Three-dimensional quantitative prediction method and system for concealed mine

Abstract

The invention provides a blind mine three-dimensional quantitative prediction method and system, and relates to the technical field of geological prospecting, and the method comprises the steps: collecting metallogenic and non-metallogenic model samples with the same number, carrying out the three-dimensional division of grid units, obtaining the feature values of a stratum system, a rock type, a structural feature and the like, taking whether there is a mine as a target variable, dividing a data set through stratified sampling, and carrying out the three-dimensional quantitative prediction of a blind mine. And a random forest model is constructed, and parameters are set and optimized. The method comprises the following steps: querying geological records, extracting geological feature data such as a stratum system, a rock type and structural features of a to-be-detected region, arranging the data into a geological feature data table, and importing the data table into professional geological modeling software to construct an accurate three-dimensional geological model, thereby providing a reliable geological basis for blind mine prediction; and then grids are divided, and required feature data are input into the model to predict the hidden mine, so that three-dimensional space accurate positioning prediction of the hidden mine is realized, a scientific basis is provided for mineral exploration and development, and the resource evaluation accuracy is improved.

Description

Technical Field

[0001] The invention relates to the technical field of geological exploration, and in particular to a three-dimensional quantitative prediction method and system for concealed mines. Background Art

[0002] With the growing demand for mineral resources, accurate detection and quantitative prediction of hidden mines have become key tasks in the field of geological exploration. Traditional geological exploration methods often have limitations when facing complex geological conditions and hidden ore bodies.

[0003] In the prior art, the publication number CN118070971A discloses a three-dimensional quantitative prediction method, computer equipment, medium and product for concealed mines. The method mainly obtains geological data of the target mining area through geological maps, geophysical data and geochemical data, analyzes the mineralization process based on the mineralization system theory and the three-in-one exploration area prospecting prediction theory to determine the ore-controlling factors, and determines the probability value of the existence of an ore body at each spatial position in the target mining area based on a machine learning algorithm. Its data type is relatively traditional and single. In a complex concealed environment, it is difficult to accurately capture the unique ore-controlling factors of a specific mining area by relying solely on these data. For example, some key information such as special geological structures, changes in rock microstructures, and mineralization and alteration characteristics may not be fully reflected, thereby affecting the accurate analysis of the mineralization process and the prediction of ore body positioning.

[0004] At the same time, the existing prediction methods are lacking in visualization and cannot intuitively display the three-dimensional structure of hidden mines in geological bodies. This makes it difficult for geologists to understand and analyze the prediction results, and it is difficult to deeply explore the relationship between geological bodies and hidden mines from a three-dimensional spatial perspective, which is not conducive to formulating scientific and reasonable exploration and mining plans.

[0005] In addition, as the amount of geological data continues to increase, traditional data processing and analysis methods are also facing challenges in terms of efficiency and accuracy. In the era of big data, how to more efficiently integrate and analyze massive geological data and mine the hidden mineralization information has become an urgent problem to be solved.

[0006] In view of this, the present invention aims to provide a more advanced and comprehensive three-dimensional quantitative prediction solution for concealed mines.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0008] The purpose of the present invention is to provide a three-dimensional quantitative prediction method and system for concealed mines to solve the problems raised in the above background technology.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A three-dimensional quantitative prediction method for concealed mines, comprising the following specific steps:

[0011] Step 1: Collect an equal number of mineralized and non-mineralized model samples, divide all model samples into grid units in three dimensions, take the stratigraphic thickness, strike, dip and inclination of the stratigraphic system of each grid unit as stratigraphic characteristic values, classify and code the rock types as rock characteristic values, and take the fault strike, dip, dip and fault throw of the structural characteristics as structural characteristic values;

[0012] Step 2: Take whether there is a mine as the target variable, associate the target variable with the corresponding stratigraphic characteristic value, rock characteristic value and structural characteristic value to form a set of geological characteristic data, calibrate the set of all geological characteristic data as model data, use the model data as the training set, build a random forest model, set the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes, and then use the accuracy rate to evaluate and optimize the random forest model;

[0013] Step 3: By querying the geological catalog, extracting the geological characteristic data of the area to be detected, saving it as a geological characteristic data table, and importing it into the geological modeling software to build a three-dimensional geological model. The geological characteristic data includes stratigraphic system, rock type, and structural characteristic data;

[0014] Step 4: Divide the constructed three-dimensional geological model into cubic grids, and input the geological characteristic data corresponding to each grid unit into the optimized random forest model according to the geological characteristic data table. The optimized random forest model predicts the grid units where hidden mines are likely to exist, thereby realizing the positioning prediction of hidden mines in three-dimensional space.

[0015] Furthermore, the collected mineralized and non-mineralized model samples are firstly gridded, and then the stratigraphic thickness of the jth measurement point in the i-th grid unit is set as h ij , where the total number of measurement points in a grid cell is n, i is used to distinguish different grid cells, each of which has its own independent set of measurement point data and the corresponding characteristic value calculation process of formation thickness, and j is a variable that numbers each measurement point in the i-th grid cell;

[0016] The calculation formula of the characteristic value of the formation thickness is:

[0017]

[0018] The formation strike angle of the jth measurement point in the i-th grid cell is set to α ij , first set the angle value α ijConvert to radian value for calculation. The converted radian value is:

[0019]

[0020] In the formula, θ ij It is expressed as the radian value of the stratigraphic strike angle of the jth measurement point in the i-th grid cell. Then x is calculated. ij = cosθ ij and ij = sinθ ij ,

[0021] The calculation formula of the characteristic value of the formation trend is:

[0022]

[0023] In the formula, θ i Expressed as the characteristic value of the stratigraphic trend of the ith grid cell under the arc value

[0024] Then the radian value θ i Converted into angle values ​​as characteristic values ​​of stratum direction:

[0025]

[0026] In the formula, α i is represented as the characteristic value of the stratigraphic trend of the i-th grid cell,

[0027] The formation dip angle of the jth measurement point in the i-th grid cell is expressed as β ij , first convert to radians:

[0028]

[0029] In the formula, It is expressed as the radian value of the formation dip angle at the jth measurement point in the i-th grid cell, and then calculated and

[0030] The calculation formula of the characteristic value of formation dip is:

[0031]

[0032] In the formula, It is expressed as the characteristic value of the stratigraphic dip of the ith grid cell under the arc value;

[0033] Then convert it into an angle value as the characteristic value of the formation dip:

[0034]

[0035] The formation dip angle of the jth measurement point in the i-th grid cell is expressed as γ ij , then the calculation formula of the characteristic value of the formation dip is:

[0036]

[0037] In the formula, γ i is represented as the characteristic value of the formation dip of the i-th grid cell,

[0038] Identify all possible rock types and create unique codes for each of them. k , where k = 1, 2, 3, ..., f, and f represents the total number of all different rock types that may appear in the i-th grid unit, forming a rock type code list, and setting the distribution area of ​​the k-th rock type in the i-th grid unit to A ik , the total area is S i , then the area ratio of the kth rock type in the grid unit is expressed as:

[0039]

[0040] In the formula, p ik It is expressed as the area proportion of the kth rock type in the i-th grid unit, A ik It is expressed as the distribution area of ​​the kth rock type in the ith grid unit, S i is expressed as the total area of ​​the ith grid cell,

[0041] In each grid cell, the area proportions of different rock types are compared. ik , extract the rock type with the highest proportion, and use the code of the rock type as the rock type coding feature value of this grid unit.

[0042] Furthermore, whether or not there is a mineralization is used as a stratification variable, and the total number of mineralized samples and non-mineralized samples is e. For all mineralized samples, a random number generator is used to generate Non-repeating random numbers are added to the mineralized samples with corresponding numbers in the training set. For non-mineralized samples, a random number generator is used to generate Non-repeating random numbers are added to all non-mineralized samples with corresponding numbers in the training set. When the overall accurate prediction ratio of the model for mineralized samples and non-mineralized samples in the training set, that is, the accuracy index, is higher than 90%, the training is stopped.

[0043] Furthermore, the maximum depth was initially set to 8, and the random forest model was constructed using the divided training set. The minimum number of samples required for node splitting was set to 8, and the minimum number of samples for leaf nodes was set to 5.

[0044] Furthermore, in each iteration of random forest model training, the sample data in the training set is input into the current random forest model for prediction. Let the number of correctly predicted mineralized samples in the training set be TP, the number of correctly predicted non-mineralized samples be TN, and the total number of mineralized samples and non-mineralized samples be e. After each round of training, the accuracy is calculated according to the formula:

[0045]

[0046] In the formula, Accuracy is expressed as the accuracy rate, e is expressed as the total number of mineralized samples and non-mineralized samples, TP is expressed as the correctly predicted number of mineralized samples, and TN is expressed as the correctly predicted number of non-mineralized samples;

[0047] When the accuracy is higher than 90%, training stops.

[0048] Furthermore, the name, thickness, strike, dip and inclination data of each layer in the stratigraphic system data are saved in a unified text format as a geological characteristic data table, and then imported into the geological modeling software. Different rock types are distinguished by different colors or materials, so that the model can visually reflect the distribution characteristics of rock types. The structural characteristic data is integrated into the model to show the position and shape of the fault in three-dimensional space and its intersection with the strata, so as to construct a complete three-dimensional geological model including stratigraphic structure, rock type and structural characteristics.

[0049] Furthermore, characteristic data are extracted from the constructed three-dimensional model of the area to be inspected in the same manner as when the model was trained, including stratum thickness, strike, dip, inclination, rock type code, and strike, dip, inclination and distance of the fault. The three-dimensional space of the area to be inspected is divided into cubic grids, and each grid unit is regarded as an independent sample point. All sample data are input into the trained random forest model. According to the input characteristic data, the relationship pattern between geological characteristics and mineralization learned in the training process is used. After the model prediction result is output, a probability threshold μ is set. When the probability value of the existence of a hidden mine corresponding to the prediction result of a certain grid unit is higher than the probability threshold μ, it is determined that a hidden mine exists at the location of this grid unit; otherwise, if the probability value does not reach the probability threshold μ, it is determined that no hidden mine exists at this location.

[0050] In addition, a three-dimensional quantitative prediction system for hidden mines is provided, and the system is used to execute the above-mentioned three-dimensional quantitative prediction method for hidden mines, comprising:

[0051] The feature extraction module is used to collect an equal number of mineralized and non-mineralized model samples and divide them into grid units in three dimensions. The stratigraphic thickness, strike, dip and inclination of the stratigraphic structure of each grid unit are used as feature values, the rock type classification code is used as feature values, the fault strike, dip, dip and fault distance of the structural characteristics are used as feature values, and whether there is a mine is used as the target variable;

[0052] The feature extraction module is used to collect an equal number of mineralized and non-mineralized model samples, divide all model samples into grid units in three dimensions, take the stratigraphic thickness, strike, dip and inclination of the stratigraphic structure of each grid unit as stratigraphic characteristic values, classify and encode rock types as rock characteristic values, and take the fault strike, dip, dip and fault throw of structural characteristics as structural characteristic values;

[0053] The model building module is used to take whether there is a mine as the target variable, associate the target variable with the corresponding stratum characteristic value, rock characteristic value and structural characteristic value to form a set of geological characteristic data, calibrate the collection of all geological characteristic data as model data, use the model data as a training set, build a random forest model, set the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes, and then use the accuracy rate to evaluate and optimize the random forest model;

[0054] A three-dimensional modeling module is used to extract geological characteristic data of the area to be detected by querying geological catalogs, save them as geological characteristic data tables, and import them into geological modeling software to construct a three-dimensional geological model. The geological characteristic data include stratum structure, rock type, and structural characteristic data.

[0055] The hidden mine positioning module is used to divide the constructed three-dimensional geological model into cubic grids, and input the geological characteristic data corresponding to each grid unit into the optimized random forest model according to the geological characteristic data table. The optimized random forest model predicts the grid units where hidden mines are likely to exist, thereby realizing the positioning prediction of hidden mines in three-dimensional space.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] By collecting an equal number of mineralized and non-mineralized model samples and dividing them into grid units in three dimensions, including multiple characteristic values ​​such as stratigraphic structure, rock type and structural characteristics, as well as the target variable of whether there is a mine, in the random forest model, we can avoid the model over-learning of one type of samples and under-learning of another type of samples due to unbalanced sample numbers, thereby improving the model's recognition and generalization capabilities for different mineralization situations.

[0058] After the mineralized and non-mineralized model samples are divided into grid units in three dimensions, the changes in the direction of the strata in different micro-areas can be clearly shown. For each grid unit, multiple characteristic values ​​of the stratigraphic structure, rock type and structural characteristics can be obtained in detail. The various geological characteristics in each grid unit are interrelated and influence each other, and work together in the mineralization process, providing rich information for revealing the mineralization mechanism. After many experiments and parameter adjustments, the accuracy of the model can reach a high level, effectively improving the accuracy of the prediction of hidden mines. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the overall method flow of the present invention.

[0060] Figure 2 It is a schematic diagram of the overall system module of the present invention. DETAILED DESCRIPTION

[0061] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0062] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0063] Example:

[0064] See also Figure 1 , the present invention provides a technical solution:

[0065] A three-dimensional quantitative prediction method for concealed mines, comprising the following specific steps:

[0066] Step 1: Collect an equal number of mineralized and non-mineralized model samples, divide all model samples into grid units in three dimensions, take the stratigraphic thickness, strike, dip and inclination of the stratigraphic system of each grid unit as stratigraphic characteristic values, classify and code the rock types as rock characteristic values, and take the fault strike, dip, dip and fault throw of the structural characteristics as structural characteristic values;

[0067] From the geological exploration reports of multiple explored mining areas, areas where a large amount of high-grade ore has been mined are determined as mineralized areas, while areas where no economically valuable ore bodies have been found after detailed exploration in the surrounding areas are determined as non-mineralized areas. For each area, geological characteristic data are collected, including the thickness, strike, dip, inclination, rock type, and structural characteristics of the stratigraphic system. The strike, dip, inclination and distance of the faults are finally collected. 500 mineralized and non-mineralized model samples are collected. Each sample contains complete stratigraphic structure, rock type, structural characteristic data, and an annotation of whether there is a mine.

[0068] The sample data of mineralized areas and non-mineralized areas represent different geological results. The combination of geological features they contain is the key basis for the model to learn to distinguish between mineralized and non-mineralized conditions. The model can better identify the subtle differences in geological characteristics between mineralized areas and non-mineralized areas by learning a large number of samples during the training process, so that when faced with data from unknown areas, it can more accurately predict the existence of hidden mines. For example, when the model learns that a combination of a certain rock type and a specific stratum thickness and fault structure frequently appears in a mineralized area, it can make judgments based on these feature patterns when predicting new areas, reducing the possibility of misjudgment.

[0069] First, the collected mineralized and non-mineralized model samples are divided into grids, where the grid is set to 5 meters, so that each grid unit can become a relatively independent and representative geological sample unit, which is convenient for subsequent data collection and analysis of various geological elements such as strata, structures, rock types, etc. in each small area, so as to more accurately understand the relationship between geological characteristics and mineralization.

[0070] Using the volume calculation function of the modeling software Surpac, the volume of the ore part and the volume of the entire grid are calculated respectively, and then the volume proportion of the ore part is calculated:

[0071]

[0072] Where P is the volume percentage of the ore-bearing part, V ore It is expressed as the volume of the mineralized part within the grid. In the process of model training and verification using modeling software and related algorithms, using 60% as the judgment limit can enable the model to achieve a better balance between accuracy and recall when predicting hidden mines. Therefore, when the ore content P in the grid is ≥ 60%, it is judged as a mineralized area. Conversely, when the ore content P in the grid is < 60%, it is judged as a non-mineralized area.

[0073] Then, the stratum thickness at the jth measuring point in the i-th grid cell is represented as h ij, where the total number of measurement points in a grid cell is n, i is used to distinguish different grid cells, each of which has its own independent set of measurement point data and the corresponding characteristic value calculation process of formation thickness, and j is a variable that numbers each measurement point in the i-th grid cell;

[0074] The calculation formula of the characteristic value of the formation thickness is:

[0075]

[0076] In the formula, H i It is the characteristic value of the formation thickness. The formation thickness may not be uniform within a grid unit. Multiple measurement points can more comprehensively reflect its changes and obtain a value that can represent the overall formation thickness level in the grid unit.

[0077] The formation strike angle of the jth measurement point in the i-th grid cell is set to α ij , first set the angle value α ij Convert to radian value for calculation. The converted radian value is:

[0078]

[0079] In the formula, θ ij It is expressed as the radian value of the stratigraphic strike angle of the jth measurement point in the i-th grid cell. Then x is calculated. ij = cosθ ij and ij = sinθ ij The stratigraphic strike refers to the direction of the intersection of the stratigraphic plane and the horizontal plane, which reflects the horizontal extension trend of the stratigraphic plane. First, the angle value α ij Convert to radians θ ij This is to calculate the characteristic value θ of the formation trend in the subsequent trigonometric calculation and the inverse tangent function. i It is more convenient and accurate when using trigonometric functions because the calculation rules of trigonometric functions in radians are simpler and more unified.

[0080] The calculation formula of the characteristic value of the formation trend is:

[0081]

[0082] In the formula, θ i It is expressed as the characteristic value of the stratigraphic trend of the i-th grid cell under the arc value.

[0083] Then the radian value θ i Converted into angle values ​​as characteristic values ​​of stratum direction:

[0084]

[0085] In the formula, α i It is expressed as the characteristic value of the stratigraphic direction of the i-th grid unit, and finally converted into an angle value α i For intuitive understanding.

[0086] The formation dip angle of the jth measurement point in the i-th grid cell is expressed as β ij , first convert to radians:

[0087]

[0088] In the formula, It is expressed as the radian value of the formation dip angle at the jth measurement point in the i-th grid cell, and then calculated and Stratigraphic dip refers to the direction of the inclination of the stratum layer. It is perpendicular to the stratum strike and together describes the inclination state of the stratum in space. The stratum dip angle β ij Convert to radians This is also to facilitate subsequent trigonometric operations.

[0089] The calculation formula of the characteristic value of formation dip is:

[0090]

[0091] In the formula, It is expressed as the characteristic value of the stratigraphic dip of the ith grid cell under the arc value,

[0092] Then convert it into an angle value as the characteristic value of the formation dip:

[0093]

[0094] The formation dip angle of the jth measuring point in the i-th grid cell is expressed as γ ij , then the calculation formula of the characteristic value of the formation dip is:

[0095]

[0096] In the formula, γ i It is expressed as the characteristic value of the formation dip of the i-th grid unit. The formation dip refers to the acute angle between the formation layer and the horizontal plane, which directly reflects the degree of inclination of the formation in the vertical direction. ij The characteristic value γ of the formation dip is obtained by arithmetic mean calculation i , this value can represent the overall vertical inclination of the stratum within the grid unit.

[0097] For structural characteristic data, check whether there are faults in the grid unit. If so, calculate the characteristic values ​​of the fault strike, dip, inclination and fault distance.

[0098] The fault strike angle of the jth measurement point in the i-th grid cell is set to α ′ ij , first set the angle value α ij Convert to radian value for calculation. The converted radian value is:

[0099]

[0100] In the formula, θ ′ ij It is expressed as the arc value of the fault strike angle at the jth measurement point in the i-th grid cell. Then x is calculated. ′ ij = cosθ ij and ′ ij = sinθ ij The fault strike refers to the direction in which the fault extends on the ground or underground horizontal plane, which reflects the fault's extension trend in the horizontal direction. Similar to the calculation of the formation strike, first set the angle value α ′ ij Convert to radians θ ′ ij This is for the subsequent trigonometric calculations and the calculation of the characteristic value θ of the fault trend using the inverse tangent function. ′ i It is more convenient and accurate.

[0101] The calculation formula of the characteristic value of the fault trend is:

[0102]

[0103] In the formula, θ ′ i It is expressed as the characteristic value of the fault trend of the ith grid cell under the arc value.

[0104] Then the radian value θ ′ i Convert to angle value as characteristic value of fault direction:

[0105]

[0106] In the formula, α ′ i It is expressed as the characteristic value of the fault direction of the i-th grid unit, and finally converted into an angle value α ′ i , for intuitive understanding.

[0107] The fault dip angle of the jth measuring point in the i-th grid cell is set to β ′ ij , first convert to radians:

[0108]

[0109] In the formula, It is expressed as the arc value of the fault dip angle at the jth measuring point in the i-th grid cell, and then calculated and Fault dip refers to the direction of the fault plane inclination, which is perpendicular to the fault strike and together describes the inclination state of the fault plane in space. ′ ij Convert to radians Also to facilitate subsequent trigonometric operations,

[0110] The calculation formula of the characteristic value of fault dip is:

[0111]

[0112] In the formula, It is expressed as the characteristic value of the fault dip of the ith grid cell in radians,

[0113] Then convert it into an angle value as the characteristic value of the fault dip:

[0114]

[0115] In the formula, β ′ i The characteristic value of the fault dip of the i-th grid cell is converted into an angle value β ′ i , for intuitive understanding.

[0116] The fault dip angle of the jth measurement point in the i-th grid cell is denoted as γ ′ ij , then the calculation formula of the characteristic value of the fault dip is:

[0117]

[0118] In the formula, γ ′ i It is expressed as the characteristic value of the fault dip of the ith grid unit. The fault dip refers to the acute angle between the fault plane and the horizontal plane, which directly reflects the degree of inclination of the fault plane in the vertical direction. ′ ij The characteristic value γ of the fault dip is obtained by arithmetic mean calculation ′i , this value can represent the overall inclination of the fault plane in the vertical direction within the grid unit.

[0119] Assume that the fault offset of the jth measuring point in the i-th grid cell is expressed as η ij , then the calculation formula of the characteristic value of the fault offset is:

[0120]

[0121] In the formula, A i It is expressed as the characteristic value of the fault throw of the i-th grid unit. The fault throw is a key parameter to describe the fault characteristics, which can accurately record the relative displacement distance between the two plates of the fault at different positions.

[0122] Identify all possible rock types and create unique codes for each of them. k , where k = 1, 2, 3, ..., f, and f represents the total number of all different rock types that may appear in the i-th grid unit, forming a rock type code list, and setting the distribution area of ​​the k-th rock type in the i-th grid unit to A ik , the total area is S i , then the area ratio of the kth rock type in the grid unit is expressed as:

[0123]

[0124] In the formula, p ik It is expressed as the area proportion of the kth rock type in the i-th grid unit, A ik It is expressed as the distribution area of ​​the kth rock type in the ith grid unit, S i It is expressed as the total area of ​​the i-th grid cell.

[0125] In each grid cell, the area proportions of different rock types are compared. ik , extract the rock type with the highest proportion, and use the code of the rock type as the rock type coding characteristic value of this grid unit. This is because there may be multiple rock types in a grid unit, but their influence on geological processes and mineralization is often different. By comparing the area proportions, the dominant rock type in the grid unit can be identified. This dominant rock type usually plays a key role in the composition and properties of the geological body.

[0126] Identify and code all possible rock types to facilitate the distinction and processing of different rock types in subsequent data analysis and model building. Calculate the area proportion p of each rock type in the grid cell ik, we can understand the distribution of different rock types in the area, that is, which rock type is dominant in this grid unit, which is of great significance for inferring the relationship between mineral formation and rock type.

[0127] The thickness of the stratum is one of the important basic data for estimating the resource volume;

[0128] As for the strike of the strata, it describes the orientation of the strata on the horizontal plane, helps to determine the distribution trend of the strata in space, and is an important parameter for constructing geological maps and geological models.

[0129] The dip of the strata, that is, the direction of inclination, not only helps to understand the mineralization process and infer the location, range and shape of hidden ore bodies, but also helps to reduce the difficulty of mining when formulating subsequent underground mining plans.

[0130] The dip angle of the stratum, that is, the angle between the stratum surface and the horizontal plane, has an important influence on the shape and distribution of the ore body and helps to judge the layered ore body.

[0131] The extent of the distribution of different rock types within a mining area may be associated with specific types of mineral deposits, helping to determine the presence of potential mineralized zones or deposits.

[0132] The description of the types of minerals in the ore and their relative contents helps geological prospectors understand the geological characteristics, mineral distribution and mineralization of the mining area.

[0133] The name of the fault is an identification of a specific fault, which facilitates the subsequent rapid location and understanding of the fault situation.

[0134] As for the location of the fault, the distribution of the fault in space is determined in the three-dimensional geological model, which provides key positioning information for constructing accurate three-dimensional geological structure.

[0135] The strike of the fault describes the extension direction of the fault plane on the horizontal plane, which is used to analyze the impact direction of the fault on the continuity of the formation and its control over the migration path of the ore-forming fluid on the horizontal plane.

[0136] For the dip of the fault, the inclination direction of the fault plane is clarified. Faults with different dips have different dislocation modes and results on the ore body, which directly affects the position, shape and continuity of the ore body in three-dimensional space.

[0137] The dip angle of the fault reflects the degree of inclination of the fault plane and has an important influence on the shape and distribution of the ore body.

[0138] The fault distance parameter is the relative movement distance between the two sides of the fault, including vertical fault distance and horizontal fault distance, which is of key significance for determining the true distribution range, shape and resource assessment of the ore body.

[0139] Step 2: Take whether there is a mine as the target variable, associate the target variable with the corresponding stratigraphic characteristic value, rock characteristic value and structural characteristic value to form a set of geological characteristic data, calibrate the set of all geological characteristic data as model data, use the model data as the training set, build a random forest model, set the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes, and then use the accuracy rate to evaluate and optimize the random forest model;

[0140] Data were collected from multiple explored mining areas. After screening and sorting, 1,000 samples were identified, including 500 mineralized samples and 500 non-mineralized samples. Each sample was divided into a three-dimensional grid with a grid size set to 5 meters.

[0141] For mineralized samples, a random number generator is used to generate 500 non-repeating random numbers in the range of 1 to 1000. For non-mineralized samples, a random number generator is also used to generate 500 non-repeating random numbers in the range of 1 to 1000.

[0142] Construct a random forest model. Random forest is an integrated learning algorithm that consists of multiple decision trees. It can integrate the prediction results of multiple decision trees and reduce the errors caused by overfitting or data bias in a single decision tree. First, the maximum depth is initially set to 8. After preliminary analysis of geological data characteristics and some empirical judgments, it is believed that the relationship between geological factors such as stratigraphic structure, rock type and structural characteristics and mineralization is not extremely simple and direct. A decision tree of a certain depth is needed to mine the complex associations, but it cannot be too deep to prevent overfitting. 10 is an initial attempt to balance learning ability and generalization ability. The minimum number of samples required for node splitting is set to 8. This is to prevent the decision tree from over-splitting when there is less data, resulting in an overly complex and unstable model. The minimum number of samples for leaf nodes is set to 5 to ensure that leaf nodes have enough samples to support their prediction results, making the prediction more reliable and representative.

[0143] In each round of random forest model training, the sample data in the training set is input into the current random forest model for prediction. The number of correctly predicted mineralized samples in the training set is TP, the number of correctly predicted non-mineralized samples is TN, and the total number of mineralized and non-mineralized samples is e. After each round of training, the accuracy is calculated according to the formula:

[0144]

[0145] In the formula, Accuracy represents the accuracy rate, e represents the total number of mineralized samples and non-mineralized samples, TP represents the correctly predicted number of mineralized samples, and TN represents the correctly predicted number of non-mineralized samples. When the accuracy rate is higher than 90%, the training is stopped.

[0146] Start training the model. After each round of training, use the training set to evaluate the model and calculate the overall prediction accuracy of the model for the mineralized samples and non-mineralized samples in the training set. For example, after the first round of training, the model predicts 1,000 samples in the training set, of which 400 samples are correctly predicted to be mineralized and 420 samples are correctly predicted to be non-mineralized. The accuracy is After multiple rounds of training, after the 20th round of training, the model predicted 1,000 samples in the training set, of which 460 samples were correctly predicted to be mineralized and 450 samples were correctly predicted to be non-mineralized, so the accuracy is Since the accuracy rate is higher than 90%, the training is stopped and a trained random forest model is obtained for subsequent hidden mine prediction work.

[0147] Step 3: By querying the geological catalog, extracting the geological characteristic data of the area to be detected, saving it as a geological characteristic data table, and importing it into the geological modeling software to build a three-dimensional geological model, the geological characteristic data includes stratum structure, rock type, and structural characteristic data;

[0148] By querying the geological catalog of the target area to be tested, data related to the stratigraphic structure are extracted, including the name, thickness value, strike, dip and inclination of each stratum. For rock types, the specific type of each rock is determined according to the geological identification standards and the description in the catalog, and encoded according to the pre-set classification and coding system for subsequent data processing and model recognition. For structural characteristics, including the strike, dip, dip and fault distance of the fault, these extracted stratigraphic structure, rock type and structural characteristic data are organized into a geological characteristic data table. The names of the stratigraphic layers generally correspond to specific rock types, fossil assemblages and geological structures, and some names are globally used stratigraphic names, such as the Cambrian and Ordovician, which facilitate subsequent rapid positioning and understanding of the stratigraphic conditions and help to infer the formation of minerals.

[0149] Then, the Surpac geological modeling software imports the organized stratigraphic data table into the software. Surpac geological modeling software has powerful functions and professional algorithms, which can convert two-dimensional geological data into three-dimensional visual models. It can accurately construct the morphology and mutual relationship of strata, rocks and structures in three-dimensional space based on the imported data, using mathematical models and graphic rendering technology. In the prediction of hidden mines, three-dimensional geological models can help researchers better understand the spatial coupling relationship between geological structure and mineralization, so as to more accurately determine the possible location and scope of hidden mines.

[0150] In the software environment, the stratigraphic model is constructed in sequence according to the stratigraphic names and thickness data in the stratigraphic data table. For each stratigraphic layer, the position and shape of the stratigraphic layer in three-dimensional space are set according to its strike, dip and inclination parameters, so that the stratigraphic model can truly reflect the actual distribution of the underground stratigraphic layer. For rock type data, the material assignment function of the software is used to assign corresponding colors or materials to different stratigraphic layers or geological bodies according to the pre-set correspondence between rock types and material colors. In this way, the distribution characteristics of rock types can be intuitively observed in the three-dimensional model, such as setting sandstone stratigraphic layers to yellow materials and granite stratigraphic layers to gray materials. For fault information in the structural feature data, a fault object is created in the software, and the fault is drawn in the three-dimensional model according to the strike, dip, dip and distance parameters of the fault.

[0151] The stratigraphic structure provides a spatial framework for the occurrence of hidden minerals. Different stratigraphic thickness, strike, dip and inclination will affect the migration path of ore-forming fluids and the location of mineral precipitation. Rock types have a close material connection with mineralization. Some rock types may be rich in specific ore-forming elements or have physical and chemical properties that are conducive to the enrichment of ore-forming elements. Structural features, especially faults, are often important channels for ore-forming fluids and favorable places for mineral precipitation. The activity of faults can change the permeability and pressure state of the strata and promote the migration and aggregation of ore-forming elements. Integrating these three into a three-dimensional geological model can comprehensively reflect the complex relationship between the geological environment and hidden minerals, and provide complete geological data for model-based prediction of hidden minerals.

[0152] Step 4: Divide the constructed three-dimensional geological model into cube grids, input the geological characteristic data corresponding to each grid unit into the optimized random forest model according to the geological characteristic data table, and use the optimized random forest model to predict the grid units where hidden mines are likely to exist, so as to realize the location prediction of hidden mines in three-dimensional space;

[0153] The grid division function is activated in the software, where the grid is set to 5 meters. The software will divide the spatial range of the entire three-dimensional geological model into several cubic grid units with a side length of 5 meters according to the set parameters, so that each grid unit can become a relatively independent and representative geological sample unit.

[0154] From the geological characteristic data table, obtain the geological characteristic data corresponding to each grid unit. For the stratigraphic structure data, calculate the average value of the stratigraphic thickness at each point in the grid unit as the characteristic value of the stratigraphic thickness of the unit. Calculate the average value of the stratigraphic dip as the characteristic value of the dip. For rock type data, determine the rock type coding characteristic value based on the distribution area ratio of the rock type in the grid unit or the main rock type. For structural characteristic data, detect whether there is a fault in the grid unit. If so, calculate the characteristic values ​​of the fault strike, dip, dip and fault distance.

[0155] Assume that the formation thickness at the jth measurement point in the i-th grid cell is expressed as h ij , where the total number of measurement points in a grid cell is n, i is used to distinguish different grid cells, each of which has its own independent set of measurement point data and the corresponding characteristic value calculation process of formation thickness, and j is a variable that numbers each measurement point in the i-th grid cell;

[0156] The calculation formula of the characteristic value of the formation thickness is:

[0157]

[0158] The formation strike angle of the jth measurement point in the i-th grid cell is set to α ij , first set the angle value α ij Convert to radian value for calculation. The converted radian value is:

[0159]

[0160] In the formula, θ ij It is expressed as the radian value of the stratigraphic strike angle of the jth measurement point in the i-th grid cell. Then x is calculated. ij = cosθ ij and ij = sinθ ij .

[0161] The calculation formula of the characteristic value of the formation trend is:

[0162]

[0163] In the formula, θ i It is expressed as the characteristic value of the stratigraphic trend of the i-th grid cell under the arc value.

[0164] Then the radian value θ i Converted into angle values ​​as characteristic values ​​of stratum direction:

[0165]

[0166] In the formula, α i is represented as the characteristic value of the stratigraphic trend of the i-th grid cell,

[0167] The formation dip angle of the jth measuring point in the i-th grid cell is set to β ij , first convert to radians:

[0168]

[0169] In the formula, It is expressed as the radian value of the formation dip angle at the jth measurement point in the i-th grid cell, and then calculated and

[0170] The calculation formula of the characteristic value of formation dip is:

[0171]

[0172] In the formula, It is expressed as the characteristic value of the stratigraphic dip of the ith grid cell under the arc value,

[0173] Then convert it into an angle value as the characteristic value of the formation dip:

[0174]

[0175] The formation dip angle of the jth measurement point in the i-th grid cell is expressed as γ oj , then the calculation formula of the characteristic value of the formation dip is:

[0176]

[0177] In the formula, γ i is represented as the characteristic value of the formation dip of the i-th grid cell,

[0178] For structural characteristic data, check whether there are faults in the grid unit. If so, calculate the characteristic values ​​of the fault strike, dip, inclination and fault distance.

[0179] The fault strike angle of the jth measurement point in the i-th grid cell is set to α ′ ij , first set the angle value α ij Convert to radian value for calculation. The converted radian value is:

[0180]

[0181] In the formula, θ ′ ij It is expressed as the arc value of the fault strike angle at the jth measurement point in the i-th grid cell. Then x is calculated. ′ ij = cosθ ij and ′ ij = sinθ ij ,

[0182] The calculation formula of the characteristic value of the fault trend is:

[0183]

[0184] In the formula, θ ′ i It is expressed as the characteristic value of the fault trend of the ith grid cell under the arc value.

[0185] Then the radian value θ ′ i Convert to angle value as characteristic value of fault direction:

[0186]

[0187] In the formula, α ′ i is represented as the characteristic value of the fault strike of the ith grid cell,

[0188] The fault dip angle of the jth measuring point in the i-th grid cell is set to β ′ ij , first convert to radians:

[0189]

[0190] In the formula, It is expressed as the arc value of the fault dip angle at the jth measuring point in the i-th grid cell, and then calculated and

[0191] The calculation formula of the characteristic value of fault dip is:

[0192]

[0193] In the formula, It is expressed as the characteristic value of the fault dip of the ith grid cell in radians,

[0194] Then convert it into an angle value as the characteristic value of the fault dip:

[0195]

[0196] In the formula, β ′ i The characteristic value of the fault dip of the i-th grid cell is converted into an angle value β ′ i , for intuitive understanding.

[0197] The fault dip angle of the jth measurement point in the i-th grid cell is denoted as γ ′ ij , then the calculation formula of the characteristic value of the fault dip is:

[0198]

[0199] In the formula, γ ′ i is represented as the characteristic value of the fault dip of the ith grid cell,

[0200] Assume that the fault offset of the jth measuring point in the i-th grid cell is expressed as η ij , then the calculation formula of the characteristic value of the fault offset is:

[0201]

[0202] In the formula, A i is represented as the characteristic value of the fault offset of the ith grid cell,

[0203] Identify all possible rock types and create unique codes for each of them. k , where k = 1, 2, 3, ..., f, and f represents the total number of all different rock types that may appear in the i-th grid unit, forming a rock type code list, and setting the distribution area of ​​the k-th rock type in the i-th grid unit to A ik , the total area is S i , then the area ratio of the kth rock type in the grid unit is expressed as:

[0204]

[0205] In the formula, p ik It is expressed as the area proportion of the kth rock type in the i-th grid unit, A ik It is expressed as the distribution area of ​​the kth rock type in the ith grid unit, S i It is expressed as the total area of ​​the i-th grid cell.

[0206] In each grid cell, the area proportions of different rock types are compared. ik , extract the rock type with the highest proportion, and use the code of the rock type as the rock type coding feature value of this grid unit.

[0207] After the geological feature data of all grid cells are sorted in the prescribed format, they are input into the trained random forest model. The model will predict the possibility of the existence of hidden mines in each grid cell based on the complex relationship between geological features and mineralization learned in the training phase. These feature data undergo a series of complex calculations and decision-making processes within the model, and finally output a probability value indicating the possibility of the existence of hidden mines in the grid cell.

[0208] When more than 60% of the decision trees judge that the grid cell is yes, the grid cell is marked as a grid cell with a hidden mine. By sorting and analyzing the prediction results of all grid cells, those areas judged by the model to have hidden mines can be located in three-dimensional space.

[0209] See also Figure 2 The present invention also provides a three-dimensional quantitative prediction system for hidden mines, the system is used to execute the above-mentioned three-dimensional quantitative prediction method for hidden mines, comprising:

[0210] The feature extraction module is used to collect an equal number of mineralized and non-mineralized model samples, divide all model samples into grid units in three dimensions, take the stratigraphic thickness, strike, dip and inclination of the stratigraphic structure of each grid unit as stratigraphic characteristic values, classify and encode rock types as rock characteristic values, and take the fault strike, dip, dip and fault throw of structural characteristics as structural characteristic values;

[0211] The model building module is used to take whether there is a mine as the target variable, associate the target variable with the corresponding stratum characteristic value, rock characteristic value and structural characteristic value to form a set of geological characteristic data, calibrate the collection of all geological characteristic data as model data, use the model data as a training set, build a random forest model, set the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes, and then use the accuracy rate to evaluate and optimize the random forest model;

[0212] A three-dimensional modeling module is used to extract geological characteristic data of the area to be detected by querying geological catalogs, save them as geological characteristic data tables, and import them into geological modeling software to construct a three-dimensional geological model. The geological characteristic data include stratum structure, rock type, and structural characteristic data.

[0213] The hidden mine positioning module is used to divide the constructed three-dimensional geological model into cubic grids, and input the geological characteristic data corresponding to each grid unit into the optimized random forest model according to the geological characteristic data table. The optimized random forest model predicts the grid units where hidden mines are likely to exist, thereby realizing the positioning prediction of hidden mines in three-dimensional space.

[0214] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0215] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0216] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, and may be located in one place or distributed over multiple grid units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0217] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A three-dimensional quantitative prediction method for concealed mines, characterized in that: The specific steps include: Step 1: Collect an equal number of mineralized and non-mineralized model samples, divide all model samples into grid units in three dimensions, take the thickness, strike, dip and inclination of the stratigraphic system of each grid unit as stratigraphic characteristic values, classify and code the rock types as rock characteristic values, and take the fault strike, dip, dip and fault throw of structural characteristics as structural characteristic values; Step 2: Take whether there is a mine as the target variable, associate the target variable with the corresponding stratigraphic characteristic value, rock characteristic value and structural characteristic value to form a set of geological characteristic data, calibrate the set of all geological characteristic data as model data, use the model data as the training set, build a random forest model, set the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes, and then use the accuracy rate to evaluate and optimize the random forest model; Step 3: By querying the geological catalog, extracting the geological characteristic data of the area to be detected, saving it as a geological characteristic data table, and importing it into the geological modeling software to build a three-dimensional geological model. The geological characteristic data includes stratigraphic system, rock type, and structural characteristic data; Step 4: Divide the constructed three-dimensional geological model into cubic grids, and input the geological characteristic data corresponding to each grid unit into the optimized random forest model according to the geological characteristic data table. The optimized random forest model predicts the grid units where hidden mines are likely to exist, thereby realizing the positioning prediction of hidden mines in three-dimensional space.

2. A three-dimensional quantitative prediction method for concealed mines according to claim 1, characterized in that: In step 1, the thickness, strike, dip and inclination of the stratigraphic system of each grid unit are used as stratigraphic characteristic values, the rock types are classified and coded as rock characteristic values, and the fault strike, dip, dip and fault throw of structural characteristics are used as structural characteristic values ​​as follows: First, the collected mineralized and non-mineralized model samples are gridded, and then the stratum thickness of the jth measurement point in the i-th grid unit is set as h ij , where the total number of measurement points in a grid cell is n, i is used to distinguish different grid cells, each of which has its own independent set of measurement point data and the corresponding characteristic value calculation process of formation thickness, and j is a variable that numbers each measurement point in the i-th grid cell; The calculation formula of the characteristic value of the formation thickness is: The stratigraphic strike angle of the jth measurement point in the i-th grid cell is set to α ij , first set the angle value α ij Convert to radian value for calculation. The converted radian value is: In the formula, θ ij It is expressed as the radian value of the stratigraphic strike angle of the jth measurement point in the i-th grid cell. Then x is calculated. ij = cosθ ij and ij = sinθ ij , The calculation formula of the characteristic value of the formation trend is: In the formula, θ i Expressed as the characteristic value of the stratigraphic trend of the ith grid cell under the arc value Then the radian value θ i Converted into angle values ​​as characteristic values ​​of stratum direction: In the formula, α i is represented as the characteristic value of the stratigraphic trend of the i-th grid cell, The formation dip angle of the jth measurement point in the i-th grid cell is expressed as β ij , first convert to radians: In the formula, It is expressed as the radian value of the formation dip angle at the jth measurement point in the i-th grid cell, and then calculated and The calculation formula of the characteristic value of formation dip is: In the formula, It is expressed as the characteristic value of the stratigraphic dip of the ith grid cell under the arc value; Then convert it into an angle value as the characteristic value of the formation dip: The formation dip angle of the jth measurement point in the i-th grid cell is expressed as γ ij , then the calculation formula of the characteristic value of the formation dip is: In the formula, γ i is represented as the characteristic value of the formation dip of the i-th grid cell, Identify all possible rock types and create unique codes for each of them. k , where k = 1, 2, 3, ..., f, and f represents the total number of all different rock types that may appear in the i-th grid unit, forming a rock type code list, and setting the distribution area of ​​the k-th rock type in the i-th grid unit to A ik , the total area is S i , then the area ratio of the kth rock type in the grid unit is expressed as: In the formula, p ik It is expressed as the area proportion of the kth rock type in the i-th grid unit, A ik It is expressed as the distribution area of ​​the kth rock type in the ith grid unit, S i is expressed as the total area of ​​the ith grid cell, In each grid cell, the area proportions of different rock types are compared. ik , extract the rock type with the highest proportion, and use the code of the rock type as the rock type coding feature value of this grid unit.

3. A three-dimensional quantitative prediction method for hidden mines according to claim 1, characterized in that: In step 2, using the model data as a training set includes the following steps: Whether or not there is a mineralization is used as the stratification variable. The total number of mineralized samples and non-mineralized samples is e. For all mineralized samples, a random number generator is used to generate a random number in the range of 1 to e. Non-repeating random numbers are added to the mineralized samples with corresponding numbers in the training set. For non-mineralized samples, a random number generator is used to generate Non-repeating random numbers are added to all non-mineralized samples with corresponding numbers in the training set. When the overall accurate prediction ratio of the model for mineralized samples and non-mineralized samples in the training set, that is, the accuracy index, is higher than 90%, the training is stopped.

4. A three-dimensional quantitative prediction method for hidden mines according to claim 1, characterized in that: In step 2, setting the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes includes the following steps: The maximum depth was initially set to 8, and the random forest model was constructed using the divided training set. The minimum number of samples required for node splitting was set to 8, and the minimum number of samples for leaf nodes was set to 5.

5. The three-dimensional quantitative prediction method for hidden mines according to claim 1 is characterized in that: In step 2, the method of using accuracy to evaluate and optimize the random forest model is: In each round of random forest model training, the sample data in the training set is input into the current random forest model for prediction. The number of correctly predicted mineralized samples in the training set is TP, the number of correctly predicted non-mineralized samples is TN, and the total number of mineralized and non-mineralized samples is e. After each round of training, the accuracy is calculated according to the formula: In the formula, Accuracy is expressed as the accuracy rate, e is expressed as the total number of mineralized samples and non-mineralized samples, TP is expressed as the correctly predicted number of mineralized samples, and TN is expressed as the correctly predicted number of non-mineralized samples; When the accuracy is higher than 90%, training stops.

6. A three-dimensional quantitative prediction method for concealed mines according to claim 1, characterized in that: In step 3, the method for constructing a three-dimensional geological model is: The name, thickness, strike, dip and inclination of each layer in the stratigraphic system data are saved in a unified text format as a geological characteristic data table, and then imported into the geological modeling software. Different rock types are distinguished by different colors or materials, so that the model can visually reflect the distribution characteristics of rock types. The structural characteristic data is integrated into the model to show the position and shape of the fault in three-dimensional space and its intersection with the stratum, so as to construct a complete three-dimensional geological model including stratigraphic structure, rock type and structural characteristics.

7. The three-dimensional quantitative prediction method for hidden mines according to claim 1 is characterized in that: In step 4, the location prediction of the hidden mine in three-dimensional space includes the following steps: From the constructed three-dimensional model of the area to be detected, feature data are extracted in the same way as when the model was trained, including stratum thickness, strike, dip, inclination, rock type code, and strike, dip, inclination and fault throw of the fault. The three-dimensional space of the area to be detected is divided into cubic grids, and each grid unit is regarded as an independent sample point. All sample data are input into the trained random forest model. According to the input feature data, the relationship pattern between geological characteristics and mineralization learned in the training process is used. After the model prediction result is output, a probability threshold μ is set. When the probability value of the existence of a hidden mine corresponding to the prediction result of a certain grid unit is higher than the probability threshold μ, it is determined that there is a hidden mine at the location of this grid unit; otherwise, if the probability value does not reach the probability threshold μ, it is determined that there is no hidden mine at this location.

8. A three-dimensional quantitative prediction system for concealed mines, characterized by: The system is used to execute a three-dimensional quantitative prediction method for concealed mines according to any one of claims 1 to 7, comprising: The feature extraction module is used to collect an equal number of mineralized and non-mineralized model samples, divide all model samples into grid units in three dimensions, take the thickness, strike, dip and inclination of the stratigraphic system of each grid unit as stratigraphic characteristic values, classify and encode rock types as rock characteristic values, and take the fault strike, dip, dip and fault throw of structural characteristics as structural characteristic values; The model building module is used to take whether there is a mine as the target variable, associate the target variable with the corresponding stratum characteristic value, rock characteristic value and structural characteristic value to form a set of geological characteristic data, calibrate the collection of all geological characteristic data as model data, use the model data as a training set, build a random forest model, set the maximum depth, the minimum number of samples required for node splitting, and the minimum number of samples for leaf nodes, and then use the accuracy rate to evaluate and optimize the random forest model; A three-dimensional modeling module is used to extract geological characteristic data of the area to be detected by querying geological catalogs, save them as geological characteristic data tables, and import them into geological modeling software to construct a three-dimensional geological model. The geological characteristic data include stratum structure, rock type, and structural characteristic data. The hidden mine positioning module is used to divide the constructed three-dimensional geological model into cubic grids, and input the geological characteristic data corresponding to each grid unit into the optimized random forest model according to the geological characteristic data table. The optimized random forest model predicts the grid units where hidden mines are likely to exist, thereby realizing the positioning prediction of hidden mines in three-dimensional space.

Images