Ore body prediction method and system based on multi-source geological information

Abstract

The invention relates to the field of ore body prediction, in particular to an ore body prediction method and system based on multi-source geological information. The method comprises the following steps: acquiring multi-source geological information data of a target exploration area, constructing a geological factor layer based on the multi-source geological information data, and establishing a geological factor map; constructing a metallogenic prediction model, inputting the metallogenic prediction model to the metallogenic prediction model, obtaining a metallogenic potential value of the target exploration area, obtaining a factor path set, and constructing a metallogenic interpretation path tree; dividing space control region units according to the metallogenic interpretation path tree, mapping each metallogenic interpretation path to a space control region, generating a path node space relationship, analyzing the path node space relationship, and generating a structure prediction path relationship; and when the space control area receives the newly added multi-source geological information data, analyzing the metallogenic interpretation path, and updating the corresponding metallogenic potential value and the metallogenic interpretation path. According to the method, the ore body distribution prediction precision and the dynamic response capability can be improved.

Description

Technical Field

[0001] This invention relates to the field of ore body prediction, and specifically to a method and system for ore body prediction based on multi-source geological information. Background Technology

[0002] As geological prospecting moves towards greater depth, precision, and intelligence, the rapid application of big data technology in the geological field in recent years has provided new tools for mineral resource prediction. By processing and analyzing multi-source heterogeneous geological data, including remote sensing imagery, geophysical data, geochemical data, geological mapping, and historical exploration results, a more representative geological factor layer system can be constructed, providing data support for modeling regional metallogenic regularities. Simultaneously, with the help of spatial dynamic modeling, map representation, and machine learning algorithms, it is expected that accurate inferences about metallogenic potential can be made, providing auxiliary decision-making basis for prospecting deployment.

[0003] Chinese Patent Publication No. CN114114458A discloses a method for predicting deep blind ore bodies in sandstone-type uranium deposits under thick overburden background, including: Step 1, systematically studying the geological conditions of the basin and determining the spatiotemporal location of favorable infiltration windows in the study area; Step 2, analyzing the strata of favorable infiltration windows in the study area and determining the spatiotemporal location of favorable uranium reservoirs; Step 3, conducting epigenetic modification analysis on favorable uranium reservoirs and predicting mineralization prospective areas; Step 4, drilling to verify the predicted mineralization prospective areas.

[0004] In existing technologies, the lack of analysis on the relationship between the spatial relationship of ore bodies and geological information leads to poor model interpretability, making it difficult to support subsequent geological verification and decision analysis. The analysis of influencing factors of prediction model results often relies on black box models, lacking path-based and structured causal explanation mechanisms. The model update process fails to be dynamically bound to the spatial control unit, resulting in the inability of newly added geological information data to be effectively mapped into the existing prediction structure, leading to poor timeliness. These are problems that we need to solve. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a method and system for predicting ore bodies based on multi-source geological information.

[0006] The technical solution of this invention: a method for predicting ore bodies based on multi-source geological information, comprising the following steps: S1. Obtain multi-source geological information data of the target exploration area, construct a geological factor layer based on each type of multi-source geological information data, and establish a geological factor map based on each geological factor layer; S2. Construct a mineralization prediction model. Input the geological factor map into the mineralization prediction model, obtain the mineralization potential value of the target exploration area, obtain the factor path set that affects the mineralization potential value, and construct a mineralization interpretation path tree. S3. Divide the spatial control area units according to the metallogenic interpretation path tree, map each metallogenic interpretation path to its corresponding spatial control area, generate the spatial relationship of path nodes, analyze the spatial relationship of path nodes, and generate the structural prediction path relationship. S4. When the spatial control area receives new multi-source geological information data, analyze the mineralization interpretation path based on the structural prediction path relationship, and update the corresponding mineralization potential value and mineralization interpretation path.

[0007] Preferably, the process of acquiring multi-source geological information data of the target exploration area, constructing geological factor layers based on each type of multi-source geological information data, and establishing geological factor maps based on each geological factor layer includes: Acquire multi-source geological information data of the target exploration area, including quantitative and categorical data; denote the quantitative and categorical data as geological factors; perform rasterization and numbering / coding on the quantitative and categorical data respectively; construct corresponding geological factor nodes based on their spatial location and raster or coded values, each geological factor node containing a node number, spatial range, data type, and initial attribute value; the initial attribute value refers to the raster or coded value; group multiple geological factor nodes belonging to the same data type source into corresponding geological factor layers, generating a geological factor layer set containing multiple quantitative and categorical geological factor layers. The response intensity values ​​of quantitative geological factor layers and known orebody areas are calculated using the evidence weight method, and the discrimination degree of categorical geological factor layers to known orebody areas is calculated using the information content index method. The response intensity values ​​and discrimination degrees are recorded as node weights. The spatial overlap rate and correlation coefficient between any two geological factor layers in the target exploration area are calculated. When the spatial overlap rate is greater than a preset spatial overlap threshold and the correlation coefficient is greater than a preset correlation threshold, edge connection relationships between geological factor layers are established, and edge weights are calculated. The edge weights are the weighted average of the spatial overlap rate and the correlation coefficient. Based on the geological factor layers, node weights, and edge weights, a geological factor map containing a set of nodes, a set of edges, and a set of attributes is constructed.

[0008] Preferably, the process of constructing a mineralization prediction model, inputting geological factor maps into the mineralization prediction model, and obtaining the mineralization potential value of the target exploration area includes: The spatial cells obtained from the geological factor map after rasterization are obtained. The node weights and edge weights corresponding to each spatial cell in the geological factor map are extracted to construct a geological feature matrix. Each row in the geological feature matrix corresponds to a spatial cell, and each column corresponds to the attribute features of a geological factor node, including the node weights and edge weights of the geological factor node. Using the geological feature matrix as input, a mineralization potential prediction model trained based on a supervised learning algorithm is invoked. The mineralization potential prediction model uses known mineralization units in the historical exploration area as positive samples and undiscovered ore body units as negative samples. A training set is constructed through labeling, and a stable model is obtained after parameter tuning using cross-validation. The stable model is then used to analyze all spatial units in the target area and outputs a mineralization potential score.

[0009] Preferably, the process of obtaining the set of factor paths affecting the mineralization potential value and constructing a mineralization interpretation path tree is as follows: The geological feature matrix of the input and the mineralization potential score of the mineralization potential prediction model are interpreted and analyzed. The SHAP algorithm based on Shapley value theory is adopted. The SHAP algorithm calculates the marginal contribution of the attribute features of the geological feature matrix of each geological factor to the mineralization potential score of the current spatial unit by constructing incremental prediction changes of geological factor feature combinations, and generates a set of geological contribution values. Geological factor nodes are sorted in descending order according to their corresponding marginal contributions. The top K geological factor nodes with the largest marginal contributions are selected as key geological factor nodes. A mineralization interpretation path tree structure is constructed by combining the edge weights in the geological factor map. In the mineralization interpretation path tree structure, the root node corresponds to the spatial unit of the key geological factor node, and the child nodes correspond to the key geological factor nodes. Each mineralization interpretation path is a directed node sequence that starts from the root node and connects multiple child nodes in sequence. The terminating node is denoted as the leaf node. The edge weight of the connecting edge between nodes is the weighted value of the marginal contribution of the key geological factor node to the score value and the edge weight in the map.

[0010] Preferably, the process of dividing spatial control area units based on the mineralization interpretation path tree, mapping each mineralization interpretation path to its corresponding spatial control area, and generating spatial relationships of path nodes includes: Based on the spatial coordinate range of each geological factor node in the mineralization interpretation path tree, the spatial coverage area of ​​each geological factor node in the geological factor map is extracted; the spatial location coordinate range is the geographic coordinate range of each geological factor node in the target exploration area, which is obtained through a geographic information system. Based on the co-occurrence edge connection relationship and its edge weight between geological factor nodes, the co-occurrence edge connection relationship refers to the spatial overlap or adjacency relationship between two or more geological factor nodes, which is used to represent the spatial association between geological factors; calculate the boundary coordinate range of the spatially overlapping area between geological factor nodes, and use the sliding window clustering algorithm to divide the area with a spatial overlap rate greater than the set overlap threshold into a group of continuous spatial control area units; For each spatial control area unit, record the range of geological factor nodes, node coordinate boundaries, co-occurrence edge weights, and corresponding marginal contributions contained within the spatial control area unit to generate a regional attribute record set; match each mineralization interpretation path in the mineralization interpretation path tree with the spatial control area unit it covers to determine the one-to-one correspondence between geological factor nodes and spatial control area units, and generate the spatial relationship of path nodes.

[0011] Preferably, the process of analyzing the spatial relationships of path nodes and generating structural prediction path relationships includes: The spatial relationships of path nodes and the corresponding regional attribute record sets are processed, and a unique path identifier is assigned to each mineralization interpretation path. The path identifier includes the path number, key factor sequence number and spatial unit number. Establish structural prediction path relationships, with the path identifier as the primary key field, and record the corresponding spatial control region number, edge weight set, node coordinate boundary information, and model reference information; the model reference information includes the prediction model name, training version number, and model call parameters. Consistency checks and duplicate path merging operations are performed on each record item within the structural prediction path relationship. Path records with spatial overlap rate and edge weight difference less than a set threshold are merged into a single control unit record to generate the structural prediction path relationship.

[0012] Preferably, when the spatial control area receives newly added multi-source geological information data, the process of analyzing the mineralization interpretation path based on the structural prediction path relationship and updating the corresponding mineralization potential value and mineralization interpretation path includes: When new geological information data is received within the target exploration area, the corresponding spatial coordinate information and the attribute features of each geological factor node as defined in the geological factor map are extracted to generate a new set of spatial data points. The spatial coordinates of the new data points are spatially matched with the spatial control area units recorded in the structural prediction path relationship to determine whether the new geological information data falls into any existing spatial control area unit corresponding to the mineralization interpretation path tree. If it does, the SHAP algorithm recorded in the mineralization potential prediction model is called to recalculate the marginal contribution of each geological factor based on the attribute features of the geological factor node corresponding to the new data point, and compare it with the marginal contribution recorded in the previous prediction to obtain the change in marginal contribution. When the marginal contribution change of any geological factor exceeds the set change threshold, the model reference information recorded in the structural prediction path relationship of the corresponding key geological factor node is called to retrieve the corresponding metallogenic potential prediction model; based on the newly added data points and the historical exploration data under the corresponding metallogenic interpretation path, the metallogenic potential prediction model is retrained to generate an updated metallogenic potential score; based on the updated metallogenic potential score and marginal contribution, the metallogenic interpretation path tree corresponding to the metallogenic interpretation path is reconstructed, and the metallogenic potential score and structural prediction path relationship are updated.

[0013] This invention also discloses a orebody prediction system based on multi-source geological information, including a management center, which is communicatively connected to a geological information acquisition module, a geological prediction module, a regional division module, and a dynamic update module. The geological information acquisition module is used to acquire multi-source geological information data of the target exploration area, construct geological factor layers based on each type of multi-source geological information data, and establish geological factor maps based on each geological factor layer; The geological prediction module is used to construct a mineralization prediction model. It inputs the geological factor map into the mineralization prediction model, obtains the mineralization potential value of the target exploration area, obtains the factor path set that affects the mineralization potential value, and constructs a mineralization interpretation path tree. The region division module is used to divide spatial control area units according to the metallogenic interpretation path tree, map each metallogenic interpretation path to its corresponding spatial control area, generate spatial relationships of path nodes, analyze the spatial relationships of path nodes, and generate structural prediction path relationships. The dynamic update module is used to analyze the mineralization interpretation path based on the structural prediction path relationship when the spatial control area receives new multi-source geological information data, and update the corresponding mineralization potential value and mineralization interpretation path.

[0014] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: by constructing a mineralization prediction model based on geological factor maps and combining it with the SHAP algorithm to realize a feature interpretability mechanism, it can effectively identify key geological factor paths that affect the mineralization potential value, construct a mineralization interpretation path tree, significantly improve the interpretability and credibility of the prediction results, and facilitate guidance for subsequent geological verification work; by mapping the mineralization interpretation path with spatial control area units and establishing structural prediction path relationships, after new geological information data is input, the mineralization potential value can be dynamically updated and the path structure can be adaptively adjusted based on the path structure, improving the prediction model's responsiveness to changes in geological data and enhancing its dynamic adaptability and practicality. Attached Figure Description

[0015] Figure 1 This is a flowchart of one embodiment of the present invention. Detailed Implementation

[0016] Example 1, as Figure 1 As shown, the present invention proposes a method for predicting ore bodies based on multi-source geological information, which includes the following steps: S1. Obtain multi-source geological information data of the target exploration area, construct a geological factor layer based on each type of multi-source geological information data, and establish a geological factor map based on each geological factor layer; S2. Construct a mineralization prediction model. Input the geological factor map into the mineralization prediction model, obtain the mineralization potential value of the target exploration area, obtain the factor path set that affects the mineralization potential value, and construct a mineralization interpretation path tree. S3. Divide the spatial control area units according to the metallogenic interpretation path tree, map each metallogenic interpretation path to its corresponding spatial control area, generate the spatial relationship of path nodes, analyze the spatial relationship of path nodes, and generate the structural prediction path relationship. S4. When the spatial control area receives new multi-source geological information data, analyze the mineralization interpretation path based on the structural prediction path relationship, and update the corresponding mineralization potential value and mineralization interpretation path.

[0017] It should be further explained that, in the specific implementation process, the process of acquiring multi-source geological information data of the target exploration area, constructing geological factor layers based on each type of multi-source geological information data, and establishing geological factor maps based on each geological factor layer is as follows: Acquire multi-source geological information data of the target exploration area. The multi-source geological information data includes quantitative data and categorical data. The quantitative data includes remote sensing data, geophysical data, geochemical data, and historical exploration numerical field data. The categorical data includes stratigraphic structure data, fault zone distribution data, and historical exploration category field data. Process the multi-source geological information data. The processing includes spatial registration, missing value removal, coordinate unification, and accuracy resampling. The quantitative data and categorical data are respectively denoted as geological factors, which are used to describe a certain type of geological feature. Quantitative and categorical data are rasterized and coded respectively. For each data source, a corresponding geological factor node is constructed based on its spatial location and raster or coded value. Each geological factor node includes a node number, spatial range, data type, and initial attribute value. The data type includes quantitative and categorical types. The initial attribute value refers to the raster or coded value. Multiple geological factor nodes belonging to the same data type source are grouped together to form a corresponding geological factor layer, generating a geological factor layer set containing multiple quantitative and categorical geological factor layers. Spatial cross-analysis is performed between each geological factor layer and the known ore body distribution area within the target exploration area. For quantitative geological factor layers, the evidence weight method is used to calculate the response intensity value between the quantitative geological factor layer and the known ore body area. For categorical geological factor layers, the information content index method is used to calculate the discrimination degree of the categorical geological factor layer to the known ore body area. The response intensity value and discrimination degree are used as the node weights of the corresponding quantitative geological factor layer and categorical geological factor layer, respectively. Specifically, the evidence weighting method is used to evaluate the responsiveness of continuous or quantitative geological factors to the occurrence of ore bodies. It belongs to the log response model under the Bayes probability framework, dividing the quantitative factor layer into several graded intervals; calculating the number of known ore body pixels and non-ore body pixels in each graded interval, as well as the total number of ore body and non-ore body pixels in the entire map; and calculating the response intensity value for each graded interval. The information content index method is mainly used to evaluate the ability of the category-based geological factor layer to distinguish the distribution of ore bodies, counting the number of known ore body pixels and non-ore body pixels under each category, as well as the total number of ore bodies and non-ore bodies in the entire map; and calculating the discrimination degree of each category. Calculate the spatial overlap rate R(i,j) and Pearson correlation coefficient C(i,j) between any two geological factor layers within the target exploration area. When the spatial overlap rate R(i,j) is greater than a preset spatial overlap threshold and the correlation coefficient is greater than a preset correlation threshold, establish an edge connection relationship between the geological factor layers and calculate the edge weight, which is the weighted average of the spatial overlap rate and the Pearson correlation coefficient; denoted as: Where R(i,j) represents the spatial overlap rate between geological factor layers i and j, C(i,j) represents their Pearson correlation coefficient, and α and β are preset weighting factors that satisfy... ; A geological factor atlas is constructed based on geological factor layers, node weights, and edge weights, comprising a set of nodes, an edge set, and an attribute set. The node set includes the number, location, type, and node weight of each geological factor node. The edge set includes the start and end numbers and edge weights of each pair of connected nodes. The attribute set includes the atlas number, exploration area identifier, timestamp, data source label, and coordinate system type.

[0018] It should be further explained that, in the specific implementation process, the process of constructing a mineralization prediction model, inputting geological factor maps into the mineralization prediction model, obtaining the mineralization potential value of the target exploration area, obtaining the factor path set that affects the mineralization potential value, and constructing a mineralization interpretation path tree is as follows: Obtain spatial units from the geological factor map after rasterization, wherein the spatial unit is the smallest geographic raster region generated after rasterization in the geological factor map; Extract the node weights and edge weights corresponding to each spatial unit in the geological factor map to construct a geological feature matrix. Each row in the geological feature matrix corresponds to a spatial unit, and each column corresponds to the attribute features of a geological factor node, including the node weights and edge weights of the geological factor node. Using a geological feature matrix as input, a metallogenic potential prediction model trained based on a supervised learning algorithm is invoked. This model uses known metallogenic units in historical exploration areas as positive samples and undiscovered ore body units as negative samples. A training set is constructed through labeling, and a stable model is obtained after parameter tuning using cross-validation. The stable model is then used to analyze all spatial units within the target area, extracting the geological feature vector corresponding to each spatial unit. This geological feature vector is composed of the node weights and edge weights associated with the spatial unit in the geological factor map, representing the comprehensive geological attributes of the spatial unit. The geological feature vector of each spatial unit is input into the trained metallogenic potential prediction model for analysis. Based on the mapping rules between geological factor nodes and metallogenic relationships learned during the training phase, the metallogenic potential prediction model calculates the probability value that the current spatial unit belongs to a "potential metallogenic unit" and outputs a metallogenic potential score value. This score value is a floating-point number in the interval [0, 1], used to represent the metallogenic probability of each unit.

[0019] The geological feature matrix of the input and the mineralization potential score of the mineralization potential prediction model are interpreted and analyzed. The SHAP algorithm based on Shapley value theory is adopted. The SHAP algorithm calculates the marginal contribution of the attribute features of the geological feature matrix of each geological factor to the mineralization potential score of the current spatial unit by constructing incremental prediction changes of geological factor feature combination, and generates a set of geological contribution values, denoted as SHAP(i), which is used to represent the influence intensity of geological factor i on the prediction result. Specifically, the SHAP algorithm treats each attribute feature as a participating variable in the prediction process of the mineralization potential prediction model. By sequentially adding or subtracting attribute features from the entire feature set, the change in the model output value is calculated to measure the marginal contribution of the attribute feature to the prediction result. Each column in the geological feature matrix is ​​a feature dimension, and the mineralization potential score f(x) output by the model is used as the objective function. All possible feature subsets S are traversed, and the weighted average of the difference between f(S∪{i}) and f(S) is calculated to obtain the marginal contribution SHAP(i) corresponding to each geological factor i. The SHAP(i) values ​​of all geological factors are combined into a geological contribution value set to represent the influence intensity of each geological factor on the mineralization potential score of the current spatial unit; a positive value indicates that the factor increases the mineralization potential, and a negative value indicates that it weakens the mineralization potential. Geological factor nodes are sorted in descending order according to their corresponding marginal contributions. The top K geological factor nodes with the largest marginal contributions are selected as key geological factor nodes, and a mineralization interpretation path tree structure is constructed by combining the edge weights in the geological factor map. In the mineralization interpretation path tree structure, the root node corresponds to the spatial unit of the key geological factor node, and the child nodes correspond to the key geological factor nodes. Each mineralization interpretation path is a directed node sequence formed by connecting multiple child nodes sequentially from the root node. The terminating node is denoted as a leaf node. The edge weight of the connection edge between nodes is the weighted value of the SHAP marginal contribution of the key geological factor node to the score value and the edge weight in the map. The weighting formula is: ;in, This refers to the weighted value. This refers to the marginal contribution. This refers to the edge weight. ; The mineralization interpretation path tree structure is used to represent the contribution order and synergistic relationship of each key geological factor node in the mineralization potential prediction model. It also supports users in visualizing the causal path, evaluating the model's credibility, and designing subsequent local geological verification for the mineralization results of specific spatial units, thereby improving the interpretability and practicality of the model.

[0020] It should be further explained that, in the specific implementation process, the spatial control area units are divided according to the metallogenic interpretation path tree, each metallogenic interpretation path is mapped to its corresponding spatial control area, the spatial relationships of path nodes are generated, and the spatial relationships of path nodes are analyzed to generate the structural prediction path relationships. The process is as follows: Based on the spatial coordinate range of each geological factor node in the mineralization interpretation path tree, the spatial coverage area of ​​each geological factor node in the geological factor map is extracted; the spatial location coordinate range is the geographical coordinate range of each geological factor node in the target exploration area, which is obtained through a geographic information system. Based on the co-occurrence edge connection relationship and its edge weight between geological factor nodes, the co-occurrence edge connection relationship refers to the spatial overlap or adjacency relationship between two or more geological factor nodes, which is used to represent the spatial association between geological factors; calculate the boundary coordinate range of the spatially overlapping area between geological factor nodes, and use the sliding window clustering algorithm to divide the area with a spatial overlap rate greater than a set overlap threshold into a group of continuous spatial control area units; For each spatial control area unit, record the range of geological factor nodes, node coordinate boundaries, co-occurrence edge weights, and corresponding marginal contributions contained within the spatial control area unit, and generate a regional attribute record set; Each mineralization interpretation path in the mineralization interpretation path tree is matched with the spatial control area unit it covers to determine the one-to-one correspondence between geological factor nodes and spatial control area units, and to generate the spatial relationship of path nodes, which is used to describe the correspondence structure between paths and spatial distribution.

[0021] The spatial relationships of path nodes and the corresponding regional attribute record sets are processed to assign a unique path identifier to each mineralization interpretation path. The path identifier includes path number, key factor sequence number and spatial unit number. Establish a structural prediction path relationship, wherein the structural prediction path relationship uses the path identifier as the primary key field and records the corresponding spatial control region number, edge weight set, node coordinate boundary information and model reference information; The model reference information includes the prediction model name, training version number, and model calling parameters, which are used to quickly call the prediction configuration of the corresponding spatial region when the model is updated or retrained. Consistency checks and duplicate path merging operations are performed on each record item within the structural prediction path relationship. Path records with spatial overlap rate and edge weight difference less than a set threshold are merged into a single control unit record to generate structural prediction path relationships, which are used for subsequent dynamic updates of the mineralization potential model and spatial result callbacks.

[0022] It should be further explained that, in the specific implementation process, when the spatial control area receives newly added multi-source geological information data, the process of analyzing the mineralization interpretation path based on the structural prediction path relationship and updating the corresponding mineralization potential value and mineralization interpretation path is as follows: When new geological information data is received from the target exploration area, the corresponding spatial coordinate information and the attribute features of each geological factor node as defined in the geological factor map are extracted to form a new spatial data point set. Spatial matching is performed between the spatial coordinates of the newly added data points and the spatial control area units recorded in the structural prediction path relationship. It is determined whether the newly added geological information data falls into any existing spatial control area unit corresponding to the mineralization interpretation path tree. If it does, the SHAP algorithm recorded in the mineralization potential prediction model is called. Based on the attribute characteristics of the geological factor node input corresponding to the newly added data points, the marginal contribution of each geological factor is recalculated and compared with the marginal contribution recorded in the previous prediction to obtain the change in marginal contribution. When the marginal contribution of any geological factor changes beyond the set threshold, the model reference information recorded in the structural prediction path relationship of the corresponding key geological factor node is invoked to retrieve the corresponding mineralization potential prediction model. Based on the newly added data points and the historical exploration data under the corresponding mineralization interpretation path, the historical exploration data includes historical exploration numerical field data and historical exploration category field data, a retraining process is carried out on the mineralization potential prediction model to generate an updated mineralization potential score value, with the score value ranging from [0, 1]. Based on the updated mineralization potential score and marginal contribution, the mineralization interpretation path tree corresponding to the mineralization interpretation path is reconstructed, and the updated results are written into the corresponding spatial position in the structural prediction path relationship, so as to realize the dynamic replacement of the mineralization potential score and the structural prediction path relationship.

[0023] Example 2: The orebody prediction system based on multi-source geological information proposed in this invention is applied to the orebody prediction method based on multi-source geological information described in Example 1. Specifically, it includes a management center, which is communicatively connected to a geological information acquisition module, a geological prediction module, a regional division module, and a dynamic update module. The geological information acquisition module is used to acquire multi-source geological information data of the target exploration area, construct geological factor layers based on each type of multi-source geological information data, and establish geological factor maps based on each geological factor layer; The geological prediction module is used to construct a mineralization prediction model. It inputs the geological factor map into the mineralization prediction model, obtains the mineralization potential value of the target exploration area, obtains the factor path set that affects the mineralization potential value, and constructs a mineralization interpretation path tree. The region division module is used to divide spatial control area units according to the metallogenic interpretation path tree, map each metallogenic interpretation path to its corresponding spatial control area, generate spatial relationships of path nodes, analyze the spatial relationships of path nodes, and generate structural prediction path relationships. The dynamic update module is used to analyze the mineralization interpretation path based on the structural prediction path relationship when the spatial control area receives new multi-source geological information data, and update the corresponding mineralization potential value and mineralization interpretation path.

[0024] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for predicting ore bodies based on multi-source geological information, characterized in that, Includes the following steps: S1. Obtain multi-source geological information data of the target exploration area, construct a geological factor layer based on each type of multi-source geological information data, and establish a geological factor map based on each geological factor layer; S2. Construct a mineralization prediction model. Input the geological factor map into the mineralization prediction model, obtain the mineralization potential value of the target exploration area, obtain the factor path set that affects the mineralization potential value, and construct a mineralization interpretation path tree. S3. Divide the spatial control area units according to the metallogenic interpretation path tree, map each metallogenic interpretation path to its corresponding spatial control area, generate the spatial relationship of path nodes, analyze the spatial relationship of path nodes, and generate the structural prediction path relationship. S4. When the spatial control area receives new multi-source geological information data, analyze the mineralization interpretation path based on the structural prediction path relationship, and update the corresponding mineralization potential value and mineralization interpretation path.

2. The orebody prediction method based on multi-source geological information according to claim 1, characterized in that, The process of acquiring multi-source geological information data of the target exploration area, constructing geological factor layers based on each type of multi-source geological information data, and establishing geological factor maps based on each geological factor layer includes: Acquire multi-source geological information data of the target exploration area, including quantitative and categorical data; denote the quantitative and categorical data as geological factors; perform rasterization and numbering / coding on the quantitative and categorical data respectively; construct corresponding geological factor nodes based on their spatial location and raster or coded values, each geological factor node containing a node number, spatial range, data type, and initial attribute value; the initial attribute value refers to the raster or coded value; group multiple geological factor nodes belonging to the same data type source into corresponding geological factor layers, generating a geological factor layer set containing multiple quantitative and categorical geological factor layers. The response intensity values ​​of quantitative geological factor layers and known orebody areas are calculated using the evidence weight method, and the discrimination degree of categorical geological factor layers to known orebody areas is calculated using the information content index method. The response intensity values ​​and discrimination degrees are recorded as node weights. The spatial overlap rate and correlation coefficient between any two geological factor layers in the target exploration area are calculated. When the spatial overlap rate is greater than a preset spatial overlap threshold and the correlation coefficient is greater than a preset correlation threshold, edge connection relationships between geological factor layers are established, and edge weights are calculated. The edge weights are the weighted average of the spatial overlap rate and the correlation coefficient. Based on the geological factor layers, node weights, and edge weights, a geological factor map containing a set of nodes, a set of edges, and a set of attributes is constructed.

3. The orebody prediction method based on multi-source geological information according to claim 2, characterized in that, The process of constructing a mineralization prediction model, inputting geological factor maps into the model, and obtaining the mineralization potential value of the target exploration area includes: The spatial cells obtained from the geological factor map after rasterization are obtained. The node weights and edge weights corresponding to each spatial cell in the geological factor map are extracted to construct a geological feature matrix. Each row in the geological feature matrix corresponds to a spatial cell, and each column corresponds to the attribute features of a geological factor node, including the node weights and edge weights of the geological factor node. Using the geological feature matrix as input, a mineralization potential prediction model trained based on a supervised learning algorithm is invoked. The mineralization potential prediction model uses known mineralization units in the historical exploration area as positive samples and undiscovered ore body units as negative samples. A training set is constructed through labeling, and a stable model is obtained after parameter tuning using cross-validation. The stable model is then used to analyze all spatial units in the target area and outputs a mineralization potential score.

4. The orebody prediction method based on multi-source geological information according to claim 3, characterized in that, The process of obtaining the set of factor paths that influence mineralization potential and constructing a mineralization interpretation path tree is as follows: The geological feature matrix of the input and the mineralization potential score of the mineralization potential prediction model are interpreted and analyzed. The SHAP algorithm based on Shapley value theory is adopted. The SHAP algorithm calculates the marginal contribution of the attribute features of the geological feature matrix of each geological factor to the mineralization potential score of the current spatial unit by constructing incremental prediction changes of geological factor feature combinations, and generates a set of geological contribution values. The geological factor nodes are sorted in descending order according to their corresponding marginal contribution. The top K geological factor nodes with the largest marginal contribution are selected as key geological factor nodes. The mineralization interpretation path tree structure is constructed by combining the edge weights in the geological factor map. In the metallogenic interpretation path tree structure, the root node corresponds to the spatial unit of the key geological factor node, the child nodes correspond to the key geological factor node, each metallogenic interpretation path is a directed node sequence formed by connecting multiple child nodes in sequence from the root node, the terminating node is denoted as the leaf node, and the edge weight of the connecting edge between nodes is the weighted value of the marginal contribution of the key geological factor node to the score value and the edge weight in the graph.

5. The orebody prediction method based on multi-source geological information according to claim 4, characterized in that, The process of dividing spatial control area units based on the metallogenic interpretation path tree, mapping each metallogenic interpretation path to its corresponding spatial control area, and generating spatial relationships of path nodes includes: Based on the spatial coordinate range of each geological factor node in the mineralization interpretation path tree, the spatial coverage area of ​​each geological factor node in the geological factor map is extracted; the spatial location coordinate range is the geographic coordinate range of each geological factor node in the target exploration area, which is obtained through a geographic information system. Based on the co-occurrence edge connection relationship and its edge weight between geological factor nodes, the co-occurrence edge connection relationship refers to the spatial overlap or adjacency relationship between two or more geological factor nodes, which is used to represent the spatial association between geological factors; calculate the boundary coordinate range of the spatially overlapping area between geological factor nodes, and use the sliding window clustering algorithm to divide the area with a spatial overlap rate greater than the set overlap threshold into a group of continuous spatial control area units; For each spatial control area unit, record the range of geological factor nodes, node coordinate boundaries, co-occurrence edge weights, and corresponding marginal contributions contained within the spatial control area unit to generate a regional attribute record set; match each mineralization interpretation path in the mineralization interpretation path tree with the spatial control area unit it covers to determine the one-to-one correspondence between geological factor nodes and spatial control area units, and generate the spatial relationship of path nodes.

6. The orebody prediction method based on multi-source geological information according to claim 5, characterized in that, The process of analyzing the spatial relationships of path nodes and generating structural prediction path relationships includes: The spatial relationships of path nodes and the corresponding regional attribute record sets are processed, and a unique path identifier is assigned to each mineralization interpretation path. The path identifier includes the path number, key factor sequence number and spatial unit number. Establish structural prediction path relationships, with the path identifier as the primary key field, and record the corresponding spatial control region number, edge weight set, node coordinate boundary information, and model reference information; the model reference information includes the prediction model name, training version number, and model call parameters. Consistency checks and duplicate path merging operations are performed on each record item within the structural prediction path relationship. Path records with spatial overlap rate and edge weight difference less than a set threshold are merged into a single control unit record to generate the structural prediction path relationship.

7. The orebody prediction method based on multi-source geological information according to claim 6, characterized in that, When the spatial control area receives new multi-source geological information data, the process of analyzing the mineralization interpretation path based on the structural prediction path relationship and updating the corresponding mineralization potential value and mineralization interpretation path includes: When new geological information data is received within the target exploration area, the corresponding spatial coordinate information and the attribute features of each geological factor node as defined in the geological factor map are extracted to generate a new set of spatial data points. The spatial coordinates of the new data points are spatially matched with the spatial control area units recorded in the structural prediction path relationship to determine whether the new geological information data falls into any existing spatial control area unit corresponding to the mineralization interpretation path tree. If it does, the SHAP algorithm recorded in the mineralization potential prediction model is called to recalculate the marginal contribution of each geological factor based on the attribute features of the geological factor node corresponding to the new data point, and compare it with the marginal contribution recorded in the previous prediction to obtain the change in marginal contribution. When the marginal contribution change of any geological factor exceeds the set change threshold, the model reference information recorded in the structural prediction path relationship of the corresponding key geological factor node is called to retrieve the corresponding metallogenic potential prediction model; based on the newly added data points and the historical exploration data under the corresponding metallogenic interpretation path, the metallogenic potential prediction model is retrained to generate an updated metallogenic potential score; based on the updated metallogenic potential score and marginal contribution, the metallogenic interpretation path tree corresponding to the metallogenic interpretation path is reconstructed, and the metallogenic potential score and structural prediction path relationship are updated.

8. A mineral body prediction system based on multi-source geological information, specifically applied to the mineral body prediction method based on multi-source geological information as described in any one of claims 1 to 7, comprising a management center, characterized in that, The management center's communication connections include a geological information acquisition module, a geological prediction module, a regional division module, and a dynamic update module. The geological information acquisition module is used to acquire multi-source geological information data of the target exploration area, construct geological factor layers based on each type of multi-source geological information data, and establish geological factor maps based on each geological factor layer; The geological prediction module is used to construct a mineralization prediction model. It inputs the geological factor map into the mineralization prediction model, obtains the mineralization potential value of the target exploration area, obtains the factor path set that affects the mineralization potential value, and constructs a mineralization interpretation path tree. The region division module is used to divide spatial control area units according to the metallogenic interpretation path tree, map each metallogenic interpretation path to its corresponding spatial control area, generate spatial relationships of path nodes, analyze the spatial relationships of path nodes, and generate structural prediction path relationships. The dynamic update module is used to analyze the mineralization interpretation path based on the structural prediction path relationship when the spatial control area receives new multi-source geological information data, and update the corresponding mineralization potential value and mineralization interpretation path.

Images