Mineral resource prediction method based on three-dimensional geological information and knowledge graph

Abstract

This application provides a mineral resource prediction method based on three-dimensional geological information and knowledge graphs, comprising the following steps: S1, collecting geological borehole data, geophysical data, and geochemical data, and constructing a voxelized three-dimensional geological information database through spatial registration and voxel division; S2, extracting ore-controlling indicators of lithology, fault structures, and alteration zoning based on known mineralization distribution, and calculating the mineralization favorability index of spatial units; S3, constructing a geological knowledge graph representing the relationship between deposit type, ore-controlling structures, metallogenic epoch, and mineralization markers, and weighting the mineralization favorability index based on knowledge rules; S4, classifying the mineralization favorability index according to thresholds, identifying continuous mineralization favorable areas based on three-dimensional connected bodies, then screening target areas, and outputting the spatial location and extent of the target areas. This application can improve the accuracy and efficiency of mineral resource prediction, reduce human intervention, and thus provide reliable exploration target area data.

Description

Technical Field

[0001] This application belongs to the field of mineral resource exploration and prediction technology, specifically involving a mineral resource prediction method based on three-dimensional geological information and knowledge graph. Background Technology

[0002] With the increasing global demand for mineral resources, the exploration and prediction of these resources have become increasingly important. Mineral resource exploration not only provides fundamental support for national economic development but also safeguards the development of industries such as energy and manufacturing. Traditional mineral resource prediction methods largely rely on techniques such as geological drilling, geophysical exploration, and geochemical analysis. While these methods have achieved certain results in resource prediction in localized areas, they face problems such as high cost, low efficiency, and insufficient accuracy in large-scale mineral resource assessments. Especially in regions with complex and diverse resource distribution, traditional methods struggle to provide high-precision and high-efficiency mineral resource predictions. Therefore, how to utilize modern technologies to improve the accuracy and efficiency of mineral resource prediction has become a critical issue that urgently needs to be addressed in the field of mineral resource exploration.

[0003] Existing mineral resource prediction methods primarily rely on geological borehole data, geophysical data, and geochemical data, employing a series of statistical methods, interpolation algorithms, or regression analyses for resource assessment. Among these methods, three-dimensional geological modeling is commonly used to construct spatial distribution models of mineral deposits. By analyzing the geological characteristics, physical properties, and chemical composition of the deposits, the potential distribution of mineral resources is obtained. However, existing technologies suffer from several problems. First, multi-source data fusion is difficult. Different data sources (such as borehole, geophysical, and geochemical data) differ in spatial scale, data format, and dimensions, potentially leading to errors after fusion. Second, traditional metallogenic favorability analyses largely depend on expert experience or statistical analysis, making the results somewhat subjective and difficult to apply over large areas.

[0004] Furthermore, although some methods attempt to combine knowledge graph technology to improve prediction accuracy, most existing knowledge graphs are limited to describing surface features such as deposit type and mineralization indicators, lacking in-depth analysis of complex factors such as deposit genesis and tectonic control. This prevents existing knowledge graphs from being effectively integrated with three-dimensional geological models, limiting their application in practical mineral resource prediction.

[0005] Therefore, the main problems faced by existing technologies are the difficulty in fusing multi-source geological data and the low accuracy of mineralization favorability analysis. Thus, it is necessary to improve existing technologies and provide a mineral resource prediction method that can enhance the accuracy of mineral resource prediction by effectively fusing multi-source geological data and incorporating knowledge graphs. Summary of the Invention

[0006] This application provides a mineral resource prediction method based on three-dimensional geological information and knowledge graphs to solve the problems of difficulty in fusion of multi-source geological data and low accuracy of mineralization favorability analysis in the existing technology, especially in mineral resource exploration over a large area, thereby improving the accuracy and reliability of mineral resource prediction.

[0007] This application provides a mineral resource prediction method based on three-dimensional geological information and knowledge graph, including the following steps:

[0008] S1. Collect geological borehole data, geophysical data, and geochemical data, and construct a voxelized three-dimensional geological information database through spatial registration and voxel division;

[0009] S2. Based on the known distribution of mineralized bodies, extract the ore-controlling indicators of lithology, fault structure and alteration zoning, and calculate the mineralization favorability index of spatial units;

[0010] S3. Construct a geological knowledge map representing the relationship between deposit type, ore-controlling structures, metallogenic epoch and mineralization markers, and adjust the weight of the metallogenic favorability index based on knowledge rules;

[0011] S4. Classify mineralization favorable index according to threshold, identify continuous mineralization favorable areas based on three-dimensional connected bodies, then screen target areas, and output the spatial location and range of the target areas.

[0012] As an optional approach in this application, before constructing the voxelized 3D geological information database, the collected geological borehole data, geophysical data, and geochemical data are preprocessed, including: integrity detection and anomaly removal of the original data to remove errors and duplicate records; completion of missing values ​​using linear interpolation or neighborhood averaging; standardization of data from different sources according to their dimensional differences, wherein borehole data are converted to a unified interval using interval scaling, and geophysical and geochemical data are standardized using Z-score to achieve statistical equilibrium; and local noise is removed using data filtering or smoothing methods to obtain a cleaned geological dataset that can be used for subsequent spatial registration and voxel division.

[0013] As an optional approach to this application, the process of constructing a voxelized three-dimensional geological information database through spatial registration and voxel partitioning includes the following steps:

[0014] (1) Spatial registration:

[0015] The cleaned geological borehole data, geophysical data, and geochemical data were unified into the same spatial coordinate system, and the coordinate transformation matrix was calculated using the least squares method. T And perform translation and rotation registration on various data point sets, the coordinate transformation matrix T's The calculation formula is:

[0016] ;

[0017] in, The coordinates of the source data points. For the target coordinates, For rotation matrix, It is a translation vector;

[0018] (2) Voxel division and attribute assignment:

[0019] Based on the boundary of the study area, and according to the set voxel side length The study area was divided into a regular cubic grid, resulting in multiple voxel units. To form a three-dimensional voxel set The volume of a single voxel ;

[0020] For each voxel unit, interpolation or weighted averaging is performed based on the spatial projection of borehole penetration depth, geophysical anomalies, and geochemical indices to obtain the voxel attribute vector:

[0021] ;

[0022] Each component This represents the value of the nth geological attribute.

[0023] (3) Establishment of adjacency relationships:

[0024] Using the 6-adjacency principle of shared surfaces, adjacency relationships between voxels are defined, and an adjacency matrix is ​​established. When voxels With voxels When sharing a face, let A(i,j)=1; otherwise, A(i,j)=0. The adjacency relationship between voxels is represented by an adjacency matrix to achieve fast retrieval of spatial data.

[0025] (4) Database generation:

[0026] The data structure containing spatial coordinates, attribute vectors, and adjacency relationships is stored in a three-dimensional geological information database for subsequent calculation of mineralization favorability index, identification of connected components, and prediction and analysis of mineral resources.

[0027] As an alternative approach in this application, the mineralization favorability index of spatial units The evidence weight method based on probability logarithmic weights is used for calculation, and the results are calculated according to voxel units. Implemented unit by unit, specifically including:

[0028] Let the set of voxels corresponding to the known mineralization be denoted as the positive sample set. S1. The set of voxels corresponding to non-mineralized bodies is denoted as the negative sample set. S 2; The probability of each mineral control index k appearing in the two sets is denoted as P. 1,k With P 2,k The evidentiary value for calculating this mineral control index:

[0029] ;

[0030] voxels The fracture geometric center distance is denoted as The alteration coverage ratio is denoted as The lithological mineralization indicator value is denoted as The above-mentioned ore-controlling indicators are then weighted and summed to obtain the mineralization favorability index. :

[0031] ;

[0032] in For bias terms, , , ∈[0,1], and , >0 represents the distance attenuation coefficient; when ≥ At that time, voxels are identified as candidate units for mineralization-favorable voxels, and the threshold is set. This is a preset value.

[0033] As an optional approach to this application, the construction of the geological knowledge map includes the following steps:

[0034] Entity and relationship descriptions related to deposit type, ore-controlling structures, metallogenic epoch and mineralization indicators were extracted from geological literature, geological reports and exploration databases;

[0035] The extracted text undergoes named entity recognition and relation extraction, mapping geological concepts to triples. h , r , e (), where h represents the initial geological entity, r represents the relationship type between entities, and e represents the associated geological entity;

[0036] The extracted triples are subjected to semantic disambiguation, attribute completion, and consistency verification to form a semantically normalized set of entity relations.

[0037] The entity relationship set is stored in a graph database, and the association strength and confidence between entities are calculated based on a knowledge reasoning algorithm to generate a complete geological knowledge graph, which is used to support subsequent mineral resource prediction and metallogenic regularity analysis.

[0038] As an optional approach in this application, the weight correction based on knowledge rules includes the following steps:

[0039] (1) Rule setting: Based on the relationship types between geological entities in the geological knowledge map, a set of weight correction rules is established. The relationship types include “genetic control”, “symbiotic enrichment”, “temporal dependence” and “spatial proximity”.

[0040] (2) Weight correction calculation: For each voxel unit Based on the knowledge association strength matched in the knowledge rules With rule weight coefficient Calculate the correction amount The formula for calculating the correction amount is:

[0041] ;

[0042] Where m represents the number of knowledge rules associated with the voxel;

[0043] (3) Index update: based on the correction amount The revised mineralization favorability index is obtained by updating according to the following formula. :

[0044] ;

[0045] in, This represents the knowledge correction coefficient, which ranges from [0, 1] and is used to control the degree of influence of knowledge constraints on the prediction results.

[0046] As an optional approach to this application, the identification of continuous mineralization favorable areas based on three-dimensional connected volumes includes:

[0047] Mineralization Favorability Index after Voxel Unit Correction By threshold Grading yields a set of voxels favorable for mineralization. ;

[0048] Based on the adjacency matrix Starting with each favorable voxel for mineralization, the system recursively searches for spatially connected voxel units according to their adjacency relationships, merging consecutively connected voxels into a connected unit to form a set of connected units. ;

[0049] For each connected component Statistical voxel count According to voxel side length Calculate the volume of a connected solid ,when At that time, the connected body was identified as a favorable area for continuous mineralization; among which, Minimum volume threshold;

[0050] Output the geometric center and three-dimensional boundary range of the continuous mineralization favorable area for subsequent target area screening.

[0051] As an optional approach in this application, the process of screening the target region includes:

[0052] From all the identified continuous mineralization favorable areas, select those that meet the preset minimum area based on their area. Required target area;

[0053] Output the spatial location, extent, geometric center coordinates, and area information of the filtered target area for subsequent exploration deployment and visualization.

[0054] Compared with the prior art, this application has the following beneficial effects:

[0055] 1. This application provides a mineral resource prediction method based on three-dimensional geological information and knowledge graphs. Through spatial registration and voxel partitioning, collected geological borehole data, geophysical data, and geochemical data are unified to the same coordinate system. Furthermore, voxelization processing divides the study area into regular voxel units, with accurate data assignment and integration within each unit. This process optimizes the integration and registration of multi-source data, improves data consistency and usability, and provides a reliable three-dimensional geological information database, laying a solid foundation for subsequent mineral resource prediction and mineralization favorability analysis.

[0056] 2. This application calculates the mineralization favorability index of spatial units by extracting ore-controlling indicators such as lithology, fault structure, and alteration zoning, and combining them with the known distribution of mineralized bodies. These ore-controlling indicators provide key geological information for mineral resource prediction. By quantifying this information, a numerical index reflecting the mineralization potential of the target area is provided, thereby improving the objectivity and accuracy of mineralization favorability analysis. This method replaces traditional empirical judgment with quantitative analysis based on geological characteristics, making mineral resource prediction results more scientific and reliable, and reducing the influence of subjective factors on the analysis results.

[0057] 3. This application constructs a geological knowledge graph to systematically represent the relationships between deposit types, ore-controlling structures, metallogenic epochs, and mineralization indicators, providing a more comprehensive geological background for metallogenic aptitude analysis. This method enables a more accurate assessment of the metallogenic potential of each spatial unit. Based on the knowledge graph construction, and combined with knowledge rule correction, the metallogenic aptitude index was further optimized, improving the objectivity and accuracy of the analysis. In the target area screening process, this application first automatically identifies all continuous metallogenic aptitude areas using three-dimensional connected body recognition technology. Then, by setting a minimum area threshold, the identified metallogenic aptitude areas are screened, excluding areas that are too small and lack practical exploration value. This approach avoids human intervention and subjective judgment, improves the accuracy and automation of target area screening, and provides precise data support for subsequent exploration deployment and resource optimization, thereby further improving the efficiency and scientific nature of mineral resource exploration. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 A flowchart of a mineral resource prediction method based on three-dimensional geological information and knowledge graph provided in an embodiment of this application. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application are described clearly and completely below. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are also within the scope of protection of this application.

[0061] like Figure 1 As shown in the figure, this application provides a mineral resource prediction method based on three-dimensional geological information and knowledge graph, including the following steps:

[0062] S1. Collect geological borehole data, geophysical data, and geochemical data, and construct a voxelized three-dimensional geological information database through spatial registration and voxel division;

[0063] S2. Based on the known distribution of mineralized bodies, extract the ore-controlling indicators of lithology, fault structure and alteration zoning, and calculate the mineralization favorability index of spatial units;

[0064] S3. Construct a geological knowledge map representing the relationship between deposit type, ore-controlling structures, metallogenic epoch and mineralization markers, and adjust the weight of the metallogenic favorability index based on knowledge rules;

[0065] S4. Classify mineralization favorable index according to threshold, identify continuous mineralization favorable areas based on three-dimensional connected bodies, then screen target areas, and output the spatial location and range of the target areas.

[0066] This embodiment unifies the collected geological borehole data, geophysical data, and geochemical data into the same coordinate system through spatial registration and voxel division methods. Voxelization processing divides the study area into regular voxel units, with accurate data assignment and integration within each unit. This process optimizes the integration and registration of multi-source data, improves data consistency and usability, and provides a reliable three-dimensional geological information database, laying a solid foundation for subsequent mineral resource prediction and metallogenic aptitude analysis.

[0067] Meanwhile, in mineral resource prediction, the accuracy of mineralization favorability analysis directly affects the reliability of the prediction results. Traditional methods often rely on expert experience and subjective judgment, which can easily lead to biased prediction results and lack sufficient objectivity. This embodiment extracts ore-controlling indicators such as lithology, fault structure, and alteration zoning, and combines them with known mineralization distribution to calculate the mineralization favorability index of spatial units. These ore-controlling indicators provide key geological information in mineral resource prediction. By quantifying this information, a numerical index reflecting the mineralization potential of the target area is provided, thereby improving the objectivity and accuracy of mineralization favorability analysis. Through this method, traditional experience-based judgment is replaced by quantitative analysis based on geological characteristics, making mineral resource prediction results more scientific and reliable, and reducing the influence of subjective factors on the analysis results.

[0068] Furthermore, this embodiment systematically represents the relationship between deposit types, ore-controlling structures, metallogenic epochs, and mineralization indicators by constructing a geological knowledge graph, providing a more comprehensive geological background for metallogenic aptitude analysis. This method enables a more accurate assessment of the metallogenic potential of each spatial unit. Based on the knowledge graph, the metallogenic aptitude index is further optimized by combining knowledge rule correction, improving the objectivity and accuracy of the analysis. This correction process eliminates the subjective factors that rely on expert experience in traditional methods, providing a more scientific analytical framework, thus making the metallogenic aptitude analysis more precise and reliable. In the target area screening process, this embodiment first automatically identifies all continuous metallogenic aptitude areas using three-dimensional connected body recognition technology. Then, by setting a minimum area threshold, the identified metallogenic aptitude areas are screened, excluding areas that are too small and lack practical exploration value. This approach avoids human intervention and subjective judgment, improves the accuracy and automation of target area screening, and provides precise data support for subsequent exploration deployment and resource optimization, thereby further improving the efficiency and scientific nature of mineral resource exploration.

[0069] In some embodiments, before constructing a voxelized 3D geological information database, the collected geological borehole data, geophysical data, and geochemical data are preprocessed, including: integrity detection and anomaly removal of the original data to remove errors and duplicate records; completion of missing values ​​using linear interpolation or neighborhood averaging; standardization of data from different sources according to their dimensional differences, wherein borehole data is converted to a unified interval using interval scaling, and geophysical and geochemical data are standardized using Z-score to achieve statistical equilibrium; and local noise is removed using data filtering or smoothing methods to obtain a cleaned geological dataset that can be used for subsequent spatial registration and voxel division.

[0070] In mineral resource prediction, data quality directly impacts the accuracy and reliability of the final analysis results. To ensure data accuracy, this embodiment preprocesses the collected geological borehole data, geophysical data, and geochemical data. First, integrity checks and outlier removal steps eliminate erroneous data and duplicate records, preventing the impact of data inconsistencies or redundant information on the analysis results. For missing data points, linear interpolation and neighborhood averaging are used to fill in the gaps, inferring the missing parts using neighboring data, thus reducing the interference of missing data on subsequent analysis. Through these steps, the preprocessed dataset has higher integrity and accuracy, providing a reliable data foundation for subsequent spatial registration and voxel division, thereby improving the quality of mineral resource prediction.

[0071] Furthermore, in mineral resource forecasting, data from different measurement devices and technologies often exhibit differences in dimensions, resolution, and statistical distribution, which can easily lead to inconsistencies during data fusion. To address this issue, this embodiment employs standardization processing to enable effective fusion of different data sources within a unified framework.

[0072] Specifically, borehole data was scaled using an interval scaling method to transform it into a uniform interval range, thereby eliminating dimensional differences. Geophysical and geochemical data were processed using Z-score standardization to ensure that the data remained statistically balanced. This standardization method eliminated differences between different data sources, enabling various types of data to be compared and analyzed under the same standard, further improving the accuracy of data fusion.

[0073] Furthermore, data filtering or smoothing methods are used to remove local noise, improving data reliability and providing a more accurate basis for subsequent analysis. This process allows the cleaned geological dataset to be efficiently used for subsequent spatial registration and voxel division, further enhancing the quality and reliability of mineral resource prediction.

[0074] In practice, geological borehole data typically originates from underground drilling operations, recording lithological information at different depths; geophysical data comes from ground or aerial remote sensing measurements, involving responses to different physical fields, such as electromagnetic and gravitational fields; and geochemical data is obtained through sample analysis, reflecting the distribution of elements within geological bodies. To ensure the reliability of subsequent data analysis, all data must undergo rigorous preprocessing steps. First, the raw data undergoes integrity checks, using automated algorithms to identify missing or duplicate records and remove erroneous data. For missing values, the neighborhood averaging method is used to impute local missing values. This method fills in missing values ​​based on the mean of adjacent data points, thus avoiding errors that may be introduced by simple linear interpolation. For data from different sources, interval scaling is used to transform the borehole data to a uniform numerical range, and Z-score standardization is applied to geophysical and geochemical data to ensure that the data conforms to a standard normal distribution. This helps eliminate dimensional differences and allows various data types to participate in subsequent analysis in a balanced manner. Furthermore, median filtering is used to remove local noise, further improving the smoothness and stability of the data, making the cleaned dataset better suited for subsequent spatial registration and voxel partitioning.

[0075] Furthermore, in step S1, the process of constructing a voxelized three-dimensional geological information database through spatial registration and voxel division includes the following steps:

[0076] (1) Spatial registration:

[0077] The cleaned geological borehole data, geophysical data, and geochemical data were unified into the same spatial coordinate system, and the coordinate transformation matrix was calculated using the least squares method. T And perform translation and rotation registration on various data point sets, this coordinate transformation matrix T's The calculation formula is:

[0078] ;

[0079] in, The coordinates of the source data points. For the target coordinates, For rotation matrix, It is a translation vector.

[0080] (2) Voxel division and attribute assignment:

[0081] Based on the boundary of the study area, and according to the set voxel side length The study area was divided into a regular cubic grid, resulting in multiple voxel units. To form a three-dimensional voxel set The volume of a single voxel ;

[0082] For each voxel unit, interpolation or weighted averaging is performed based on the spatial projection of borehole penetration depth, geophysical anomalies, and geochemical indices to obtain the voxel attribute vector:

[0083] ;

[0084] Each component This represents the value of the nth geological attribute.

[0085] In this step, attribute values ​​are assigned to each voxel unit, and the components in the attribute vector ( , ,… These represent different types of geological attributes, reflecting multiple geological factors associated with the mineralization process. Specifically, , ,… These correspond to different geological quality attributes, including but not limited to lithology, fault structures, alteration characteristics, geophysical response values ​​(such as resistivity, gravity anomalies, etc.), and geochemical element concentrations. These attributes play a crucial role in mineral resource prediction; through their comprehensive analysis, mineralization potential and mineral distribution can be inferred.

[0086] (3) Establishment of adjacency relationships:

[0087] Using the 6-adjacency principle of shared surfaces, adjacency relationships between voxels are defined, and an adjacency matrix is ​​established. When voxels With voxels When sharing a face, let A(i,j)=1; otherwise, let A(i,j)=0. The adjacency relationship between voxels is represented by the adjacency matrix, which enables fast retrieval of spatial data.

[0088] (4) Database generation:

[0089] The data structure containing spatial coordinates, attribute vectors, and adjacency relationships is stored in a three-dimensional geological information database for subsequent calculation of mineralization favorability index, identification of connected components, and prediction and analysis of mineral resources.

[0090] In the above embodiment, the calculation of the coordinate transformation matrix T is optimized using the least squares method, aiming to minimize the coordinate difference between the source data points and the target data points. In the formula, Represents the coordinates of the source data points. Indicates the target coordinates. For rotation matrix, Let be the translation vector. The minimization objective is:

[0091] ;

[0092] This formula calculates the rotation matrix and translation vector to ensure that the spatial positions of the source and target data point sets are as consistent as possible. The least squares method optimizes this formula to obtain the transformation matrix T, ensuring that different data sources (borehole, geophysical, and geochemical data) can be unified into the same spatial coordinate system, thereby guaranteeing the accuracy and consistency of data fusion.

[0093] The 6-adjacency rule for shared faces is used to define the adjacency relationships between voxels. According to this rule, two voxels are adjacent voxels if they share a face. In this case, the adjacency matrix... A value of 1 is assigned; if two voxels do not share a face, a value of 0 is assigned. The 6-adjacency principle is a commonly used spatial adjacency definition method, applicable to adjacent voxels in a regular cubic mesh. This method clearly represents the relationships between voxels, providing foundational data for subsequent connected component identification. This definition is not only simple and efficient but also accurately describes the connectivity of voxel units in space, facilitating the rapid identification and analysis of the continuity of favorable mineralization areas.

[0094] This embodiment first uses the least squares method to calculate the coordinate transformation matrix T, transforming all data sources into a unified spatial coordinate system and eliminating the impact of spatial coordinate differences on data fusion. Next, voxel partitioning technology is used to divide the study area into a uniform cubic grid, with each voxel unit having the same spatial scale and attribute data, thus solving the problem of spatial consistency between different data sources. This method improves the accuracy of data fusion, providing a unified and reliable data foundation for subsequent mineral resource prediction and mineralization favorability analysis, ensuring the accuracy of the analysis results.

[0095] Building upon this foundation, voxel partitioning divides the study area into regular voxel units, ensuring that each voxel has the same spatial scale, facilitating systematic management and processing of data from different spatial locations. By assigning attribute values ​​to each voxel unit and performing interpolation or weighted averaging based on the spatial projections of borehole data, geophysical anomalies, and geochemical indicators, the attribute vector for each voxel is obtained. This method not only improves the spatial accuracy of the data but also integrates various geological information, providing complete and structured spatial data for subsequent resource prediction and mineralization favorability analysis.

[0096] This spatial registration and voxel division method effectively integrates all geological data within a unified three-dimensional spatial framework, resolving the issue of spatial inconsistency. Simultaneously, voxelization not only preserves the accuracy of spatial relationships but also facilitates subsequent spatial analysis and calculations, such as the calculation of mineralization favorability indices and the identification of interconnected components. Ultimately, this approach provides an efficient and reliable data foundation for the generation of a three-dimensional geological database, ensuring the accuracy and systematic nature of subsequent analyses.

[0097] In step S2, the mineralization favorability index of the spatial unit The evidence weight method based on probability logarithmic weights is used for calculation, and the results are calculated according to voxel units. Implemented unit by unit, specifically including:

[0098] Let the set of voxels corresponding to the known mineralization be denoted as the positive sample set. S 1. The set of voxels corresponding to non-mineralized bodies is denoted as the negative sample set. S 2; The probability of each mineral control index k appearing in the two sets is denoted as P. 1,k With P 2,k The evidentiary value for calculating this mineral control index:

[0099] ;

[0100] voxels The fracture geometric center distance is denoted as The alteration coverage ratio is denoted as The lithological mineralization indicator value is denoted as The above-mentioned ore-controlling indicators are then weighted and summed to obtain the mineralization favorability index. :

[0101] ;

[0102] in For bias terms; , , These are weighting coefficients used to adjust the relative influence of various geological factors in the calculation of the mineralization favorability index. These are the weighting coefficients for lithological factors. These are the weighting coefficients of alteration characteristic factors. These are the weighting coefficients of fracture structural factors. , , ∈[0,1], and. This indicates the contribution of lithological factors to the favorable conditions for mineralization; This indicates the contribution of alteration characteristics to mineralization. This indicates the contribution of fracture structures to mineralization. >0 represents the distance attenuation coefficient;

[0103] when ≥ At that time, voxels are identified as candidate units for mineralization-favorable voxels, and the threshold is set. This is a preset value. Threshold. It is an important parameter used to screen for favorable areas of continuous mineralization, and represents the lowest value of the mineralization favorableness index.

[0104] The logarithmic weighted evidence weight method is used in this embodiment to calculate the mineralization favorability index because it can more effectively handle and quantify the uncertainties in geological information and comprehensively consider the influence of multiple ore-controlling indicators. The logarithmic weighted evidence weight method quantifies the difference in the performance of each ore-controlling factor in known mineralized bodies (positive samples) and non-mineralized bodies (negative samples), and calculates the evidence weight of each ore-controlling factor. By comparing the occurrence probability of each ore-controlling factor in positive and negative samples, this method can accurately reflect the differences between geological factors inside and outside mineralized bodies, thus providing a more scientific and data-driven basis for the calculation of the mineralization favorability index. Specifically, the evidence weight formula calculates the logarithmic ratio of the occurrence probability of ore-controlling indicators, considering both the positive effects of ore-controlling factors in mineralized bodies and their inhibitory effects in non-mineralized bodies. This makes the calculated results of the mineralization favorability index more consistent with actual geological laws.

[0105] Furthermore, this embodiment calculates the mineralization favorability index using the above-described method, primarily based on the comprehensive influence mechanism of geological factors on mineralization. Lithology, alteration, and faulting are key elements controlling mineral distribution, but their modes of action and scope of influence differ, thus requiring a weighted comprehensive evaluation. In the formula for calculating the mineralization favorability index, the lithological mineralization indicator value... This reflects the influence of stratigraphic material properties on mineralization, through functions. Mapping to quantitative indicators makes the contributions of different lithological types to mineralization comparability; alteration cover ratio The influence of hydrothermal alteration on mineralization was described by a function. Adding weighted superposition after transformation helps to reflect the spatial distribution characteristics of alteration; fracture geometric center distance It reflects the spatial distance relationship with the tectonic fracture zone and uses an exponential decay function. The adjustments are made so that voxels closer to the fault contribute more to mineralization, thus better aligning with the spatial patterns of mineralization. (Weight coefficients for each item are listed below.) , , Determined based on geostatistical results, this weighted superposition method balances the influence of different ore-controlling factors. By comprehensively considering the influence characteristics of multi-source geological information at different scales, the calculation of the mineralization favorability index possesses both physical meaning and quantifiable reflection of the intensity of each factor's effect, thereby improving the scientific rigor and reliability of the prediction results. Ultimately, the mineralization favorability index value exceeds the set minimum threshold. At that time, voxels were identified as promising candidate areas for mineralization, which promoted the scientific and automated development of mineral resource prediction and exploration.

[0106] In step S3, the construction of the geological knowledge map includes the following steps:

[0107] Entity and relationship descriptions related to deposit type, ore-controlling structures, metallogenic epoch and mineralization indicators were extracted from geological literature, geological reports and exploration databases;

[0108] The extracted text undergoes named entity recognition and relation extraction, mapping geological concepts to triples. h , r , e (), where h represents the initial geological entity, r represents the relationship type between entities, and e represents the associated geological entity;

[0109] The extracted triples are subjected to semantic disambiguation, attribute completion, and consistency verification to form a semantically normalized set of entity relations.

[0110] The entity relationship set is stored in a graph database, and the association strength and confidence between entities are calculated based on a knowledge reasoning algorithm to generate a complete geological knowledge graph, which is used to support subsequent mineral resource prediction and metallogenic regularity analysis.

[0111] In this embodiment, the geological knowledge graph constructed using the above method can efficiently integrate and correlate geological data, forming a structured and operable knowledge base. First, by extracting entity and relation descriptions from geological literature, geological reports, and exploration databases, rich geological knowledge can be gathered, covering multiple aspects such as deposit types, ore-controlling structures, metallogenic epochs, and mineralization indicators. This information is of great significance in geological exploration. Through named entity recognition and relation extraction, the information in the original text is transformed into structured triples, allowing various geological concepts and their interrelationships to be clearly and accurately expressed. In this way, tacit knowledge in the geological field is revealed, providing a solid theoretical foundation for mineral resource prediction.

[0112] Secondly, the semantic disambiguation, attribute completion, and consistency verification steps of triples can resolve ambiguity and inconsistency in natural language, ensuring the accuracy and standardization of the knowledge graph. Finally, based on knowledge reasoning algorithms, the association strength and confidence between entities are calculated, providing quantitative support for the connections between geological concepts.

[0113] Furthermore, in step S3, the weight correction based on knowledge rules includes the following steps:

[0114] (1) Rule setting: Based on the relationship types between geological entities in the geological knowledge map, a set of weight correction rules is established. The relationship types include “genetic control”, “symbiotic enrichment”, “temporal dependence” and “spatial proximity”.

[0115] (2) Weight correction calculation: For each voxel unit Based on the knowledge association strength matched in the knowledge rules With rule weight coefficient Calculate the correction amount The formula for calculating this correction amount is:

[0116] ;

[0117] Where m represents the number of knowledge rules associated with the voxel;

[0118] (3) Index update: based on the correction amount The revised mineralization favorability index is obtained by updating according to the following formula. :

[0119] ;

[0120] in, This represents the knowledge correction coefficient, which ranges from [0, 1] and is used to control the degree of influence of knowledge constraints on the prediction results.

[0121] In this embodiment, the above-described method of correcting the mineralization favorability index based on knowledge rules is used to effectively utilize the geological laws and expert experience reflected in the geological knowledge graph, thereby improving the prediction accuracy of the mineralization favorability index. The geological knowledge graph contains relationships between important geological factors such as deposit types, ore-controlling structures, and metallogenic epochs. However, these relationships are often complex and multidimensional. Correcting the mineralization favorability index using knowledge rules can further enhance the accuracy of predictions. In particular, the relationships of "genetic control," "symbiotic enrichment," "temporal dependence," and "spatial proximity" are all important factors influencing mineral distribution. Using these knowledge rules to correct the original mineralization favorability index makes the analysis results more consistent with actual geological laws, thereby improving the reliability of predictions.

[0122] Specifically, the formula for calculating the correction amount This indicates that by matching the types of relationships between geological entities, combined with the weight coefficient of each knowledge rule... and correlation strength The formula is used to calculate the correction amount for each voxel unit. This formula integrates the influence of multiple factors, ensuring that the correction process reflects the contribution of different geological features to the mineralization favorability. The magnitude of the correction amount depends on the number of knowledge rules m associated with the voxel and the degree of influence of each rule. This allows for dynamic adjustment of the mineralization favorability of each voxel unit, making the exponential correction more closely aligned with the actual geological features.

[0123] Next, the revised mineralization favorability index is updated through the formula. Calculations are performed, in which, It is a coefficient that controls the degree of influence of knowledge revision. By adjusting... The value of the ore-forming favorability index allows for flexible control over the impact of knowledge rules on the final prediction results. This ensures the rationality of the knowledge rules while avoiding overfitting caused by excessive reliance on them. In this way, the ore-forming favorability index can be scientifically corrected based on the original data, making the prediction results more accurate and consistent with actual geological conditions.

[0124] As can be seen from the above, this embodiment, through a knowledge-rule-based weight correction method combined with the calculation of the mineralization favorability index, effectively integrates geological knowledge and data-driven analysis methods, improving the scientific rigor and reliability of mineral resource prediction. The advantage of this method lies in its ability to flexibly adjust the degree of influence of knowledge constraints while avoiding subjective biases caused by traditional experience-based judgments, further enhancing the automation and accuracy of mineral resource prediction.

[0125] In step S4, identifying continuous mineralization favorable areas based on three-dimensional connected volumes includes:

[0126] Mineralization Favorability Index after Voxel Unit Correction By threshold Grading yields a set of voxels favorable for mineralization. ;

[0127] Based on the adjacency matrix Starting with each favorable voxel for mineralization, the system recursively searches for spatially connected voxel units according to their adjacency relationships, merging consecutively connected voxels into a connected unit to form a set of connected units. ;

[0128] For each connected component Statistical voxel count According to voxel side length Calculate the volume of a connected solid ,when At that time, the connected body was identified as a favorable area for continuous mineralization; among which, Minimum volume threshold;

[0129] Output the geometric center and three-dimensional boundary range of the continuous mineralization favorable area for subsequent target area screening.

[0130] In mineral resource prediction, accurately identifying continuous mineralization-favorable areas is crucial for further resource assessment and exploration. This embodiment employs a three-dimensional connected volume identification method to efficiently identify areas with high mineralization potential, avoiding the manual judgment and inefficient screening process that may occur in traditional methods. First, the modified mineralization favorability index is classified using threshold grading to identify which voxel units have high mineralization potential. In this way, all voxel units with mineralization potential are automatically identified as mineralization-favorable voxels, avoiding manual inspection of each voxel one by one and improving analysis efficiency.

[0131] Next, the connected component clustering identification method based on the adjacency relation matrix can identify spatially connected mineralization-favorable voxels and merge these connected voxels into a single connected component. This process efficiently aggregates spatially distributed mineralization-favorable voxels into a complete mineralization-favorable area, thereby achieving holistic identification of mineralization potential areas rather than isolated voxel analysis. This method can handle complex spatial relationships, ensuring that the identification of mineralization-favorable areas conforms to actual geological characteristics.

[0132] Finally, based on the calculation and screening of connected volumes, connected volumes meeting the minimum volume threshold are defined as continuous mineralization favorable areas. This screening process automatically determines whether to include each connected volume in the target area, avoiding bias from subjective judgment and ensuring that the selected areas have actual exploration value. Through this automated, spatial connectivity-based identification method, this embodiment can more accurately determine the potential distribution areas of mineral resources, improving the efficiency and scientific rigor of exploration work, while providing a reliable basis for subsequent target area screening and visualization.

[0133] Furthermore, in step S4, the process of screening the target area includes:

[0134] From all the identified continuous mineralization favorable areas, select those that meet the preset minimum area based on their area. Required target area;

[0135] Output the spatial location, extent, geometric center coordinates, and area information of the filtered target area for subsequent exploration deployment and visualization.

[0136] This embodiment employs a region-area-based screening method when selecting target areas, primarily to effectively eliminate mineralized areas that are too small and lack practical exploration value. This embodiment sets a minimum area threshold (…). This method retains only mineralized areas that meet the area requirements, thus avoiding excessive interference from small, localized areas and focusing on target areas with actual exploration potential. The advantages of this method are that it simplifies the target area selection process, improves selection efficiency, and ensures that the selected target areas have a certain scale to meet actual exploration needs. This screening method can quickly identify eligible target areas, providing clear data support for subsequent exploration deployment and visualization, further enhancing the scientific rigor and accuracy of resource exploration.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A mineral resource prediction method based on three-dimensional geological information and knowledge graph, characterized in that, Includes the following steps: S1. Collect geological borehole data, geophysical data, and geochemical data, and construct a voxelized three-dimensional geological information database through spatial registration and voxel division; S2. Based on the known distribution of mineralized bodies, extract the ore-controlling indicators of lithology, fault structure and alteration zoning, and calculate the mineralization favorability index of spatial units; S3. Construct a geological knowledge map representing the relationship between deposit type, ore-controlling structures, metallogenic epoch and mineralization markers, and adjust the weight of the metallogenic favorability index based on knowledge rules; S4. Classify mineralization favorable index according to threshold, identify continuous mineralization favorable areas based on three-dimensional connected body, then screen target areas, and output the spatial location and range of the target areas. The process of constructing a voxelized three-dimensional geological information database through spatial registration and voxel division includes the following steps: (1) Spatial registration: The cleaned geological borehole data, geophysical data, and geochemical data were unified into the same spatial coordinate system, and the coordinate transformation matrix was calculated using the least squares method. T And perform translation and rotation registration on various data point sets, the coordinate transformation matrix T's The calculation formula is: ； in, The coordinates of the source data points. For the target coordinates, For rotation matrix, It is a translation vector; (2) Voxel division and attribute assignment: Based on the boundary of the study area, and according to the set voxel side length The study area was divided into a regular cubic grid, resulting in multiple voxel units. To form a three-dimensional voxel set The volume of a single voxel ; For each voxel unit, interpolation or weighted averaging is performed based on the spatial projection of borehole penetration depth, geophysical anomalies, and geochemical indices to obtain the voxel attribute vector: ； Each component This represents the value of the nth geological attribute. (3) Establishment of adjacency relationships: Using the 6-adjacency principle of shared surfaces, adjacency relationships between voxels are defined, and an adjacency matrix is ​​established. When voxels With voxels When sharing a face, let A(i,j)=1; otherwise, A(i,j)=0. The adjacency relationship between voxels is represented by an adjacency matrix to achieve fast retrieval of spatial data. (4) Database generation: The data structure containing spatial coordinates, attribute vectors, and adjacency relationships is stored in the three-dimensional geological information database for subsequent calculation of mineralization favorability index, identification of connected components, and mineral resource prediction and analysis. The method for identifying continuous mineralization favorable areas based on three-dimensional connected volumes includes: Mineralization Favorability Index after Voxel Unit Correction By threshold Grading yields a set of voxels favorable for mineralization. ; Based on the adjacency matrix Starting with each favorable voxel for mineralization, the system recursively searches for spatially connected voxel units according to their adjacency relationships, merging consecutively connected voxels into a connected unit to form a set of connected units. ; For each connected component Statistical voxel count According to voxel side length Calculate the volume of a connected solid ,when At that time, the connected body was identified as a favorable area for continuous mineralization; among which, Minimum volume threshold; Output the geometric center and three-dimensional boundary range of the continuous mineralization favorable area for subsequent target area screening.

2. The mineral resource prediction method based on three-dimensional geological information and knowledge graph as described in claim 1, characterized in that, Before constructing the voxelized 3D geological information database, the collected geological borehole data, geophysical data, and geochemical data were preprocessed, including: integrity checks and anomaly removal of the raw data to eliminate errors and duplicate records; completion of missing values ​​using linear interpolation or neighborhood averaging; standardization of data from different sources based on their dimensional differences, with borehole data converted to a unified interval using interval scaling and geophysical and geochemical data standardized using Z-score to achieve statistical equilibrium; and removal of local noise using data filtering or smoothing methods to obtain a cleaned geological dataset that can be used for subsequent spatial registration and voxel division.

3. The mineral resource prediction method based on three-dimensional geological information and knowledge graph according to claim 1, characterized in that, Mineralization Favorability Index of Spatial Units The evidence weight method based on probability logarithmic weights is used for calculation, and the results are calculated according to voxel units. Implemented unit by unit, specifically including: Let the set of voxels corresponding to the known mineralization be denoted as the positive sample set. S 1. The set of voxels corresponding to non-mineralized bodies is denoted as the negative sample set. S 2; The probability of each mineral control index k appearing in the two sets is denoted as P. 1,k With P 2,k The evidentiary value of calculating this mineral control index: ； voxels The fracture geometric center distance is denoted as The alteration coverage ratio is denoted as The lithological mineralization indicator value is denoted as The above-mentioned ore-controlling indicators are then weighted and summed to obtain the mineralization favorable index. : ； in For bias terms, , , ∈[0,1], and , >0 represents the distance attenuation coefficient; when ≥ At that time, voxels are identified as candidate units for mineralization-favorable voxels, and the threshold is set. This is a preset value.

4. The mineral resource prediction method based on three-dimensional geological information and knowledge graph according to claim 1, characterized in that, The construction of the geological knowledge map includes the following steps: Entity and relationship descriptions related to deposit type, ore-controlling structures, metallogenic epoch and mineralization indicators were extracted from geological literature, geological reports and exploration databases; The extracted text undergoes named entity recognition and relation extraction, mapping geological concepts to triples. h , r , e (), where h represents the initial geological entity, r represents the relationship type between entities, and e represents the associated geological entity; The extracted triples are subjected to semantic disambiguation, attribute completion, and consistency verification to form a semantically normalized set of entity relations. The entity relationship set is stored in a graph database, and the association strength and confidence between entities are calculated based on a knowledge reasoning algorithm to generate a complete geological knowledge graph, which is used to support subsequent mineral resource prediction and metallogenic regularity analysis.

5. The mineral resource prediction method based on three-dimensional geological information and knowledge graph according to claim 1, characterized in that, The weight correction based on knowledge rules includes the following steps: (1) Rule setting: Based on the relationship types between geological entities in the geological knowledge map, a set of weight correction rules is established. The relationship types include genetic control, symbiotic enrichment, temporal dependence and spatial proximity. (2) Weight correction calculation: For each voxel unit Based on the knowledge association strength matched in the knowledge rules With rule weight coefficient Calculate the correction amount The formula for calculating the correction amount is: ； Where m represents the number of knowledge rules associated with the voxel; (3) Index update: based on the correction amount The revised mineralization favorability index is obtained by updating according to the following formula. : ； in, This represents the knowledge correction coefficient, which ranges from [0, 1] and is used to control the degree of influence of knowledge constraints on the prediction results.

6. The mineral resource prediction method based on three-dimensional geological information and knowledge graph according to claim 1, characterized in that, The process of screening target regions includes: From all the identified continuous mineralization favorable areas, select those that meet the preset minimum area based on their area. Required target area; Output the spatial location, extent, geometric center coordinates, and area information of the filtered target area for subsequent exploration deployment and visualization.

Images