex underground structural information, and also suffer from low data processing efficiency and poor accuracy. With the development of three-dimensional modeling technology, the combination of BIM and GIS can better display underground structure and geological information, greatly improving data accuracy and visualization.

[0023] With the increasing demands of modern urban construction and mineral resource development, the field of geological exploration is placing higher demands on efficient and accurate geological data management systems. Geological data management is not just about data storage and organization; it also impacts the rational utilization of underground resources, environmental protection, and disaster prevention. Against this backdrop, traditional geological data management methods based on two-dimensional maps and annotations are increasingly insufficient to meet the demands of modern geological exploration.

[0024] In traditional geological data management systems, most work relies on two-dimensional graphics, tables, and manual annotation to represent geological information. This approach presents several problems: First, two-dimensional graphics cannot intuitively display the spatial relationships of geological bodies and changes in underground structures. Information such as the spatial relationships of geological bodies, changes in rock layer thickness, and groundwater flow cannot be effectively expressed using simple two-dimensional graphics. Second, due to the complexity of geological bodies, two-dimensional graphics struggle to accurately represent the heterogeneity and distribution patterns of underground resources, making analysis and prediction difficult. Finally, traditional methods rely on manual analysis and updates, which not only reduces the real-time and accuracy of data but also greatly increases the risk of human error.

[0025] To address these issues, 3D geological data management technology has emerged. 3D modeling technology can present geological data in the form of 3D models, helping geological prospectors gain a more intuitive and accurate understanding of underground geological structures and resource distribution. 3D models clearly demonstrate the spatial relationships of geological bodies, better reflect their hierarchical structures, and provide a deeper understanding of the distribution of underground resources through simulation and prediction methods. However, 3D geological modeling still faces many challenges, particularly in terms of modeling accuracy and computational efficiency. To address these challenges, geological data management systems based on BIM (Building Information Modeling) and GIS (Geographic Information System) technologies have been increasingly adopted in recent years, bringing new opportunities for geological data visualization, analysis, and management.

[0026] BIM technology was originally applied in the construction industry, aiming to provide efficient design and construction solutions through 3D modeling and data integration. By constructing digital 3D models of buildings or infrastructure, BIM integrates all design, construction, and operation data into a unified platform, providing an accurate basis for project management and decision-making. Having achieved tremendous success in the construction industry, BIM has gradually penetrated other fields in recent years, particularly in urban planning, infrastructure construction, and mineral resource management, where it has begun to play an increasingly important role. GIS technology focuses on the collection, storage, analysis, and display of geospatial data. GIS processes geographic data through a variety of techniques, including spatial analysis, network analysis, and spatial interpolation, helping users understand and analyze the physical features and geographical phenomena of the Earth's surface. In geological exploration, GIS technology provides powerful tools for the management, analysis, and visualization of large-scale geological data. GIS technology allows users to query and analyze data from a spatial perspective, supporting the prediction of mineral resource distribution, the construction of groundwater flow models, and earthquake hazard assessment.

[0027] When BIM is combined with GIS technology, comprehensive geological data management can be achieved on a unified platform. This integration not only integrates geospatial information used in building design, construction, and operations, but also processes large-scale 3D geographic data related to geological exploration and mineral resource development. Combining the advantages of BIM and GIS can significantly improve the efficiency and accuracy of geological data management, enabling geological prospectors to conduct more precise and detailed analyses.

[0028] While geological data management systems based on BIM and GIS technologies have demonstrated advantages in many areas, existing technologies still have some shortcomings. First, the data modeling process is complex. Existing 3D geological models mostly rely on complex geometric modeling processes. Traditional geometric modeling methods struggle to meet the high-precision modeling requirements, particularly for representing complex geological structures and heterogeneous resources. Especially when processing large-scale geological data, traditional modeling methods can encounter performance bottlenecks due to the large data volume and complex model calculations, thereby reducing model accuracy and computational efficiency. Second, modeling accuracy is insufficient. Some existing geological data management systems exhibit low modeling accuracy, particularly when representing complex geological volumes, making it difficult to accurately reflect the heterogeneity and complexity of the subsurface. For some heterogeneous resources (such as mineral deposits and groundwater flow zones), traditional modeling methods struggle to effectively capture their spatial distribution and physical properties, resulting in errors in underground resource distribution predictions and impacting resource development efficiency. Limited visualization is also a problem with current technologies. Existing geological data management systems mostly use simple geometric models to display geological data, making it difficult to present complex subsurface structures. Especially when it comes to 3D visualization, the visualization technology of many systems remains limited to 2D graphical displays. For the three-dimensional structure of underground geological bodies and mineral resources, the existing system lacks effective visualization technology, cannot present the fine structure of complex underground strata, and cannot effectively display the spatial relationship between different geological bodies, resulting in limitations in visualization effects.

[0029] This method leverages the advantages of BIM (Building Information Modeling) and GIS (Geographic Information System) technologies, ensuring not only the accuracy and efficiency of geological data modeling but also effectively enhancing visualization. Through efficient data management and analysis, it provides more precise and comprehensive support for geological exploration, mineral resource development, urban construction, and environmental protection. This method enables geological prospectors to model and analyze complex underground structures more intuitively and accurately, thereby optimizing resource utilization, improving decision-making efficiency, and reducing risks during exploration.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Reference Figure 1 and Figure 2 As shown, this embodiment provides a BIM-GIS three-dimensional twin geological modeling method based on cognitive enhancement, including the following steps: S101. Collect geological data based on different data sources and perform preprocessing to obtain preprocessed geological data.

[0032] In this embodiment, geological data is collected through multiple data sources such as drilling data, seismic exploration data, and remote sensing image data in the mining area. Data types include but are not limited to borehole data, point cloud data, raster data, etc. The collected raw data often contains noise and missing values, so data processing operations such as denoising and interpolation are required first. The missing values ​​are filled using the Kriging interpolation algorithm. The Kriging interpolation method dynamically adjusts the nugget value, partial base value, and range parameter through particle swarm optimization, and converts data from different sources into a standard format for subsequent processing. For example, point cloud data is converted into a voxel grid to ensure the spatial consistency of the data.

[0033] Gaussian filter kernel function for noise removal: (1) Where x and y are the offsets relative to the center point of the filter, and σ is the standard deviation of the Gaussian distribution.

[0034] Kriging interpolation method prediction value calculation: (2) Among them, λ is the weighting coefficient, is the estimated value at the point (x0, y0).

[0035] In this embodiment, the Kriging algorithm is the most widely used spatial interpolation method. It is an interpolation method that performs unbiased optimal estimation of regional variation values ​​within a limited area based on real structure analysis and variation function theory.

[0036] Preferably, the variogram in the Kriging algorithm is a core component, and the covariance or semivariance between each pair of points is calculated through the variogram to describe the degree of variation between any two points in the spatial data. However, no matter what kind of variation, it involves the values ​​of the three fitting parameters: nugget value C0, partial base value C, and range a. Therefore, in this embodiment, the traditional Kriging interpolation method is improved, specifically: the particle swarm optimization method is used to improve it, and the PSO method can quickly find the global optimal solution to determine the three parameters C0, C, and a in the variogram model. The specific process is as follows: 1) Initialize the particle swarm parameters. Set the parameters C0, C, and a in the Kriging interpolation method to be optimized as one-dimensional particles, and initialize the position and velocity of each particle; 2) Determine the fitness function. In the optimization process, the sum of the difference between the actual variation function and the theoretical variation function of the interpolation point is selected as the fitness function in the particle swarm algorithm. Specifically: (3) in, is the actual variation function, is the theoretical variation function.

[0037] 3) Particle swarm parameter update. In this embodiment, the improved particle swarm formula is as follows: The particle velocity update formula is: (4) The position update formula is: (5) Among them, the inertia weight The maximum value is set to 0.9, the minimum value is set to 0.4, and it is adjusted according to the nonlinear attenuation rule. , is the acceleration factor, r1, r2 are random factors between [0,1], v j (k), x j (k) and v j (k+1), x j (k+1) represents the speed and position of the jth particle at time k and time k+1 respectively, p j (k) is the individual optimal position, and g(k) is the global optimal position.

[0038] The fitness function takes the root mean square error between the measured data at the interpolation point and the theoretical variation function as the target, and constrains the range parameter satisfy ; When the error change rate for 10 consecutive iterations is less than Or the optimization is terminated when the maximum number of iterations is reached.

[0039] In formula (4) The value of will affect the motion state of the particles, while the values ​​of c1 and c2 determine the search ability of the particle swarm. Considering that the fixed weights in the traditional particle swarm method easily cause the algorithm to fall into the local optimal solution, the nonlinear attenuation rule is adjusted. Therefore, the improved strategy is as follows: (5) Where k is the current number of iterations, is the maximum value of inertia weight, is the minimum inertia weight, K max is the maximum number of iterations, the specific value is 200.

[0040] In the early stages of iteration, a larger inertia weight can prevent the algorithm from falling into local optimality and facilitate global search. In the later stages of iteration, a smaller inertia weight is conducive to local search. Specifically: (6) (7) The acceleration factor of this example It decreases linearly from the initial value of 2.5 to the final value of 0.5. It increases linearly from the initial value of 0.5 to the final value of 2.5.

[0041] 4) Establish the optimal mutation function model. When the termination condition is met, that is, the maximum number of iterations is reached or the minimum fitness function value is obtained, the calculation process is stopped and the optimized C0, C, and a are output, thereby obtaining the ideal mutation function parameter values.

[0042] In this embodiment, the fitness function takes the root mean square error between the measured data of the interpolation point and the theoretical variation function as the target, and constrains the range parameter satisfy ; When the error change rate for 10 consecutive iterations is less than Or the optimization is terminated when the maximum number of iterations is reached.

[0043] In this embodiment, the inertia weight of dynamic decay is introduced into the speed update equation ( nonlinear decay from 0.9 to 0.4) and the time-varying acceleration constant ( From 2.5 to 0.5, From 0.5 to 2.5), by adjusting the balance between the global search and local development capabilities of particles, the number of optimization iterations in measured data is reduced by 42% compared with the traditional fixed-parameter PSO algorithm.

[0044] Variable Range The value range (10-500 meters) is constrained by geological prior knowledge to avoid distortion of stratum continuity caused by excessive pursuit of mathematical optimal solutions. The dual composite termination condition ensures the balance between efficiency and accuracy of the algorithm. Actual engineering data shows that 85% of cases converge within 150 iterations, and the average time consumption is shortened from 4.2 minutes of traditional methods to 1.8 minutes. The optimized Kriging interpolation results are synchronized to the BIM model in real time through the API interface. When the GIS detects that the stratum dip angle changes by more than 2°, it automatically triggers the recalculation of local parameters. The data synchronization delay is controlled within 300 milliseconds, meeting the real-time requirements of dynamic modeling.

[0045] In this embodiment, data acquisition and intelligent preprocessing involve multi-source data integration. By integrating data from various exploration methods (such as remote sensing, drilling, and seismic wave propagation), a unified geological data integration platform is created. Next, machine learning algorithms are used for automated data cleaning to identify abnormal data and noise, removing or replacing them. Adaptive algorithms are used to optimize the interpolation process and automatically fill in missing data. Furthermore, data conversion and standardization operations convert data in different formats into a unified format, ensuring data consistency and, therefore, accuracy in subsequent model construction.

[0046] The method also includes: binding the compressive strength and elastic modulus of the rock formation in the BIM model to voxel units, constructing a three-dimensional topological network of the geological body using Delaunay triangulation based on GIS spatial topological analysis, and generating an influence zone with a buffer radius of 5 times the fault displacement for the fault area; when the GIS detects that the fault displacement exceeds 10 cm or the rock formation inclination offset exceeds 5°, triggering the recalculation of the BIM model parameters, and controlling the bidirectional synchronization error between BIM and GIS data within ±0.5%.

[0047] S102: Convert the geological body into a voxel three-dimensional model according to the pre-processed geological data.

[0048] Based on preprocessed multi-source geological data, the discretization of geological structures was first achieved through dynamic multi-resolution voxel meshing. Geological complexity was quantified based on local curvature and density gradients. High-precision voxels (0.1m×0.1m×0.1m) were used for high-complexity areas such as faults and ore body boundaries, while low-resolution voxels (1m×1m×1m) were used for homogeneous strata. Adjacent low-complexity voxels were dynamically merged using an octree structure, reducing the total number of meshes by 40%-60%. During the boundary extraction phase, an improved Marching Cubes algorithm was used to process multi-attribute voxel data such as density and resistivity. A density threshold of 2.8 g / cm³ was used as the stratum boundary criterion. Geological dip angles and inclinations were introduced to constrain the connection direction of isosurfaces. Canny edge detection (low threshold 0.1, high threshold 0.3) was combined to mitigate boundary jaggedness. For fault regions, radial basis function interpolation was used to complete the data. The kernel function was defined as a smooth transition function related to the fault displacement to ensure the continuity and geological rationality of the fault zone boundary.

[0049] To enhance modeling automation, a 3D U-Net deep learning model was introduced to identify lithologic distribution. It inputs 64×64×64 voxel blocks and outputs classification labels such as sandstone and shale. The network was trained by synthesizing 100,000 sets of geostatistical simulation data, combining Dice Loss and cross-entropy loss functions to mitigate class imbalance. The network predictions were post-processed using morphological closing operations and an empirical rule library. These constraints enforce a minimum sedimentary rock thickness of 0.3m and generate condensation edges when the intrusive contact zone dips greater than 45°. Isolated noise points were eliminated, and anomalous areas that violated geological laws were corrected. BIM attribute mapping was implemented simultaneously during the modeling process, incorporating rock mechanical parameters such as compressive strength and elastic modulus into voxel metadata. The associated density calculation formula was dynamically updated, triggering parameter recalculation when the voxel density changes by more than ±5%, supporting real-time engineering mechanics analysis.

[0050] During the model integration phase, a spatial topological network of geological bodies and engineering structures is constructed through Delaunay triangulation, and the minimum distance between the rock mass and the building structure is calculated. If the distance is less than 2m and the rock mass compressive strength is less than 30MPa, it is marked as a high-risk construction area. The resulting integrated model supports multi-scale visualization, rendering high-precision voxel details from a near-field perspective, switching to a low-polygon mesh in the far field, combining transparency mapping to display the mineral layer and surrounding rock in layers, and dynamically optimizing the rendering load through LOD technology. Experiments show that compared with traditional modeling, this method reduces the voxel error from 12.3% to 6.7%, reduces the boundary jaggedness by 75%, and improves the fault interpolation efficiency by 74%, providing a high-precision three-dimensional foundation for collaborative design and risk warning of complex geological engineering. This embodiment uses BIM technology to integrate the geological body model with the BIM model of the building or engineering project to form a complete three-dimensional spatial model for further analysis and decision-making. In the integrated BIM model, spatial relationship analysis is performed to identify potential building risks, geological problems, etc., and the accuracy of the model is improved through optimization algorithms. In addition, based on the integrated model, the system supports multi-scenario analysis, such as construction risk assessment, underground resource exploration, etc., providing multi-dimensional analysis and application.

[0051] like Figure 3 Specifically, in this embodiment, the specific process of voxel 3D modeling is as follows: First, voxel calculation preparation is performed on the geological body to confirm the required geological data, grid accuracy and other parameters.

[0052] The geological body to be modeled is selected in three-dimensional space, and based on its structural characteristics or dip angle and other information, the appropriate geological direction is determined to facilitate subsequent grid division.

[0053] Based on the initially acquired geological body, it is divided into triangular meshes to better identify the geological body boundaries and extract geometric characteristics in subsequent steps.

[0054] In order to limit the spatial extent of the geological body, a bounding box is constructed, and the collective properties of the bounding box and the geological body are retrieved, including vertex coordinates, normal vectors, etc.

[0055] Extract the diagonal vector from the bounding box and split it into several parts as input for the subsequent voxelization process. The diagonal vector is decomposed into three components, x, y, and z, to facilitate fine control of the center point and voxel size.

[0056] Based on the x, y, and z components obtained from the split, the center point of the bounding box is determined. The first "center voxel" is created at the center point to initialize the voxelization process.

[0057] The central voxel is expanded in each coordinate direction, and several adaptive points are generated by cross-referencing surrounding components (such as mesh nodes or triangle faces). These adaptive points can dynamically adjust the voxel size or shape according to the complexity of the geological volume.

[0058] The initial center voxel is combined with the surrounding adaptive points to achieve voxel coverage of the entire bounding box. This process can be used to perform hierarchical or multi-resolution partitioning of voxels based on the structural characteristics of the geological volume.

[0059] After completing the initial voxelization of the bounding box, it is necessary to determine whether the voxelized result is similar to the actual shape of the geological volume. If so, the next step is to determine whether the voxel size is within the optimal range. If not, a collision test is required (such as detecting overlap or missing between the voxels and the geological volume) to further evaluate whether the voxel size needs to be adjusted.

[0060] Determine whether the voxel size is within the optimal range. If so, voxelize the geological volume directly to generate the final 3D voxel model, completing the voxel modeling. If not, continue the collision test or return to the re-division input step (i.e., readjust the diagonal vectors, central voxels, or adaptive points) until the appropriate voxel size is obtained.

[0061] Through the above multi-step iteration and collision tests, the computational efficiency can be improved while ensuring the modeling accuracy, and ultimately a complete and accurate voxel 3D model can be obtained.

[0062] After obtaining a high-precision three-dimensional voxel model, the linear regression model can be further used to predict the mineral resources. Its mathematical form can be expressed as: (3) in, is the target variable, representing the estimated mineral resource value (in 10,000 tons) based on the weighted fusion of drill core grade and 3D voxel density. It is the geological background benchmark value, reflecting the background resource volume determined by regional mineralization background statistics. is the regression coefficient, which is used to quantify the contribution weight of each ore-controlling factor to resource distribution, such as the influencing factor of fault zone intersection density and the coefficient of surrounding rock alteration intensity; It is a multi-source geological indicator whose independent variables include geophysical parameters (Bouguer gravity anomaly, magnetic susceptibility), geochemical parameters (Cu / Au element contrast, alteration mineral abundance) and structural parameters (fault density, fold curvature); is the geological uncertainty error term, which obeys the normal distribution , covering unmodeled factors such as microstructural variation and ore-forming fluid activity; In order to optimize the number of mineral-controlling factors, the stepwise regression method was used to screen and determine the number, with a typical value range of 5-8 items to balance the model complexity and overfitting risk.

[0063] S103 , performing multi-level optimization processing on the generated voxel three-dimensional model to construct an optimized model, and performing three-dimensional display of the optimized model through a three-dimensional visualization engine.

[0064] In this embodiment, the generated three-dimensional model is subjected to multi-level optimization processing, especially the boundaries and geological layers are fine-tuned to improve the accuracy of the model. The generated voxel three-dimensional model is subjected to multi-level optimization processing. First, the Laplacian smoothing and bilateral filtering combined algorithm is used to eliminate model noise, wherein the Laplacian smoothing weight is set to 0.3 to maintain the topological relationship of the geological structure. The bilateral filtering is double-constrained by the spatial standard deviation of 0.5 and the color domain standard deviation of 0.1, while removing discrete voxel jump artifacts, retaining the sharpness of the ore body boundary. Model accuracy verification is divided into two levels: coarse scale and fine scale: at the coarse scale level, the model is aligned with the seismic exploration profile data, and local voxel resampling is triggered when the calculated root mean square error exceeds 8%; at the fine scale level, the drill core data is used to verify each layer, and radial basis function interpolation correction is performed on the rock layer with a thickness error exceeding 15 cm. The optimized model is managed through an octree-LOD hybrid structure. 0.1-meter precision voxels are loaded within 50 meters in the near field, a 0.5-meter simplified grid is switched between 50-200 meters, and 1-meter voxel block aggregation is used for expression beyond 200 meters. This increases the rendering efficiency of billion-level voxel scenes by more than 3 times.

[0065] Secondly, this embodiment uses a 3D visualization engine to display the geological model in 3D. A dedicated geological visualization engine, developed based on the Unity High Definition Rendering Pipeline (HDRP), implements dynamic color mapping and stratigraphic perspective. Lithology codes are mapped to the voxel surface in real time using an HSL gradient color palette, with granite marked orange and sandstone tan. The hue and saturation gradient varies with the mineral composition. Stratigraphic perspective utilizes depth peeling technology for layered rendering, supporting the simultaneous display of up to eight geological layers. Transparency is controlled by an exponential decay formula with depth, starting at 0.8 at a depth of 20 meters and decreasing to 0.1 at a depth of 100 meters. The virtual reality module integrates the SteamVR SDK and dynamically optimizes resolution using foveated rendering technology. Multi-user collaborative analysis utilizes CRDT-based conflict-free replicated data types, supporting concurrent multi-user operations. Command transmission latency is less than 50 milliseconds, and the automatic merging of version differences has a success rate exceeding 99%, meeting the real-time collaboration needs of geological exploration teams.

[0066] Mineral resource prediction is achieved through a cascaded deep learning architecture. Three-dimensional density gradient fields, alteration mineral abundances, and geochemical anomaly indices are extracted from the voxel model as feature inputs. The data augmentation stage uses sequential Gaussian simulation to generate 10,000 virtual ore deposit voxel blocks, covering mineralization types such as layered and vein-like. Elastic deformations of ±5 voxels and Gaussian noise with a standard deviation of 0.1 are applied to the raw data to enhance model generalization. The prediction network consists of a five-layer 3D convolutional encoder and a Transformer decoder. The convolution kernel size is 3×3×3, and the number of channels is expanded from 64 to 512. Each layer is followed by group normalization and Gaussian error linear activation. The decoder fuses 512-dimensional feature vectors using a four-head self-attention mechanism. Positional encodings are incorporated into the voxel space coordinates to enhance spatial correlation. Training uses a cross-entropy loss with a gradient consistency constraint weighted by 0.2 to force the predicted ore body boundaries to align with the density gradient field. During the deployment phase, the model was accelerated and optimized by TensorRT. The inference time for a 256×256×256 voxel block was 1.2 seconds, and the mineral probability heat map and resource estimation results were output. The resource volume of high-confidence target areas (probability > 0.7) was calculated according to the density-grade regression model after morphological closing operation, and the error rate was controlled within 8%.

[0067] This solution addresses distortion issues in complex structural modeling through a multi-scale optimization algorithm and physically constrained interpolation. Compared to traditional methods, it reduces boundary aliasing from 18.2% to 4.5%, and improves fault interpolation efficiency by 74%. The visualization engine integrates dynamic LOD management and GPU parallel computing technology to achieve smooth 4K / VR multimodal interaction in scenes with billions of voxels, achieving a 3.2x increase in frame rate compared to traditional OpenGL solutions.

[0068] S104. Predict the spatial distribution of mineral resources based on the voxel three-dimensional model, and combine historical exploration data and convolutional neural networks to predict the location and quantity of mineral resources, and conduct groundwater flow simulation analysis in combination with a groundwater flow model.

[0069] In this embodiment, mineralization and alteration characteristics, including pyrite content, magnetic susceptibility gradient, and spatial distribution continuity index, are extracted from the voxel model; a 3D convolutional neural network is constructed with a 64×64×64 voxel block as input. The network includes five three-dimensional convolutional layers connected to a global average pooling layer and a fully connected layer. A geological continuity constraint is introduced into the loss function, and the constraint weight is set to 0.1. Pre-trained weights are loaded through transfer learning, and Focal Loss is used to alleviate the sample imbalance problem; a probability distribution map of mineral resources is output, and the resource volume is calculated based on a linear regression model. The regression coefficient is solved by ridge regression regularization, and the characteristic variables include voxel density and electromagnetic anomaly values.

[0070] In the data analysis and deep learning phase, this embodiment combines deep learning algorithms for underground resource prediction, using algorithms such as convolutional neural networks (CNNs) to analyze the spatial distribution and compositional characteristics of mineral resources and automatically generate analysis reports that include spatial distribution maps of mineral resources, geological layer maps, and prediction errors. Furthermore, leveraging cloud computing platforms and big data analysis technologies, the system can deeply mine and analyze massive amounts of geological data, automatically discovering potential geological patterns.

[0071] In summary, the method of the present invention first conducts preliminary collection and analysis of the required geological data (such as geological models, soil layers, bedrock, etc.), and establishes a geological data model based on the existing geological information, laying the foundation for subsequent algorithm development and model generation.

[0072] After completing the development of the geological data model, geotechnical engineering simulation software and related algorithm descriptions are used to perform boundary representation (B-rep) of the geological body, and construct corresponding three-dimensional voxel models based on the different characteristics of homogeneous or heterogeneous strata to form a more detailed digital expression of the geological body.

[0073] The generated geological model is combined with IFC (Industry Foundation Classes) and CityGML (City Geography Markup Language) to achieve bidirectional data mapping and extension. This process allows the geological model to be integrated into the BIM-GIS platform, thereby constructing a city-level hybrid geological model, supporting a wider range of spatial analysis and visualization applications.

[0074] Demonstrate the 3D geological modeling and analysis framework of the present invention through actual service or operational environments. At this stage, demonstrate application examples of the system, such as the integration and scheduling of geological models in engineering scenarios and the presentation of hybrid models at the case level, to visually demonstrate the feasibility and applicability of the solution.

[0075] Finally, the operating performance of the entire system is compared, analyzed and verified with the data results, the basic data and model prediction results are statistically analyzed, and a comprehensive evaluation of the model's accuracy, efficiency and scalability is made based on actual needs, providing a reference basis for subsequent improvements and further promotion.

[0076] like Figure 4 As shown, this embodiment also provides an intelligent three-dimensional geological data modeling system based on BIM-GIS fusion, including: a data acquisition and processing module 1, a three-dimensional modeling module 2, a visualization module 3 and a data analysis module 4; The data acquisition and processing module 1 is used to collect geological data based on different data sources and perform preprocessing to obtain preprocessed geological data; The three-dimensional modeling module 2 is used to convert the geological body into a voxel three-dimensional model according to the pre-processed geological data; The visualization module 3 is used to perform multi-level optimization processing on the generated voxel 3D model, construct an optimized model, and display the optimized model in 3D through a 3D visualization engine; wherein the visualization module 3 includes a stereoscopic view display module 31, a cross-sectional view display module 32, and a bottom perspective view display module 33; The stereoscopic view display module 31 is used to display the overall structure of the three-dimensional geological body using the voxel three-dimensional model; the user can view the entire geological model from different angles to understand the spatial distribution of the geological body; The cross-sectional view display module 32 is used to provide the user with a cut view of the geological model of a specified cross section and to display the geological hierarchy and ore body distribution information; The bottom perspective view display module 33 is used to display the stratum distribution of the geological body and distinguish different geological layers and mineral distributions through mapping of different colors and textures.

[0077] The data analysis module 4 is used to predict the spatial distribution of mineral resources based on the voxel three-dimensional model, and combine historical exploration data and convolutional neural networks to predict the location and quantity of mineral resources, and combine the groundwater flow model to perform groundwater flow simulation analysis; among them, the data analysis module 4 includes a spatial relationship analysis module 41, a mineral resource prediction module 42 and a groundwater flow simulation module 43.

[0078] The spatial relationship analysis module 41 is used to perform spatial analysis on 3D geological data, constructing spatial relationships between different geological bodies and analyzing the spatial distribution characteristics of underground resources. Using spatial interpolation algorithms and regional analysis techniques, it can accurately identify the spatial layout of key geological bodies such as mineral layers and rock formations.

[0079] The mineral resource prediction module 42 is used to predict the distribution of underground mineral resources based on historical exploration data and convolutional neural network algorithms, and provide potential locations and quantities of mineral resources; The groundwater flow simulation module 43 is used to simulate the flow path of groundwater and the distribution of water resources in combination with hydrogeological data.

[0080] like Figure 5 As shown, specifically, in this embodiment, the specific working process of the three-dimensional modeling module 2 is: First, establish and develop a geotechnical engineering exploration database as the basis for subsequent geological information management and retrieval.

[0081] After the database is built, the required geological information, including rock structure, exploration parameters, etc., is retrieved from it to provide data support for three-dimensional geological modeling.

[0082] Based on the retrieved geological information, we begin to build a 3D geological model. If we encounter data gaps or anomalies, we will promptly supplement or correct the geological data to ensure model accuracy.

[0083] The constructed 3D geological model is integrated into the BIM (Building Information Modeling) environment to generate a BIM model for further analysis and management.

[0084] According to common industry standards, the geological model in the BIM model is exported as an IFC (Industry Foundation Classes) format file to facilitate subsequent interoperability in GIS and other systems.

[0085] Use data mapping tools or interfaces to map geological model data in IFC format to the CityGML (City Geography Markup Language) standard to generate a GIS model.

[0086] The geological model processed by CityGML standardization becomes a GIS model and can be directly imported into the GIS system for geospatial analysis, visualization and subsequent voxel modeling.

[0087] Based on the GIS model, the B-rep (boundary representation) method is used to identify and process the geometric boundaries, and finally the voxel construction is completed to obtain a three-dimensional voxel model.

[0088] The complete three-dimensional voxel model can be further applied to various scenarios such as geological exploration, mineral resource assessment, and urban planning, providing data support for the overall analysis and decision-making of the system.

[0089] Through the organic connection of the above steps, the 3D modeling module 2 realizes the complete process from database construction to voxel model generation, and is closely integrated with BIM and GIS technologies, providing a reliable data foundation and visualization support for subsequent multi-field applications.

[0090] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present disclosure.

[0091] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0092] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A BIM-GIS three-dimensional twin geological modeling method based on cognitive enhancement, characterized in that: The following steps are involved: Collect geological data based on different data sources and perform preprocessing to obtain preprocessed geological data; Converting the geological body into a voxel three-dimensional model based on the preprocessed geological data; Performing multi-level optimization processing on the voxel three-dimensional model to construct an optimized model, and performing three-dimensional display through a three-dimensional visualization engine; The spatial distribution of mineral resources is predicted based on the voxel three-dimensional model, and the location and quantity of mineral resources are predicted by combining historical exploration data and convolutional neural networks. The groundwater flow model is then used to perform groundwater flow simulation analysis.

2. The cognitive enhancement-based BIM-GIS three-dimensional twin geological modeling method according to claim 1 is characterized in that: The process of acquiring geological data from different data sources and performing preprocessing to obtain preprocessed geological data is as follows: Collect geological data from multiple data sources including drilling data, seismic exploration data, and remote sensing impact data in the mining area. Build a unified geological data integration platform by collecting and integrating data from different exploration methods. Gaussian filter kernel function is used for automated data cleaning to identify and remove abnormal data and noise, and improved Kriging interpolation method is used to automatically fill in missing data to obtain processed data; The processed data in different formats are converted into a unified format to obtain processed geological data.

3. The cognitive enhancement-based BIM-GIS three-dimensional twin geological modeling method according to claim 2 is characterized in that: The improved Kriging interpolation method uses the PSO method to optimize the Kriging difference parameters, specifically: The kriging difference parameters to be optimized are set as one-dimensional particles, and the position and velocity of each particle are initialized. The kriging difference parameters to be optimized are the nugget value C0, the partial sill value C, and the range a. In the optimization process, the sum of the difference between the actual variation function and the theoretical variation function of the interpolation point is selected as the fitness function in the particle swarm method; An improved particle swarm update method is used to compare the position selections between particles to obtain the optimal position, and then compare the particle with the global optimal position, retain the optimal position, and thus update the particle speed and position; Perform iterative updates, and when the termination condition is met, stop updating and output the optimized Kriging difference parameters, where the termination condition includes reaching the maximum number of iterations or obtaining the minimum fitness function value.

4. The cognitive enhancement-based BIM-GIS three-dimensional twin geological modeling method according to claim 3 is characterized in that: The improved particle swarm update method realizes Kriging interpolation optimization through dynamic parameter adjustment rules, specifically including: ； ； ； ； ； in, is the inertia weight, which is adjusted according to the nonlinear attenuation rule. , is the acceleration factor, r1, r2 are random factors between [0,1], v j (k), x j (k) and v j (k+1), x j (k+1) represents the speed and position of the jth particle at time k and time k+1 respectively, p j (k) is the individual optimal position, g(k) is the global optimal position, k is the current number of iterations, is the maximum value of inertia weight, is the minimum inertia weight, K max is the maximum number of iterations.

5. The cognitive enhancement-based BIM-GIS three-dimensional twin geological modeling method according to claim 1 is characterized in that: The process of converting the geological body into a voxel three-dimensional model according to the pre-processed geological data is as follows: Based on the pre-processed geological data, and by performing voxel calculation preparation on the geological body, determining the geological data and network accuracy parameters required for modeling; Selecting a geological body to be modeled in three-dimensional space, and determining a geological direction based on structural characteristics or dip angle information of the geological body; Performing triangular meshing on the geological body and constructing a bounding box, and simultaneously retrieving collective properties of the bounding box and the geological body; Extracting a diagonal vector from the bounding box and splitting it into x, y, and z components, determining a center point of the bounding box based on the x, y, and z components obtained by the splitting, and creating a central voxel at the center point; Expanding the central voxel in each coordinate direction, generating a plurality of adaptive points by cross-referencing surrounding components, and combining the central voxel with the adaptive points to complete voxelization of the bounding box and obtain a voxelized bounding box; Determine whether the geological volume and the voxelized bounding box are similar. If so, determine whether the voxel size is in the optimal range. If not, perform a collision test and re-evaluate the voxel size and make adjustments. If it is within the optimal range, voxelization is directly performed on the geological body to generate the final three-dimensional voxel model and complete the voxel modeling; If it is not within the optimal range, continue the collision test or go back and redo the diagonal vector splitting until the appropriate voxel size is obtained.

6. The cognitive enhancement-based BIM-GIS three-dimensional twin geological modeling method according to claim 1 is characterized in that: The process of predicting the spatial distribution of mineral resources based on the voxel three-dimensional model and combining historical exploration data and convolutional neural networks to predict the location and quantity of mineral resources is as follows: Extracting mineralization alteration characteristics from the voxel three-dimensional model, wherein the mineralization alteration characteristics include pyrite content, magnetic susceptibility gradient, and spatial distribution continuity index; Constructing a 3D convolutional neural network, processing the mineralization and alteration characteristic input values ​​into the neural network, and outputting a mineral resource probability distribution map, wherein the input of the 3D convolutional neural network is a 64×64×64 voxel block, the neural network includes a convolutional layer connected to a global average pooling layer and a fully connected layer, and the loss function of the neural network introduces a geological continuity constraint term; Based on the probability distribution map of mineral resources, the resource volume is calculated using a linear regression model, specifically: ; in, is the target variable, representing the estimated mineral resource based on the weighted fusion of drill core grade and 3D voxel density. It is the geological background benchmark value, reflecting the background resource volume determined by regional mineralization background statistics. is the regression coefficient, which is used to quantify the contribution weight of each mineral-controlling factor to resource distribution. It is a multi-source geological indicator including independent variables including geophysical parameters, geochemical parameters and structural parameters. is the geological uncertainty error term, which obeys the normal distribution , In order to optimize the number of ore-controlling factors, the stepwise regression method was used to screen and determine them.

7. The cognitive enhancement-based BIM-GIS three-dimensional twin geological modeling method according to claim 1, characterized in that: The process of performing multi-level optimization processing on the generated voxel three-dimensional model to construct the optimized model is as follows: A data smoothing method is used to eliminate irregular shapes in the voxel 3D model, and the voxel 3D model is compared with the field survey data to verify the accuracy of the voxel 3D model and perform error correction to obtain an optimized model.

8. An intelligent 3D geological data modeling system based on BIM-GIS fusion, used to implement the cognitive enhancement-based BIM-GIS 3D twin geological modeling method according to any one of claims 1 to 7, characterized in that: include: Data acquisition and processing module (1), 3D modeling module (2), visualization module (3) and data analysis module (4); The data acquisition and processing module (1) is used to acquire geological data based on different data sources and perform preprocessing to obtain preprocessed geological data; The three-dimensional modeling module (2) is used to convert the geological body into a voxel three-dimensional model according to the pre-processed geological data; The visualization module (3) is used to perform multi-level optimization processing on the generated voxel three-dimensional model, construct an optimized model, and display the optimized model in three dimensions through a three-dimensional visualization engine; The data analysis module (4) is used to predict the spatial distribution of mineral resources based on the voxel three-dimensional model, and to predict the location and quantity of mineral resources in combination with historical exploration data and convolutional neural networks, and to perform groundwater flow simulation analysis in combination with a groundwater flow model.

9. The intelligent three-dimensional geological data modeling system based on BIM-GIS fusion according to claim 8 is characterized in that: The visualization module (3) includes a stereoscopic view display module (31), a cross-sectional view display module (32), and a bottom perspective view display module (33); The stereoscopic view display module (31) is used to display the overall structure of the three-dimensional geological body using a voxel three-dimensional model; The cross-sectional view display module (32) is used to provide the user with a geological model cutting diagram of a specified cross section and to display geological hierarchy and ore body distribution information; The bottom perspective view display module (33) is used to display the stratum distribution of the geological body and distinguish different geological layers and mineral distributions through mapping of different colors and textures.

10. The intelligent three-dimensional geological data modeling system based on BIM-GIS fusion according to claim 8 is characterized in that: The data analysis module (4) includes a spatial relationship analysis module (41), a mineral resource prediction module (42) and a groundwater flow simulation module (43); The spatial relationship analysis module (41) is used to perform spatial analysis on three-dimensional geological data, construct spatial relationships between different geological bodies, and analyze and obtain spatial distribution characteristics of underground resources; The mineral resource prediction module (42) is used to predict the distribution of underground mineral resources based on historical exploration data and a convolutional neural network algorithm, and provide potential locations and quantities of mineral resources; The groundwater flow simulation module (43) is used to simulate the flow path of groundwater and the distribution of water resources in combination with hydrogeological data.

Images