Mineral resource metallogenic regularity simulation and target area prediction system based on big data

Abstract

The application discloses a kind of mineral resources metallogenic regularity simulation and target area prediction system based on big data, and relates to the technical field of geological exploration;Multi-source data acquisition and standardization module supports five kinds of data collection, which is transmitted to distributed database after pretreatment;Key elements are mined using improved Apriori algorithm, and association graph is constructed;Model is built based on Unity3D, and three types of dynamic simulation are integrated;CNN-LSTM model is used for three-level prediction;Support three-dimensional rendering and multifunctional display;Data, model and rules are automatically updated, forming an optimized closed loop.The present application integrates multi-source data to improve the comprehensiveness of ore-forming element identification, dynamically simulates the geological reality, accurately predicts AI and includes uncertainty analysis, is practical for visual interaction, iteratively optimized to continuously improve efficiency, reduces the cost of ineffective exploration, and efficiently guides mineral exploration.

Description

Technical Field

[0001] This invention relates to the field of geological exploration technology, and in particular to a big data-based system for simulating mineralization patterns and predicting target areas for mineral resources. Background Technology

[0002] In the field of mineral resource exploration, metallogenic regularity simulation and target area prediction are core components guiding exploration work and reducing mineral exploration costs. Traditional metallogenic analysis has long relied on single types of data, primarily geological maps and limited geophysical and geochemical data, neglecting the integrated utilization of high-value data such as remote sensing, borehole data, and microfacies analysis. For example, metallogenic analysis of metallic deposits often relies solely on stratigraphic and structural information from 1:50,000 geological maps, lacking detailed analysis of mineralization alteration anomalies from remote sensing and borehole core logging. This leads to incomplete identification of key metallogenic elements—such as failing to capture microfracture structures and elemental assemblage anomalies closely related to mineralization—thus affecting the comprehensiveness of metallogenic regularity judgments. Furthermore, traditional analysis relies heavily on manual interpretation, which is insufficient for large areas (e.g., exceeding 1000 km²). 2 The correlation analysis of ore-forming elements often takes weeks or even months, which is extremely inefficient and greatly affected by the experience of the analysts. Different people have significantly different judgments on the weight of ore-forming elements in the same area, resulting in poor consistency of analysis results and difficulty in forming a standardized understanding of ore formation.

[0003] Traditional metallogenic simulations often remain at a two-dimensional, static level, lacking a dynamic depiction of geological evolution. Existing simulations frequently employ GIS software for simple feature overlays, such as superimposing structural lines, stratigraphic boundaries, and geochemical anomalies to delineate potential metallogenic zones. However, they fail to consider the temporal evolution of tectonic activity—such as differences in fault activity rates at different stages and pressure changes during fold formation—and cannot simulate the migration paths of ore-forming fluids and the mineralization enrichment process. Even some three-dimensional simulations often set key parameters like fluid temperature and pressure to fixed values, ignoring their spatiotemporal dynamic changes, leading to simulation results that are out of sync with actual geological scenarios. For example, in simulating hydrothermal copper deposits, if the temperature gradient change (from 400℃ to 100℃) during fluid migration from deep to shallow layers is not considered, the location and scale of mineral precipitation cannot be accurately calculated. The simulation results often have a lower than 70% agreement with known deposits, limiting their reference value and making them unsuitable for effectively guiding target area prediction.

[0004] Technical shortcomings in the target area prediction stage further constrain exploration efficiency. Traditional prediction methods often employ simple models such as logistic regression and weighted evidence methods. These models struggle to handle complex nonlinear relationships from multiple data sources—such as the synergistic influence of geological structures, fluid parameters, and elemental content on mineralization probability—leading to low prediction accuracy, especially in areas with complex geological conditions (such as areas with overlapping tectonic phases), where the accuracy rate often falls below 60%. Furthermore, traditional prediction methods lack uncertainty analysis, only outputting the potential target area range without specifying the reliability of the prediction results. Users cannot determine whether a target area should be prioritized for exploration with high confidence or require verification with low confidence, easily leading to blind drilling and an invalid borehole rate exceeding 40%. In addition, traditional systems lack a closed-loop mechanism for prediction-verification-optimization, resulting in a disconnect between prediction results and actual exploration verification data (such as borehole mineralization). The inability to update model parameters and mineralization rules based on new verification data hinders continuous improvement in model accuracy, leaving it stagnant at its initial level and unable to adapt to exploration needs under different geological conditions. Summary of the Invention

[0005] The present invention proposes a big data-based mineral resource metallogenic law simulation and target area prediction system to solve the problems mentioned in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a mineral resource metallogenic regularity simulation and target area prediction system based on big data, comprising:

[0007] Mineral Big Data Acquisition and Standardization Module: Supports the acquisition of five types of data: geological, geophysical, geochemical, remote sensing, and borehole data. Geological data includes 1:50,000 geological maps and thin-section microscopic images of rocks. Geophysical data includes gravity, magnetic, and electrical resistivity data. Geochemical data includes the content of 39 elements. Remote sensing data includes 8-band multispectral data and InSAR topographic data. Borehole data includes core logging data and well logging curve data. Equipped with edge computing nodes, it enables real-time data preprocessing, including format conversion to GeoJSON / GeoTIFF, outlier removal using the 3σ criterion, filling in missing values ​​using the K-nearest neighbor algorithm, and data transmission to a distributed database via 5G / fiber optic.

[0008] The ore-forming element correlation analysis module employs the Apriori association rule algorithm, which sets clear minimum support and minimum confidence standards. The algorithm mines ore-forming elements and identifies ore-controlling structures and geophysical anomalies that meet the requirements. Simultaneously, it constructs an ore-forming element correlation map with a sufficient number of nodes and automatically labels core elements; for example, the correlation between faults and mineralized bodies has a high confidence level. In terms of analysis efficiency, the correlation analysis time for specific areas is controllable, meeting the time requirements of practical work.

[0009] The 3D mineralization dynamic simulation module is built using the Unity3D engine to construct a 1:10,000 scale 3D geological model with reasonable mesh accuracy settings. The module integrates three types of dynamic simulation functions: tectonic evolution, fluid transport dynamics, and mineralization enrichment. Each type of simulation has corresponding parameter range settings. During the simulation, users can adjust parameters in real time. After the simulation is completed, the module outputs the mineralization probability distribution and the coordinates of the mineral enrichment center. Not only does the resolution of the mineralization probability distribution match the model, but the error of the mineral enrichment center coordinates is also small, and the simulation results are accurately consistent with known mineral deposits.

[0010] AI Target Area Prediction and Priority Assessment Module: This module uses a CNN-LSTM fusion model for target area prediction and priority assessment. The model input consists of feature maps of various elements with specific specifications, covering four categories of features: geological, geophysical, geochemical, and simulation results. The model is pre-trained based on labeled samples of known mineral deposits, covering various mineral types, and supports incremental learning. The prediction process is divided into three levels: Level 1 clarifies the target area range and mineralization probability requirements; Level 2 calculates the target area potential value and determines the value range; Level 3 prioritizes the target areas based on the potential value.

[0011] The simulation prediction result visualization and interaction module supports 3D scene rendering with a rendering frame rate that meets the requirements for smooth display. It also provides five types of display functions, including spatial distribution display of ore-forming elements, playback of 3D ore-forming simulation animation, display of target area distribution heat map, comparison of borehole-simulation results, and plotting of reserve estimation value lines. Each display function has corresponding operation or display standards. It also supports data export, provides commonly used export formats, and implements distance measurement and profile extraction functions. It is compatible with PC and mobile devices and sets corresponding resolution standards for different terminals.

[0012] System Iteration and Optimization and Data Update Module: An automatic update mechanism is set up. In terms of data updates, new data from the Geological Survey is synchronized monthly, and incremental updates take less time. In terms of model optimization, after a certain number of exploration projects are completed, the AI ​​model is fine-tuned based on new validation data, using specific algorithms and setting corresponding parameters. In terms of rule updates, the mineralization association rule library is updated quarterly, equipped with a user feedback interface, supporting various feedback forms, and recording the target area validation results, forming a complete closed loop of data-simulation-prediction-validation-optimization.

[0013] Furthermore, it also includes:

[0014] The module for dynamic calculation of ore-forming element weights calculates the contribution weights of ore-forming elements in different geological periods through integration. The calculation formula is as follows: W i Let f be the dynamic weight of the i-th type of ore-forming element, t1 and t2 be the start and end times of the target ore-forming period, and f be the dynamic weight of the i-th type of ore-forming element. ig(t) represents the activity intensity of the i-th type of element at time t, g(t) represents the decay coefficient of the mineralization contribution time, and t0 represents the peak mineralization period. By calculating and distinguishing the mineralization contribution of elements in different periods, the mineralization simulation can be made to fit the actual geological evolution.

[0015] Target area mineralization potential correction module; The initial potential value is corrected based on borehole verification data, calculated as follows: Among them, P' j P is the corrected potential value of the j-th target region. j N represents the initial potential value of the target area; α is the correction coefficient, and N... j,min The number of boreholes encountering mineralization within the target area; N j,total This represents the total number of boreholes drilled within the target area.

[0016] Furthermore, the mineral big data acquisition and standardization module also includes a UAV remote sensing submodule; equipped with a 6-rotor UAV, a 12-band multispectral camera and a ground laser scanner, the UAV remote sensing submodule is used to acquire fine images of mineralization alteration and surface microstructure data. The fine images of mineralization alteration identify silicification and carbonatization; the surface microstructure data identifies fractures. The UAV data is stitched together in real time through edge computing nodes and also undergoes radiometric correction through a diffuse reflection reference plate. Then it is fused with satellite remote sensing data, and a pixel-level fusion algorithm is used to supplement the microscopic details of the remote sensing data.

[0017] Furthermore, the metallogenic element correlation analysis module also includes a spatiotemporal correlation mining submodule; it introduces a time dimension to expand traditional spatial correlation, models structural, stratigraphic, and mineralization data from different geological periods, including the Caledonian and Yanshanian periods, and calculates the cross-period element correlation degree R. i ,j(t1,t2); where t1 is the formation time of the early elements and t2 is the mineralization time of the later stages; the spatiotemporal correlation results are displayed in the form of a time series map, and the node size represents the element intensity, which helps to identify the mineralization mode of early ore-forming material reserves and later tectonic activation, avoiding the limitation of traditional static correlation ignoring the time evolution.

[0018] Furthermore, the three-dimensional dynamic simulation module for mineralization regularity also includes a fluid-rock interaction simulation submodule; using the PHREEQC hydrogeochemical model, the reaction rate between fluid and rock is calculated, and the input parameters include rock mineral composition, fluid chemical composition, temperature and pressure; during the simulation, the amount of mineral dissolution / precipitation, fluid pH changes and isotopic fractionation are calculated in real time, and the output results include a comparison diagram of mineral composition before and after the reaction and curves of fluid chemical parameter changes, providing quantitative basis for the analysis of the source and migration path of mineralization fluids.

[0019] Furthermore, the AI ​​target area prediction and priority assessment module also includes an uncertainty analysis submodule. Monte Carlo simulation is used to assess prediction errors, with 1000 iterations. Random noise is added to the input ore-forming element data, which includes geochemical element content and geophysical anomaly amplitude. The ore-forming element probability of the target area is recalculated in each iteration. The uncertainty analysis results are displayed synchronously with the target area prediction results, helping users objectively assess the reliability of the target area.

[0020] Furthermore, the simulation prediction result visualization and interaction module also includes a virtual borehole design submodule; it supports users in designing virtual boreholes in a 3D scene, with users inputting borehole diameter, dip angle, azimuth angle, and design depth; the system automatically calculates the intersection of the borehole trajectory and the mineralization simulation results, and outputs the predicted mineral type, grade, and thickness at the intersection; based on the virtual borehole data, it generates borehole columnar sections and well logging curve prediction diagrams, with the well logging curve prediction diagrams including density and resistivity curves; virtual boreholes are generated in batches, and can be exported as borehole design documents, which include coordinates, depth, and predicted mineralized sections, providing guidance for actual drilling operations.

[0021] Furthermore, the system iteration optimization and data update module also includes a model deviation diagnosis submodule. By comparing the AI ​​prediction results with the actual exploration verification data, three types of deviation indicators are calculated: location deviation, grade deviation, and reserve deviation. Location deviation is the distance between the predicted target area center and the actual mineralization center; grade deviation is the relative error between the predicted grade and the actual grade; and reserve deviation is the relative error between the predicted reserve and the actual reserve. For prediction samples with deviations exceeding the standard, SHAP value analysis is used to identify the cause of the deviation. Elements with an absolute SHAP value greater than 0.1 are the source of deviation, and a deviation correction report is automatically generated. The report includes data supplementation suggestions and model parameter adjustment schemes.

[0022] Furthermore, it also includes:

[0023] The user-to-user collaborative analysis module supports 5-20 users collaborating online simultaneously. It is implemented using the WebSocket protocol and features three levels of permissions: administrator, analyst, and viewer. Administrators can modify system parameters and review data; analysts can perform simulations and export results; and viewers can only browse results and add annotations. The module provides collaborative tools, including real-time annotation, online discussion, and task assignment. Real-time annotation supports text and graphic annotations. File transfer is supported, and task assignment assigns target area verification tasks to designated users with set completion deadlines. The collaborative process automatically records operation logs, including user, time, and operation content, and supports log backtracking.

[0024] Compared with existing technologies, the beneficial effects of this invention are:

[0025] In terms of data processing and ore-forming element identification, the system supports comprehensive collection and standardized integration of multi-source mineral big data, covering various types of data such as geology, geophysics, geochemistry, remote sensing, and borehole data. In particular, the UAV remote sensing submodule supplements the data with fine images of mineralization and alteration, and microscopic data such as surface microstructures, solving the problems of single data and lack of detail in traditional analysis. Data standardization processing ensures efficient fusion of data from different sources and in different formats, and edge computing nodes enable real-time data preprocessing, significantly improving data utilization efficiency. The ore-forming element correlation analysis module mines complex correlations between elements through improved algorithms, and expands traditional static correlations by combining spatiotemporal dimensions. It can accurately identify key mineralization models such as "early ore-forming material reserves - later tectonic activation," making the identification of ore-forming elements more comprehensive and the correlation analysis more in line with the actual geological evolution, laying a solid data and cognitive foundation for subsequent simulation and prediction.

[0026] In terms of simulating mineralization patterns, the system's constructed three-dimensional dynamic simulation framework completely overcomes the limitations of traditional two-dimensional static simulations. Based on the Unity3D engine, the three-dimensional geological model accurately depicts the spatial morphology of geological bodies. Three types of dynamic simulations (tectonic evolution, fluid migration, and mineralization enrichment) allow for real-time adjustment of key parameters. The simulation process fully considers the temporal dimension of geological evolution and the spatiotemporal dynamic changes of parameters, such as the gradient changes in fluid temperature and pressure along the migration path, and the adjustment of mineral precipitation rates with the intensity of tectonic activity. The fluid-rock interaction simulation submodule further enables quantitative calculation of the mineralization process, significantly improving the consistency between simulation results and actual deposits. It accurately outputs the mineralization probability distribution and mineral enrichment centers, providing a scientific basis for understanding mineralization mechanisms and delineating potential mineralization areas, thus avoiding the drawbacks of traditional simulations that are detached from reality and have low reference value.

[0027] In terms of target area prediction and practical value, the AI ​​target area prediction module employs a CNN-LSTM fusion model that effectively handles the nonlinear relationships of multi-source data. Combined with incremental learning, it adapts to different geological backgrounds, resulting in significantly higher prediction accuracy than traditional simple models. The addition of an uncertainty analysis submodule and a target area potential correction module makes the prediction results more reliable—the former indicates the reliability of the target area, helping users avoid the risks of blind exploration; the latter optimizes the potential value by combining borehole verification data, making the target area priority assessment more realistic. The simulation prediction result visualization and interaction module supports functions such as 3D scene rendering and virtual borehole design, which not only intuitively displays mineralization patterns and target area distribution but also provides precise guidance for actual drilling, reducing invalid boreholes. The multi-user collaborative analysis module solves the problems of difficult data sharing and asynchronous analysis in traditional multi-unit collaboration, improving team collaboration efficiency.

[0028] Furthermore, the data-simulation-prediction-verification-optimization closed loop constructed by the system iteration optimization and data update modules can continuously improve model accuracy and the applicability of mineralization rules based on new data and verification results, ensuring that the system maintains high performance in the long term and adapts to ever-changing exploration needs. Overall, this invention promotes the transformation of mineral resource exploration from experience-driven to data and model-driven, significantly improving the scientific nature of mineralization analysis, the accuracy of target area prediction, and the efficiency of exploration work, providing strong support for reducing mineral exploration costs and increasing the success rate of mineral exploration. Attached Figure Description

[0029] Figure 1 This is a schematic block diagram of the mineral resource metallogenic regularity simulation and target area prediction system based on big data proposed in this invention.

[0030] Figure 2 This is a schematic diagram comparing the identification accuracy of different data types of mineralization elements in the big data-based mineralization law simulation and target area prediction system proposed in this invention.

[0031] Figure 3 This diagram illustrates the comparison of the degree of agreement between different simulation types and known mineral deposits in the big data-based mineral resource metallogenic regularity simulation and target area prediction system proposed in this invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0034] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0035] Reference Figures 1 to 3 A big data-based mineral resource metallogenic regularity simulation and target area prediction system is applied to the metallogenic research and exploration target area delineation of metallic, non-metallic, and energy minerals. It solves the problems of traditional metallogenic analysis data being singular, static simulation, and low target area prediction accuracy. The system includes the following modules:

[0036] Multi-source mineral big data acquisition and standardization module: Supports the acquisition of five types of data: geological, geophysical, geochemical, remote sensing, and borehole data. Geological data includes 1:50,000 geological maps, with stratigraphic lithology and structural lines accurate to within 50m, and thin-section microscopic images of rocks with a resolution of 2048×2048 pixels; geophysical data includes gravity, magnetic, and electrical resistivity data, with gravity data accuracy of ±0.1mGal, magnetic data accuracy of ±1nT, and electrical resistivity data resolution of 10m×10m; geochemical data includes the content of 39 elements, with a detection limit of no more than 0.01ppm; remote sensing data includes 8-band multispectral data and InSAR topographic data, with multispectral data spatial resolution of 15m and InSAR topographic data elevation accuracy of ±2m; borehole data includes core logging data and well logging curve data, with core logging depth error not exceeding 0.1m and well logging curve density and resistivity sampling intervals of 0.1m. This module is equipped with edge computing nodes with a computing power of 16 TOPS, enabling real-time data preprocessing. Preprocessing includes format conversion to GeoJSON / GeoTIFF, outlier removal using the 3σ criterion, and missing value imputation using the K-nearest neighbor algorithm, with an imputation error not exceeding 5%. Data is transmitted to a distributed database via 5G / fiber optics at a transmission rate of no less than 100Mbps. The distributed database has a capacity of 100TB, with read / write latency not exceeding 10ms. After standardization, the data complies with the "Mineral Resources Big Data Sharing Specification".

[0037] The ore-forming element correlation analysis module employs an improved Apriori association rule algorithm with a minimum support of 20% and a minimum confidence of 80%. This algorithm is used to mine key ore-forming elements. Ore-controlling structures and favorable strata are identified from geological elements. Ore-controlling structures must meet the following criteria: fault length of at least 1 km and fold core width of at least 500 m. Favorable strata include the Anshan Group corresponding to iron ore and the Kunyang Group corresponding to copper ore. Gravity high anomalies and magnetic anomalies are extracted from geophysical elements, with gravity high anomaly amplitudes at least 5 mGal and magnetic anomaly intensities at least 100 nT. Element anomaly combinations are delineated from geochemical elements, including Cu-Pb-Zn co-occurrence anomalies and Au-As-Sb indicator anomalies. Mineralization and alteration are identified from remote sensing elements, with hydroxyl and iron staining anomaly extraction accuracy of at least 90%. This module constructs a correlation map of ore-forming elements, with no fewer than 1000 nodes and edge weights representing correlation confidence levels. It can automatically label core elements; for example, the correlation confidence level between faults and mineralized bodies is no less than 95%, and the correlation analysis takes no more than 2 hours per 1000km. 2 area.

[0038] The 3D mineralization dynamic simulation module: Based on the Unity3D engine, a 1:10,000 scale 3D geological model is constructed with a mesh resolution of 10m×10m×5m, integrating three types of dynamic simulations. Tectonic evolution simulation includes calculations of fault activity rates and fold formation pressures. Fault activity rates range from 0.1-1 mm / year, and fold formation pressures range from 20-50 MPa, with a simulation timescale of 1-100 Ma. Fluid transport dynamic simulation calculates fluid velocity, temperature, and pressure. Fluid velocity ranges from 0.1-1 m / day, temperature ranges from 100-400℃, and pressure ranges from 5-20 MPa, solved using the finite element method with an iteration step of 1000 years. Mineralization enrichment simulation, based on the mass balance equation, calculates mineral precipitation rates ranging from 0.01-0.1 kg / m³. 2 • The simulation period is 1-10 Ma. The simulation process supports real-time parameter adjustment, and parameters such as fluid salinity and tectonic stress direction can be adjusted. The output simulation results include mineralization probability distribution and mineral enrichment center coordinates. The mineralization probability distribution resolution is 10m×10m×5m, the mineral enrichment center coordinate error does not exceed 20m, and the simulation results are in agreement with known deposits at a rate of not less than 85%.

[0039] The AI ​​target area prediction and priority assessment module employs a CNN-LSTM fusion model. The model input is a 256×256×12 multi-element feature map, which includes four categories of features: geological, geophysical, geochemical, and simulation results. The model is pre-trained on over 100,000 labeled samples of known mineral deposits, covering 12 types of minerals including iron, copper, and gold. It supports incremental learning, with an update time of no more than one hour when adding 500 new samples. The prediction process is divided into three levels: the first level delineates the target area, with a target area of ​​no less than 1 km².2 The probability of mineralization is no less than 70%; the second level calculates the target area potential value, which comprehensively considers the probability of mineralization, element integrity, and transportation accessibility, with a value range of 0-100; the third level assesses priority, arranged in descending order of potential value, with the top 20% being high priority. The model's prediction accuracy is no less than 88%, an improvement of no less than 30% compared to the traditional logistic regression model, and the target area prediction time does not exceed 1 hour / 500km. 2 area.

[0040] The simulation prediction result visualization and interaction module supports 3D scene rendering with a frame rate of no less than 30fps and provides five types of display functions. The spatial distribution display of ore-forming elements supports layered display of strata, structures, and anomalies, with transparency adjustable from 0-100%. The 3D ore-forming simulation animation playback speed is 1-100Ma / second, supporting pause and rewind operations. The target area distribution heatmap color scale range is 0-100%, with 5% intervals, and high-priority target areas can be marked. The borehole-simulation result comparison supports slicing along the borehole trajectory to display the consistency between simulated mineralization and actual core mineralization. Reserve estimation contour lines are drawn at 1 million ton intervals with an error of no more than 10%. This module supports data export in formats including Shapefile, BIM, and Excel. It also supports distance measurement and profile extraction, with a distance measurement accuracy of no more than 1m. Profiles can be extracted in any direction, with a generation time of no more than 10 seconds. It is compatible with PC and mobile displays, with a PC resolution of 3840×2160 and a mobile resolution of 2340×1080.

[0041] The system iteration optimization and data update module features an automatic update mechanism. Data updates are synchronized monthly with new data from the Geological Survey, with incremental updates taking no more than 4 hours. Model optimization involves fine-tuning the AI ​​model based on new validation data after every 5 exploration projects, using a small-batch gradient descent algorithm with a batch size of 32 and 50 iterations, resulting in an accuracy improvement of at least 2%. Rule updates include quarterly updates to the mineralization association rule library, adding at least 10 new associations. This module is equipped with a user feedback interface, supporting text and image feedback with a response time of no more than 24 hours. It records target area validation results, including borehole ore encounter rate and reserve deviation, forming a closed loop of "data-simulation-prediction-validation-optimization." After one year of system operation, the target area prediction accuracy stabilizes at no less than 90%.

[0042] This invention also includes:

[0043] The module for dynamic calculation of ore-forming element weights calculates the contribution weights of ore-forming elements in different geological periods through integration. The calculation formula is as follows: Where Wi is the dynamic weight of the i-th type of ore-forming element, with a value range of 0-1; t1 and t2 are the start and end times of the target ore-forming period, in Ma, for example, the BIF ore-forming period for iron ore is 2500-2300 Ma; f i (t) represents the activity intensity of the i-th type of element at time t, with a value ranging from 0 to 1. During periods of fault activity, the value is 0.8 to 1.0, and during periods of quiescence, it is 0.1 to 0.3. g(t) is the time decay coefficient of mineralization contribution, with a value ranging from 0 to 1. During peak mineralization, the coefficient is 1.0, and it decays exponentially away from the mineralization period, specifically g(t) = e^-0.1(t-t0), where t0 is the time of peak mineralization. This calculation can distinguish the mineralization contribution of elements in different periods. For example, for Paleoproterozoic iron ore, the weight of the Anshan Group strata increases from 0.6 to 0.8, while the weight of later tectonic structures decreases from 0.4 to 0.2, making the mineralization simulation more consistent with actual geological evolution.

[0044] Target area mineralization potential correction module; The initial potential value is corrected based on borehole verification data, calculated as follows: Among them, P' j P represents the corrected potential value for the j-th target region, ranging from 0 to 100. j α is the initial potential value of the target area; α is the correction coefficient, ranging from 0.1 to 0.3, with 0.2 for metallic minerals and 0.1 for non-metallic minerals; Nj,min is the number of ore-bearing boreholes in the target area; N j,total This represents the total number of boreholes within the target area. For example, if the initial potential value of a copper mine target area is 80, with 12 ore-bearing boreholes and a total of 15 boreholes, and α = 0.2, then P'j = 80 × [1 + 0.2 × ln(1 + 12 / 15)] ≈ 80 × 1.158 ≈ 92.6. This correction more accurately reflects the actual potential of the target area and avoids deviations caused by neglecting verification data in the initial prediction.

[0045] In this invention, the multi-source mineral big data acquisition and standardization module also includes a UAV remote sensing submodule. This submodule is equipped with a 6-rotor UAV, a 12-band multispectral camera, and a ground-based laser scanner. The UAV has a flight time of 60 minutes and a positioning accuracy of ±0.5m; the 12-band multispectral camera has a spectral range of 400-1000nm and a spatial resolution of 0.5m at an altitude of 100m; the ground-based laser scanner has a point cloud density of 50 points / m². 2The ranging accuracy is ±3cm. The UAV remote sensing submodule is used to acquire fine images of mineralization and alteration and surface microstructure data. The fine images of mineralization and alteration can identify silicification and carbonatization with an accuracy of no less than 92%. The surface microstructure data can identify fractures, with fractures of no less than 0.5m in length and no less than 0.1m in width being identifiable. UAV data is stitched in real time through edge computing nodes with a stitching error of no more than 1m. It is also radiometrically corrected using a diffuse reflection reference plate with a reflectivity error of no more than 3%. Then it is fused with satellite remote sensing data using a pixel-level fusion algorithm. The fused data has a resolution of 0.5m and an information retention rate of no less than 95%, supplementing the microscopic details of large-scale remote sensing data and improving the precision of mineralization element identification.

[0046] In this invention, the metallogenic element correlation analysis module further includes a spatiotemporal correlation mining submodule. This submodule introduces a time dimension to extend traditional spatial correlation, modeling tectonic, stratigraphic, and mineralization data from different geological periods, including the Caledonian and Yanshanian periods, and calculating the cross-period element correlation degree R. i The correlation coefficient is defined as j(t1,t2), where t1 represents the formation time of early-stage elements, and t2 represents the later-stage mineralization time. The correlation coefficient ranges from -1 to 1, with a correlation coefficient greater than 0.7 indicating a strong positive correlation. For example, the correlation coefficient between the Yanshanian fault and the Caledonian strata was found to be 0.85. The Yanshanian fault formed at time t2 = 150 Ma, and the Caledonian strata formed at time t1 = 450 Ma. This result indicates that later-stage fault activation controlled mineralization enrichment in earlier-stage strata. The spatiotemporal correlation results are presented in the form of a time series map. The time axis of the time series map has an accuracy of 10 Ma, and the node size represents the element intensity, helping to identify the mineralization model of "early-stage ore-forming reserves - later-stage tectonic activation" and avoiding the limitations of traditional static correlation that ignores temporal evolution.

[0047] In this invention, the three-dimensional mineralization dynamic simulation module further includes a fluid-rock interaction simulation submodule. This submodule uses the PHREEQC hydrogeochemical model to calculate the reaction rate between the fluid and rock. Input parameters include rock mineral composition, fluid chemical composition, temperature, and pressure. For example, the rock mineral composition might be 25% plagioclase, 30% quartz, and 15% calcite; the fluid chemical composition includes Ca... 2 ⁺、Mg 2 ⁺ Plasma concentration, in mmol / L; temperature range: 100-350℃; pressure range: 5-18MPa. During the simulation, mineral dissolution / precipitation, fluid pH changes, and isotopic fractionation are calculated in real time. For example, the calcite precipitation rate is 0.005-0.05 mol / m³. 2 •h, the fluid pH value varies from 4 to 9, and the δ in isotopic fractionation 18The range of O variation is from -10‰ to +10‰. The output results include a comparison chart of mineral composition before and after the reaction, and curves showing changes in fluid chemical parameters. The agreement rate with the isotopic data of rocks sampled in the field is no less than 80%, providing quantitative basis for the analysis of the source and migration path of ore-forming fluids.

[0048] In this invention, the AI ​​target area prediction and priority assessment module also includes an uncertainty analysis submodule. This submodule uses Monte Carlo simulation to assess prediction errors, with 1000 iterations. Random noise is added to the input ore-forming element data, including geochemical element content and geophysical anomaly amplitude, with a noise intensity of 0.01-0.1 times the data standard deviation. The ore-forming probability of the target area is recalculated in each iteration. The probability distribution of 1000 iterations is statistically analyzed, and the standard deviation σp is calculated. σp below 5% is considered low uncertainty, 5%-10% is medium uncertainty, and above 10% is high uncertainty. High-uncertainty target areas are automatically marked in orange, while low-uncertainty target areas are marked in green. Data supplementation suggestions are also provided, such as adding 3-5 boreholes for verification in high-uncertainty target areas. The uncertainty analysis results are displayed synchronously with the target area prediction results, helping users objectively assess the reliability of the target area and avoid blind exploration.

[0049] In this invention, the visualization and interaction module for simulation prediction results also includes a virtual borehole design submodule. This submodule allows users to design virtual boreholes in a 3D scene. Users input the borehole diameter, inclination angle, azimuth angle, and design depth. The borehole diameter ranges from 50-200 mm, the inclination angle from 0-90°, the azimuth angle from 0-360°, and the design depth from 100-1000 m. The system automatically calculates the intersection of the borehole trajectory and the mineralization simulation results with a calculation accuracy of no more than 0.5 m, and outputs the predicted mineral type, grade, and thickness at the intersection. Based on the virtual borehole data, a borehole columnar section and a well logging curve prediction map are generated. The borehole columnar section has a scale of 1:200, and the well logging curve prediction map includes density and resistivity curves, with a consistency of no less than 85% with known actual boreholes. Virtual boreholes can be generated in batches, up to 100 at a time. They can be exported as borehole design documents, which include coordinates, depth, and predicted mineralized sections, providing guidance for actual drilling operations and reducing the number of invalid boreholes by no less than 25%.

[0050] In this invention, the system iterative optimization and data update module also includes a model deviation diagnosis submodule. This submodule calculates three types of deviation indicators by comparing AI prediction results with actual exploration verification data: location deviation, grade deviation, and reserve deviation. Location deviation is the distance between the predicted target area center and the actual mineralization center; a distance not exceeding 50m is considered optimal. Grade deviation is the relative error between the predicted grade and the actual grade; a relative error not exceeding 10% is considered optimal. Reserve deviation is the relative error between the predicted reserves and the actual reserves; a relative error not exceeding 15% is considered optimal. For prediction samples with excessive deviations, SHAP value analysis is used to identify the causes of the deviations. Elements with an absolute SHAP value greater than 0.1 are considered key deviation sources. For example, the deviation in a gold mine originated from errors in the detection of Au elements in geochemical exploration, with a SHAP value of 0.15. A deviation correction report is automatically generated, which includes data supplementation suggestions and model parameter adjustment schemes. Data supplementation suggestions include re-collecting geochemical samples, and model parameter adjustment schemes include increasing the feature weight of Au elements. After correction, the model deviation is reduced by no less than 30%, continuously improving prediction accuracy.

[0051] This invention also includes:

[0052] The inter-user collaborative analysis module supports 5-20 users collaborating online simultaneously. Based on the WebSocket protocol, the collaboration latency is no more than 500ms. It features three levels of permissions: administrator, analyst, and viewer. Administrators can modify system parameters and review data; analysts can perform simulations and export results; and viewers can only browse results and add annotations. The module provides collaborative tools, including real-time annotation, online discussion, and task assignment. Real-time annotation supports text and graphic annotation with a positioning accuracy of no more than 10m. Online discussion message sending latency is no more than 1 second, and file transfer is supported with a file size limit of 100MB. Task assignment allows assigning target area verification tasks to specific users and setting completion deadlines. The collaborative process automatically records operation logs, including user, time, and operation content. Log backtracking is supported, and the logs are stored for at least one year. This is suitable for multi-unit joint exploration projects, improving team collaboration efficiency by at least 40%.

[0053] Specific implementation of the big data-based mineral resource metallogenic regularity simulation and target area prediction system:

[0054] Example 1: Simulation of metallogenic regularity and target area prediction of a large copper mine in Jiangxi Province (multi-phase tectonic superposition zone)

[0055] This example focuses on a copper mine exploration project in Jiangxi Province, covering an area of ​​120 km². 2This is a hydrothermal copper mineralization belt formed by multiple superimposed tectonic phases, having experienced two major tectonic periods: the Caledonian and Yanshanian. Three known deposits (all hydrothermal infill type) exist. The objective is to guide subsequent drilling operations by systematically simulating the mineralization patterns and delineating high-priority target areas. The specific implementation process is as follows:

[0056] 1. System Deployment and Multi-Source Data Acquisition System Hardware Deployment: Edge computing nodes utilize NVIDIA Jetson AGXXavier (32 TOPS computing power); the distributed database employs HBase (150TB capacity, 8ms read / write latency); terminal devices include 3 industrial-grade PCs (3840×2160 resolution) and 5 mobile tablets (2340×1080 resolution); data interaction is achieved via a 5G private network (150Mbps transmission rate). The multi-source mineral big data acquisition and standardization module collects data according to five categories:

[0057] Geological data: Collect 1:50,000 geological maps (strata and lithology are Kunyang Group slightly metamorphic rocks, structural line accuracy 45m), and collect thin section microscopic images of rocks (2048×2048 pixels, 2000 images in total, including microscopic features of chalcopyrite and pyrite).

[0058] Geophysical data: Gravity data were collected using a CG-5 gravimeter (accuracy ±0.08mGal, survey network density 200m×200m), magnetic data were collected using a GSM-19T magnetometer (accuracy ±0.8nT), and electrical data were collected using an EH4 electrical resistivity meter (resolution 10m×10m, detection depth 500m).

[0059] Geochemical data: 1200 soil samples were collected, and the contents of 39 elements were detected by ICP-MS (the detection limit for Cu was 0.008 ppm, and the detection limits for Pb and Zn were 0.01 ppm).

[0060] Remote sensing data: Acquire Sentinel-28 band multispectral data (spatial resolution 15m) and ALOS-2 InSAR topographic data (elevation accuracy ±1.8m).

[0061] Drilling data: Data from 15 known boreholes in the area were compiled. The core logging depth error was 0.08m, and the sampling interval for logging curves (density and resistivity) was 0.1m. Among them, 8 boreholes encountered mineralization (chalcopyrite grade 0.5%-2.3%).

[0062] In the data preprocessing stage, the edge computing nodes processed the data according to the following process: the format was converted to GeoJSON (vector data) and GeoTIFF (raster data); outliers in the geochemical data were removed using the 3σ criterion (Cu element outliers > 100ppm, a total of 28 were removed); missing values ​​in the gravity data were filled using the K-nearest neighbor algorithm (K=5) (missing rate 3%, filling error 4.2%); finally, the standardized data was uploaded to the distributed database, which took 3.5 hours.

[0063] 2. Correlation Analysis and Weight Calculation of Ore-forming Elements

[0064] The ore-forming element correlation analysis module uses the improved Apriori algorithm, setting a minimum support of 20% and a minimum confidence of 80% to mine key elements:

[0065] Geological elements: The ore-controlling faults (length ≥ 1.2 km, 12 in total) and the Kunyang Group shallow metamorphic rock strata (distributed in 65% of the area) were identified as favorable elements, with a 96% confidence level in the correlation between faults and mineralization.

[0066] Geophysical features: High gravity anomalies (amplitude ≥ 5.2 mGal, 5 locations in total) and magnetic anomalies (intensity ≥ 105 nT, 3 locations in total) were extracted. The confidence level of the correlation between gravity anomalies and mineralization is 92%.

[0067] Geochemical elements: Cu-Pb-Zn co-occurrence anomalies were delineated (Cu≥50ppm, Pb≥20ppm, Zn≥50ppm, a total of 8 locations), and the confidence level of the correlation between the anomaly combination and mineralization was 94%.

[0068] Remote sensing elements: Identified hydroxyl (2.2μm band anomaly) and iron staining (0.86μm band anomaly) alteration with an extraction accuracy of 91% and a confidence level of 88% for the correlation between alteration zones and mineralization. Simultaneously, a dynamic calculation module for the weights of mineralization elements was activated, targeting the Yanshanian period (t1=180Ma, t2=120Ma), to calculate the weights of key elements: Among them, the Kunyang Group strata (i=1) showed an activity intensity of f1(t) = 0.9 during the Yanshanian period (a consistently stable distribution), with a time decay coefficient g(t) = e⁻ 0 . 1 (t-150) (t0=150Ma is the peak period of mineralization), integrating, we get W1=∫ 180 120 0.9×e⁻ 0 . 1(t-150)dt≈0.82; the active intensity of the Yanshanian faults (i=2) is f2(t)=0.95 (active period t=150-130Ma), and the integral yields W2≈0.78; the active intensity of the Caledonian tectonic structures (i=3) is f3(t)=0.3, and the integral yields W3≈0.21. The weighting results indicate that the Yanshanian strata and faults are the core metallogenic elements, providing a basis for subsequent simulations.

[0069] The spatiotemporal correlation mining submodule further discovered that the correlation degree R between the Yanshanian fault (t2=150Ma) and the Caledonian strata (t1=460Ma) is 0.87, confirming the mineralization model of "later fault activation controlling earlier strata mineralization". When the time series map shows this correlation, the node size is set according to the element intensity (fault node diameter 8mm, strata node diameter 6mm), and the time axis accuracy is 10Ma.

[0070] 3. Three-dimensional dynamic simulation of mineralization and fluid-rock interaction

[0071] The 3D mineralization dynamic simulation module uses Unity3D 2022.3 to build a 1:10,000 3D geological model with a mesh resolution of 10m×10m×5m. After importing standardized data, three types of simulations are performed:

[0072] Tectonic evolution simulation: The fault activity rate was set to 0.8 mm / year (Yanshanian) and 0.2 mm / year (Caledonian), the fold formation pressure was 35 MPa, the simulation time scale was 180-120 Ma, and the output fault activity trajectory and stratigraphic deformation map showed 87% agreement with the tectonic morphology revealed by known boreholes.

[0073] Fluid transport dynamic simulation: Input fluid parameters (flow velocity 0.8 m / day, temperature 320-150℃, pressure 15-8 MPa), use finite element method to solve (iteration step size 1000 years), simulate the fluid transport path from deep fracture to shallow fracture, and find that the fluid accumulates in the fault intersection area (a total of 4 accumulation areas).

[0074] Mineralization enrichment simulation: Based on the mass balance equation, the chalcopyrite precipitation rate is set to 0.08 kg / m³. 2 • The simulation period is 150-130 Ma, and the output mineralization probability distribution (resolution 10m×10m×5m) shows that the high probability area (≥70%) is concentrated in the fault intersection and fluid accumulation area.

[0075] The fluid-rock interaction simulation submodule uses the PHREEQC model. Input parameters: rock mineral composition (plagioclase 28%, quartz 32%, calcite 12%), fluid chemical composition (Ca). 2 ⁺=5mmol / L, Mg 2 ⁺=2mmol / L, Cu 2(⁺=0.1mmol / L), temperature 280℃, pressure 12MPa. Simulation calculation yielded a calcite precipitation rate of 0.03mol / m³. 2 •h, the fluid pH increased from 5.2 to 6.8, δ 18 O increased from -8‰ to -2‰, consistent with rock isotope data from field sampling (δ¹⁸O). 18 The O=-7‰ to -3‰ match rate of 82% confirms that the fluid source is a mixture of deep magma water and groundwater.

[0076] 4. AI target area prediction and result verification

[0077] The AI ​​target area prediction and priority evaluation module adopts a CNN-LSTM fusion model. The input is a 256×256×12 multi-element feature map (4D geological, 3D geophysical, 3D geochemical, and 2D simulation results). It is pre-trained based on 120,000 known mineral deposit samples (including 40,000 copper deposit samples from Jiangxi and Yunnan) with a learning rate of 1e⁻. 4 The iteration count was 200. 500 new samples (including 300 mineral-bearing samples) were added to this region for incremental learning, with an update time of 55 minutes.

[0078] The prediction process is divided into three levels: Level 1 delineates 8 target areas, covering an area of ​​1.2-2.5 km². 2 The mineralization probability is 72%-85%; the initial potential value calculated at the secondary level (considering mineralization probability, element integrity, and transportation accessibility) has a maximum value of 88 (target area 3) and a minimum value of 65 (target area 8); the potential value is corrected at the tertiary level based on borehole verification data, taking target area 3 as an example (initial Pj=88, 5 ore-bearing boreholes, 6 total boreholes, α=0.2):

[0079] After correction, targets are sorted in descending order of potential value, with the first two (target area 3 and target area 5) having high priority. The uncertainty analysis submodule uses Monte Carlo simulation (1000 iterations). Noise is added to target area 3 (0.05 times the standard deviation of the noise intensity of Cu element in geochemical exploration), and σP is calculated to be 4.2% (low uncertainty, marked in green); σP for target area 8 is 9.5% (medium uncertainty, marked in yellow). It is recommended to add 4 boreholes for verification.

[0080] The simulation prediction results visualization and interaction module generates a 3D scene (35fps): The heat map of target area 3 shows that the mineralization probability area is concentrated in the fault intersection zone, with 85% of the area showing a mineralization thickness of 3.5m along the borehole trajectory, which matches the actual core mineralization thickness of 3.2m with a 91% degree of agreement; 10 virtual boreholes were designed (diameter 150mm, dip angle 60°, depth 500m), and the predicted mineralization section of virtual borehole No. 5 in target area 3 is 200-203m with a grade of 1.8%. Subsequent actual drilling verified that the mineralization section is 201-204m with a grade of 1.7%, and the error meets the requirements.

[0081] 5. Comparison of Implementation Results

[0082] Table 1: Comparison of Key Indicators between Metallogenic Simulation and Target Area Prediction in a Copper Mine in Jiangxi Province

[0083] Evaluation indicators Traditional manual methods This invention system Mineralization element identification rate 75% 94% The simulation matches the known mineral deposits. 68% 87% Target area prediction accuracy 58% 92% Ineffective drilling rate 42% 18% Total project time 60 days 8 days

[0084] Table 1 shows that this system has significant advantages in copper deposit scenarios with multiple superimposed tectonic phases. Traditional manual methods, due to incomplete data integration and static simulation, have an ore-forming element identification rate of only 75% and a simulation consistency of less than 70%. This system, through multi-source data fusion and dynamic simulation, improves the identification rate to 94% and the consistency to 87%, accurately capturing the ore-controlling patterns of multiple tectonic phases. The target area prediction accuracy has increased from 58% to 92%, and the invalid borehole rate has decreased by 24%, attributed to the nonlinear fitting capability of the AI ​​model and the risk avoidance of uncertainty analysis. The total project time has been shortened from 60 days to 8 days, with an efficiency improvement of 7.5 times, fully demonstrating the core value of the system of "data-driven, dynamic simulation, and accurate prediction," providing an efficient solution for copper exploration in complex tectonic areas.

[0085] Example 2: Simulation of mineralization regularity and target area prediction of a large limestone deposit in Hebei Province (sedimentary deposit)

[0086] This example focuses on a limestone mine exploration project in Hebei Province, covering an area of ​​150 km². 2 This is a Cambrian marine sedimentary limestone deposit, with mineralization controlled by sedimentary facies. The known ore layer thickness is 5-20m. The objective is to simulate the evolution of the sedimentary environment, predict high-purity limestone target areas, and simultaneously achieve multi-unit collaborative analysis. The specific implementation process is as follows:

[0087] 1. System Deployment and Multi-Source Data Acquisition

[0088] The system deployment includes edge computing nodes (NVIDIA Jetson Xavier NX, 21 TOPS computing power), an HBase distributed database (200TB capacity, 9ms read / write latency), and 8 collaborative terminals (3 PCs, 5 tablets), transmitted via 5G+fiber dual-mode (200Mbps speed). Multi-source data acquisition focuses on the characteristics of sedimentary deposits.

[0089] Geological data: 1:50,000 geological map (Cambrian Mantou Formation and Zhangxia Formation strata, accuracy 40m), rock thin section images (2048×2048 pixels, 1500 images, including calcite grain structure).

[0090] Geophysical data: gravity data (accuracy ±0.1mGal, survey network 150m×150m), seismic exploration data (resolution 5m×5m, detection depth 300m).

[0091] Geochemical data: 1800 rock samples were collected, and the CaO content (detection limit 0.1%) and MgO content (detection limit 0.05%) were tested.

[0092] Remote sensing data: Landsat-88 band multispectral (15m resolution), Sentinel-1 InSAR topography (elevation accuracy ±2m);

[0093] Drilling data: 20 known boreholes, core logging depth error 0.09m, logging curves (density 2.6-2.8 g / cm³). 3 (Resistivity 500-1000Ω・m) Sampling interval 0.1m, ore was encountered in 12 boreholes (CaO content ≥52%).

[0094] Data preprocessing: After format conversion, geochemical CaO outliers (>56% or <48%, a total of 32) were removed; K-nearest neighbor algorithm (K=4) was used to fill missing values ​​in seismic data (missing rate 2.5%, error 3.8%); after standardization, the data was uploaded to the database, taking 4 hours.

[0095] 2. Correlation analysis of ore-forming elements and sedimentary simulation

[0096] The ore-forming element correlation analysis module, using the improved Apriori algorithm (18% support, 78% confidence), extracted:

[0097] Geological elements: Zhangxia Formation oolitic limestone (distribution area 70%) and tidal flat sedimentary facies are favorable elements, with a 95% confidence level in their correlation with limestone mineralization;

[0098] Geophysical features: low gravity anomalies (amplitude -2 to -5 mGal, 6 locations), continuous seismic phase axes (length ≥ 5 km, 4 locations), correlation confidence level 90%;

[0099] Geochemical elements: 10 combined anomalies with CaO ≥ 52% and MgO ≤ 2%, with a correlation confidence level of 96%;

[0100] Remote sensing elements: absorption characteristics of carbonate rocks (2.3μm band anomaly), extraction accuracy 93%, correlation confidence 89%.

[0101] The weight calculation of ore-forming elements is based on the Cambrian mineralization period (t1=520Ma, t2=490Ma), and the weight of the Zhangxia Formation is W=∫ 520 490 0.95×e⁻ 0 . 1 (t-505)dt≈0.85, the later tectonic weight W=0.18, clearly indicating that the sedimentary facies is the core element.

[0102] The three-dimensional mineralization dynamic simulation focuses on the evolution of the sedimentary environment: the Unity3D model has a mesh resolution of 10m×10m×3m, simulating the Cambrian marine transgression-regression process (timescale 520-490 Ma), setting the sea-level change rate to 0.1 mm / year and the sedimentation rate to 0.5 m / ka, outputting the distribution of tidal flats and shallow marine sedimentary facies, which matches 89% with the sedimentary facies revealed by known boreholes; the fluid simulation considers the influence of groundwater (flow velocity 0.3 m / day, temperature 25-50℃), calculating the calcite cementation rate to be 0.02 kg / m³. 2 • The simulated ore layer thickness is 5-18m, with a deviation of ≤1.5m from the actual ore layer thickness.

[0103] 3. AI target area prediction and multi-user collaboration

[0104] The AI ​​target area prediction module uses a "CNN-LSTM" model. The input feature map includes sedimentary facies, geochemical, geophysical, and simulation results. It is pre-trained based on 80,000 sedimentary deposit samples and incrementally learned with 600 new samples from the local area. The update time is 50 minutes. It predicts and delineates 12 target areas with a mineralization probability of 70%-88% and an initial potential value of up to 90 (target area 6).

[0105] Target potential correction is taken for target area 6 as an example (Pj=90, 8 boreholes with mineralization, 10 total boreholes, α=0.1):

[0106] After correction, the first three locations are high priority. Uncertainty analysis shows that the target area 6σP=3.8% (green) and the target area 11σP=8.5% (yellow).

[0107] The multi-user collaborative analysis module is now enabled, allowing three organizations (an exploration institute, a mining company, and a university) to be online simultaneously.

[0108] Administrator (Exploration Institute): Review data, assign tasks, and distribute target area verification tasks to mining companies; Analyst (Mining Company): Design virtual boreholes (20, 200mm in diameter, 300m deep) to predict the ore-bearing section of the target area at 120-125m with a CaO content of 54%;

[0109] Viewer (University): Add annotation (annotate the sedimentary facies boundary of target area 6), and discuss the impact of sedimentary environment on mineralization online (message delay 0.8s).

[0110] Operation logs are automatically recorded (including user, time, and operation), and the retention period is set to 2 years, improving collaboration efficiency by 50% compared to traditional email communication.

[0111] 4. Comparison of Implementation Results

[0112] Table 2: Comparison of Key Indicators between Ore-forming Simulation and Target Area Prediction of a Limestone Mine in Hebei Province

[0113] Evaluation indicators Traditional manual methods This invention system Sedimentary phase identification accuracy 72% 93% Target area prediction accuracy 62% 91% Ineffective drilling rate 38% 15% Multi-unit collaborative efficiency Low (email communication) High (real-time collaboration) Total project time 50 days 7 days

[0114] Table 2 data demonstrates the system's advantages in sedimentary limestone deposit scenarios. Traditional methods rely on manual interpretation for sedimentary facies identification, resulting in an accuracy of only 72% and a target area prediction accuracy of 62%. This system, through dynamic simulation of the sedimentary environment and AI models, improves identification accuracy to 93% and prediction accuracy to 91%, accurately locating high-purity limestone target areas. The invalid borehole rate decreased from 38% to 15%, attributed to virtual borehole design and potential correction. Multi-user collaboration solves the problem of inefficient traditional communication, reducing total project time from 50 days to 7 days, a 7-fold efficiency improvement. Simultaneously, the system's detailed simulation of sedimentary facies provides a basis for limestone ore quality prediction (CaO content), meeting the high purity requirements of non-metallic minerals and laying the foundation for subsequent mining planning.

[0115] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A system for simulating mineralization patterns and predicting target areas based on big data, characterized in that, include: Mineral Big Data Acquisition and Standardization Module: Supports the acquisition of five types of data: geological, geophysical, geochemical, remote sensing, and borehole data. Geological data includes 1:50,000 geological maps and thin-section microscopic images of rocks. Geophysical data includes gravity, magnetic, and electrical resistivity data. Geochemical data includes the content of 39 elements. Remote sensing data includes 8-band multispectral data and InSAR topographic data. Borehole data includes core logging data and well logging curve data. Equipped with edge computing nodes, it enables real-time data preprocessing, including format conversion to GeoJSON / GeoTIFF, outlier removal using the 3σ criterion, filling in missing values ​​using the K-nearest neighbor algorithm, and data transmission to a distributed database via 5G / fiber optic. The ore-forming element correlation analysis module employs the Apriori association rule algorithm, which sets clear minimum support and minimum confidence standards. The algorithm mines ore-forming elements and identifies ore-controlling structures and geophysical anomalies that meet the requirements. Simultaneously, it constructs an ore-forming element correlation map with a sufficient number of nodes and automatic annotation of core elements. In terms of analysis efficiency, the correlation analysis time for specific areas is controllable, meeting the time requirements of practical work. The 3D mineralization dynamic simulation module is built using the Unity3D engine to construct a 1:10,000 scale 3D geological model with reasonable mesh accuracy settings. The module integrates three types of dynamic simulation functions: tectonic evolution, fluid transport dynamics, and mineralization enrichment. Each type of simulation has corresponding parameter range settings. During the simulation, users can adjust parameters in real time. After the simulation is completed, the module outputs the mineralization probability distribution and the coordinates of the mineral enrichment center. Not only does the resolution of the mineralization probability distribution match the model, but the error of the mineral enrichment center coordinates is also small, and the simulation results are accurately consistent with known mineral deposits. AI Target Area Prediction and Priority Assessment Module: This module uses a CNN-LSTM fusion model for target area prediction and priority assessment. The model input consists of feature maps of various elements with specific specifications, covering four categories of features: geological, geophysical, geochemical, and simulation results. The model is pre-trained based on labeled samples of known mineral deposits, covering various mineral types, and supports incremental learning. The prediction process is divided into three levels: Level 1 clarifies the target area range and mineralization probability requirements; Level 2 calculates the target area potential value and determines the value range; Level 3 prioritizes the target areas based on the potential value. The simulation prediction result visualization and interaction module supports 3D scene rendering with a rendering frame rate that meets the requirements for smooth display. It also provides five types of display functions, including spatial distribution display of ore-forming elements, playback of 3D ore-forming simulation animation, display of target area distribution heat map, comparison of borehole-simulation results, and plotting of reserve estimation value lines. Each display function has corresponding operation or display standards. It also supports data export, provides commonly used export formats, and implements distance measurement and profile extraction functions. It is compatible with PC and mobile devices and sets corresponding resolution standards for different terminals. System Iteration and Optimization and Data Update Module: An automatic update mechanism is set up. In terms of data updates, new data from the Geological Survey is synchronized monthly, and incremental updates take less time. In terms of model optimization, after a certain number of exploration projects are completed, the AI ​​model is fine-tuned based on new validation data, using specific algorithms and setting corresponding parameters. In terms of rule updates, the mineralization association rule library is updated quarterly, equipped with a user feedback interface, supporting various feedback forms, and recording the target area validation results, forming a complete closed loop of data-simulation-prediction-validation-optimization. The module for dynamic calculation of ore-forming element weights calculates the contribution weights of ore-forming elements in different geological periods through integration. The calculation formula is as follows: W i Let f be the dynamic weight of the i-th type of ore-forming element, t1 and t2 be the start and end times of the target ore-forming period, and f be the dynamic weight of the i-th type of ore-forming element. i g(t) represents the activity intensity of the i-th type of element at time t, g(t) represents the time decay coefficient of mineralization contribution, and t0 represents the peak mineralization period. By calculating and distinguishing the mineralization contribution of elements in different periods, the mineralization simulation can be made to fit the actual geological evolution. Target area mineralization potential correction module; The initial potential value is corrected based on borehole verification data, and the calculation formula is as follows: Among them, P' j P is the corrected potential value of the j-th target region. j N represents the initial potential value of the target area; α is the correction coefficient, and N... j,min The number of boreholes encountering mineralization within the target area; N j,total This represents the total number of boreholes drilled within the target area.

2. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, The mineral big data acquisition and standardization module also includes a UAV remote sensing submodule; equipped with a 6-rotor UAV, a 12-band multispectral camera and a ground laser scanner, the UAV remote sensing submodule is used to acquire fine images of mineralization and alteration and surface microstructure data. The fine images of mineralization and alteration identify silicification and carbonatization; the surface microstructure data identifies fractures. The UAV data is stitched together in real time through edge computing nodes and also undergoes radiometric correction through a diffuse reflection reference plate. Then it is fused with satellite remote sensing data, and a pixel-level fusion algorithm is used to supplement the microscopic details of the remote sensing data.

3. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, The metallogenic element correlation analysis module also includes a spatiotemporal correlation mining submodule; it introduces a time dimension to expand traditional spatial correlation, and models the tectonic, stratigraphic, and mineralization data of different geological periods, including the Caledonian and Yanshanian periods, to calculate the cross-period element correlation degree R. i ,j(t1,t2); Where t1 represents the formation time of early elements and t2 represents the mineralization time of later elements; the spatiotemporal correlation results are displayed in the form of time series graphs, and the node size represents the element intensity, which helps to identify the mineralization mode of early ore-forming material reserves and later tectonic activation, avoiding the limitation of traditional static correlation ignoring the time evolution.

4. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, The three-dimensional dynamic simulation module for mineralization also includes a fluid-rock interaction simulation submodule. It uses the PHREEQC hydrogeochemical model to calculate the reaction rate between fluid and rock. The input parameters include rock mineral composition, fluid chemical composition, temperature and pressure. During the simulation, the amount of mineral dissolution / precipitation, fluid pH changes and isotopic fractionation are calculated in real time. The output results include a comparison diagram of mineral composition before and after the reaction and curves of fluid chemical parameter changes, providing quantitative basis for the analysis of the source and migration path of mineralization fluids.

5. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, The AI ​​target area prediction and priority assessment module also includes an uncertainty analysis submodule; Monte Carlo simulation is used to assess the prediction error, with 1000 simulation iterations. Random noise is added to the input ore-forming element data, which includes the content of geochemical elements and the amplitude of geophysical anomalies. The probability of mineralization in the target area is recalculated in each iteration. The uncertainty analysis results are displayed synchronously with the target area prediction results to help users objectively judge the reliability of the target area.

6. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, The simulation prediction result visualization and interaction module also includes a virtual borehole design submodule; it supports users to design virtual boreholes in a 3D scene, with users inputting borehole diameter, dip angle, azimuth angle, and design depth; the system automatically calculates the intersection of the borehole trajectory and the mineralization simulation results, and outputs the predicted mineral type, grade, and thickness at the intersection; based on the virtual borehole data, it generates borehole columnar sections and well logging curve prediction diagrams, with the well logging curve prediction diagrams including density and resistivity curves; virtual boreholes are generated in batches, and can be exported as borehole design documents, which include coordinates, depth, and predicted mineralized sections, providing guidance for actual drilling operations.

7. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, The system iteration optimization and data update module also includes a model deviation diagnosis submodule. By comparing AI prediction results with actual exploration verification data, three types of deviation indicators are calculated: location deviation, grade deviation, and reserve deviation. Location deviation is the distance between the predicted target area center and the actual mineralization center; grade deviation is the relative error between the predicted grade and the actual grade; and reserve deviation is the relative error between the predicted reserve and the actual reserve. For prediction samples with deviations exceeding the standard, SHAP value analysis is used to identify the cause of the deviation. Elements with an absolute SHAP value greater than 0.1 are the source of deviation, and a deviation correction report is automatically generated. The report includes data supplementation suggestions and model parameter adjustment schemes.

8. The mineral resource metallogenic regularity simulation and target area prediction system based on big data according to claim 1, characterized in that, Also includes: The user-to-user collaborative analysis module supports 5-20 users collaborating online simultaneously. It is implemented using the WebSocket protocol and features three levels of permissions: administrator, analyst, and viewer. Administrators can modify system parameters and review data; analysts can perform simulations and export results; and viewers can only browse results and add annotations. The module provides collaborative tools, including real-time annotation, online discussion, and task assignment. Real-time annotation supports text and graphic annotations. File transfer is supported, and task assignment assigns target area verification tasks to designated users with set completion deadlines. The collaborative process automatically records operation logs, including user, time, and operation content, and supports log backtracking.

Images