Three-dimensional prediction method for gold mineralization based on multi-source data fusion

Abstract

The application relates to the technical field of mineral resource exploration, in particular to a gold deposit 3D prediction method based on multi-source data fusion, which comprises the following steps: obtaining geological attribute information of multiple grid units in a study area, ore-controlling elements of multiple preset ore-forming stages, and stage attribution and positions of multiple known ore spots; determining the associated geological data of each grid unit under each preset ore-forming stage according to the geological attribute information and the ore-controlling elements; determining the single-stage ore-forming probability of each grid unit under each preset ore-forming stage according to the positions of the multiple known ore spots and the associated geological data of each grid unit under each preset ore-forming stage; performing weighted fusion on the single-stage ore-forming probability of each grid unit under each preset ore-forming stage to obtain the comprehensive ore-forming probability of each grid unit; and determining the gold deposit prediction result according to the comprehensive ore-forming probability. The method can improve the accuracy and reliability of the ore-forming probability prediction.

Description

Technical Field

[0001] This application relates to the field of mineral resource exploration technology, specifically to a three-dimensional prediction method for gold mineralization based on multi-source data fusion. Background Technology

[0002] With the continuous development of gold resources in my country, easily identifiable shallow ore bodies have been gradually depleted, and gold exploration is gradually expanding to deep, concealed deposits. Deep metallogenic environments are complex, metallogenic information is weak, and is controlled by multiple geological factors. Therefore, three-dimensional metallogenic prediction technology has become an important tool supporting the exploration of deep gold resources.

[0003] Currently, existing three-dimensional mineralization prediction methods are usually based on multi-source geological data to establish a three-dimensional geological model, extract the multi-dimensional geological features of each grid cell in the three-dimensional geological model, and input them into machine learning models such as neural networks, random forests or support vector machines for training and mineralization probability prediction.

[0004] However, gold mineralization often involves multiple overlapping phases, and the properties of ore-forming fluids, ore-controlling factors, and mineralization types vary significantly across different phases. The aforementioned methods input all features into the machine learning model, and the favorable signals from different phases interfere with each other in the feature space, resulting in low accuracy and reliability of mineralization probability prediction, making it difficult to effectively meet the precision requirements of deep gold resource exploration. Summary of the Invention

[0005] To address the low accuracy and reliability of existing mineralization probability prediction technologies, this application aims to provide a three-dimensional gold mineralization prediction method based on multi-source data fusion. The specific technical solution adopted is as follows: This application provides a three-dimensional prediction method for gold mineralization based on multi-source data fusion, comprising: acquiring geological attribute information of multiple grid units within a study area, ore-controlling elements of multiple preset mineralization periods, and the period affiliation and location of multiple known mineral deposits; determining the associated geological data of each grid unit under each preset mineralization period based on the geological attribute information of each grid unit and the ore-controlling elements of each preset mineralization period; determining the single-period mineralization probability of each grid unit under each preset mineralization period based on the location of multiple known mineral deposits and the associated geological data of each grid unit under each preset mineralization period; weighting and fusing the single-period mineralization probabilities of each grid unit under each preset mineralization period based on the period affiliation of multiple known mineral deposits to obtain the comprehensive mineralization probability of each grid unit; and determining the gold mineralization prediction result of the study area based on the comprehensive mineralization probability of each grid unit.

[0006] Optionally, the geological attribute information of a grid cell includes the content values ​​of various ore-forming elements, the distance from each fault, the rock mass type, and the stratigraphic number. The ore-controlling elements of a preset metallogenic period include fault identification, ore-controlling rock mass type, favorable ore-bearing strata number, and ore-forming element combination. Based on the geological attribute information of each grid cell and the ore-controlling elements of each preset metallogenic period, the associated geological data of each grid cell under each preset metallogenic period is determined, including: determining the element combination score of each grid cell under each preset metallogenic period based on the content values ​​of various ore-forming elements included in the geological attribute information of each grid cell and the ore-forming element combination of each preset metallogenic period; and based on the grid cell... The system includes the rock mass types and the ore-controlling rock mass types of each preset metallogenic period. It determines the rock mass period affiliation of each grid unit under each preset metallogenic period; based on the stratigraphic number of the strata within the grid unit and the favorable ore-bearing strata number of each preset metallogenic period, it determines the stratigraphic period affiliation of each grid unit under each preset metallogenic period; based on the distance of each grid unit from each fault and the fault identifier of each preset metallogenic period, it determines the fault distance of each grid unit from each preset metallogenic period; and it uses the element combination score, rock mass period affiliation, stratigraphic period affiliation, and fault distance from each preset metallogenic period of each grid unit as the associated geological data for each grid unit under each preset metallogenic period.

[0007] Optionally, the above-mentioned determination of the single-phase mineralization probability of each grid unit under each preset mineralization period based on the location of multiple known mineral deposits and the associated geological data of each grid unit under each preset mineralization period includes: determining the mineralization contribution of each preset mineralization period to each grid unit based on the associated geological data of each grid unit under each preset mineralization period; and determining the single-phase mineralization probability of each grid unit under each preset mineralization period based on the mineralization contribution of each preset mineralization period to each grid unit, the associated geological data of each grid unit under each preset mineralization period, and the location of multiple known mineral deposits.

[0008] Optionally, the above-mentioned determination of the mineralization contribution of each preset mineralization period to each grid unit based on the associated geological data of each grid unit under each preset mineralization period includes: constructing a feature matrix with grid units as rows and associated geological data of each grid unit under each preset mineralization period as columns; performing non-negative matrix decomposition on the feature matrix to obtain the mineralization contribution of each preset mineralization period to each grid unit.

[0009] Optionally, the above-mentioned non-negative matrix decomposition of the feature matrix to obtain the mineralization contribution of each preset mineralization period to each grid cell includes: decomposing the feature matrix into the product of a base matrix and a coefficient matrix, wherein the rows of the base matrix are the same as the rows of the feature matrix, the columns of the base matrix are each preset mineralization period, the rows of the coefficient matrix are each preset mineralization period, and the columns of the coefficient matrix are the same as the columns of the feature matrix; constructing a loss function based on the element values ​​in the feature matrix, the element values ​​in the base matrix, the element values ​​in the coefficient matrix, the fault distance of each grid cell from each preset mineralization period, and the rock mass period attribution of each grid cell under each preset mineralization period; performing iterative solution with the goal of minimizing the loss function, and determining the mineralization contribution of each preset mineralization period to each grid cell based on the element values ​​of each row in the obtained base matrix.

[0010] Optionally, the above iterative solution with the goal of minimizing the loss function, and the element values ​​of each row in the obtained basis matrix as the mineralization contribution of each preset mineralization period to each grid cell, includes: using the gradient descent method to iteratively update the basis matrix and coefficient matrix with the goal of minimizing the loss function; when the rate of change of the loss function value in multiple consecutive iterations is less than a preset rate of change threshold, determining the mineralization contribution of each preset mineralization period to each grid cell based on the element values ​​of each row in the basis matrix obtained in the current iteration.

[0011] Optionally, the above-mentioned determination of the single-period mineralization probability of each grid unit under each preset mineralization period, based on the mineralization contribution of each grid unit to each grid unit under each preset mineralization period, the associated geological data of each grid unit under each preset mineralization period, and the location of multiple known mineral deposits, includes: determining the grid unit to which each known mineral deposit belongs based on the location of multiple known mineral deposits; determining the a priori mineralization probability of the study area as the ratio between the number of grid units with known mineral deposits and the total number of all grid units; using the data and mineralization contribution of each grid unit under each preset mineralization period as evidence factors, and determining the existence status of each evidence factor based on preset mineralization conditions; determining the evidence weight of each evidence factor based on the existence status of each evidence factor; and determining the single-period mineralization probability of each grid unit under each preset mineralization period based on the evidence weight of each evidence factor of each grid unit under each preset mineralization period and the a priori mineralization probability of the study area.

[0012] Optionally, the above-mentioned determination of the evidence weight of each evidence factor based on the existence state of each evidence factor includes: marking the grid cell to which the known mineral deposit belongs as a positive sample, and marking the grid cell that is spatially distant from the known mineral deposit and does not contain the known mineral deposit as a negative sample; determining the proportion of grid cells in the positive sample where the existence state of the same type of evidence factor under the same preset mineralization period is present as the first conditional probability of the existence state of the same type of evidence factor under the same preset mineralization period; determining the proportion of grid cells in the negative sample where the existence state of the same type of evidence factor under the same preset mineralization period is present as the second conditional probability of the existence state of the same type of evidence factor under the same preset mineralization period; and determining the evidence weight of the same type of evidence factor under the same preset mineralization period when it exists based on the ratio of the first conditional probability to the second conditional probability.

[0013] Optionally, determining the evidence weight of each evidence factor based on its existence status includes: determining the proportion of grid cells in the positive sample where the existence status of the same type of evidence factor under the same preset mineralization period is non-existent, as the third conditional probability of the same type of evidence factor under the same preset mineralization period when its existence status is non-existent; determining the proportion of grid cells in the negative sample where the existence status of the same type of evidence factor under the same preset mineralization period is non-existent, as the fourth conditional probability of the same type of evidence factor under the same preset mineralization period when its existence status is non-existent; and determining the evidence weight of the same type of evidence factor under the same preset mineralization period when its existence status is non-existent based on the ratio of the third conditional probability to the fourth conditional probability.

[0014] Optionally, the above-mentioned weighted fusion of the single-period mineralization probability of each grid unit under each preset mineralization period based on the period assignment of multiple known mineral deposits to obtain the comprehensive mineralization probability of each grid unit includes: determining the number of known mineral deposits in each preset mineralization period based on the period assignment of multiple known mineral deposits; determining the period weight of each preset mineralization period based on the proportion of the number of known mineral deposits in each preset mineralization period to the total number of known mineral deposits in all preset mineralization periods; and weighted summing the single-period mineralization probability of each grid unit under each preset mineralization period based on the period weight of each preset mineralization period to obtain the comprehensive mineralization probability of each grid unit.

[0015] This application has the following beneficial effects: By acquiring geological attribute information of grid cells, ore-controlling elements of preset mineralization periods, and the period affiliation and location of known mineral deposits, a deep correlation foundation was established between geological attribute information and the metallogenic regularity of preset mineralization periods. Undifferentiated geological attribute information was transformed into associated geological data closely related to each preset mineralization period, providing period-specific feature inputs for independent phased prediction. Based on this, by combining the location of known mineral deposits with associated geological data, the single-phase mineralization probability was determined for each preset mineralization period, enabling the metallogenic regularity of different preset mineralization periods to operate within an independent prediction framework, avoiding mutual interference and superposition / cancellation of signals from multiple periods. Finally, by weighting and fusing the single-phase mineralization probabilities according to the period affiliation of known mineral deposits, the comprehensive mineralization probability reflects the true contribution ratio of each preset mineralization period in the actual mineralization process, thereby ensuring a high degree of consistency between the prediction results and regional geological metallogenic regularity, significantly improving the accuracy and reliability of deep gold deposit mineralization prediction results. Attached Figure Description

[0016] Figure 1 A method flow for a three-dimensional prediction method for gold mineralization based on multi-source data fusion, provided as an embodiment of this application. Figure 1 ; Figure 2 A method flow for a three-dimensional prediction method for gold mineralization based on multi-source data fusion, provided as an embodiment of this application. Figure 2 . Detailed Implementation

[0017] The following description, in conjunction with the accompanying drawings, details a specific scheme for a three-dimensional prediction method for gold mineralization based on multi-source data fusion provided in this application.

[0018] Please see Figure 1 It illustrates the method flow of a three-dimensional prediction method for gold mineralization based on multi-source data fusion provided in an embodiment of this application. Figure 1 .

[0019] like Figure 1 As shown, the three-dimensional prediction method for gold mineralization based on multi-source data fusion includes S101-S105.

[0020] S101. Obtain geological attribute information of multiple grid units within the study area, ore-controlling elements of multiple preset mineralization periods, and the period affiliation and location of multiple known mineral deposits.

[0021] It should be understood that geological attribute information refers to the collection of multi-source geological data carried by each grid unit within the study area, used to characterize the geological features of each grid unit.

[0022] The geological attribute information of a grid cell includes the content values ​​of various mineralization-related elements, distances from each fault, rock mass type, and stratigraphic number. Mineralization-related elements include, but are not limited to, gold, silver, arsenic, antimony, copper, lead, zinc, tungsten, molybdenum, bismuth, and other elements closely related to gold mineralization. Distances from each fault refer to the spatial distance from the center point of the grid cell to each geological fault within the study area. Rock mass type refers to the type of magmatic rock exposed or inferred within the grid cell, such as granite porphyry or diorite porphyry, used to characterize the magmatic activity of that grid cell. Stratigraphic number refers to the stratigraphic code of the strata to which the grid cell belongs, used to distinguish different stratigraphic units.

[0023] Optionally, the geological attribute information of a grid cell may also include geophysical anomaly values. Geophysical anomaly values ​​refer to the geophysical anomaly attributes corresponding to the grid cell, including magnetic anomaly values, polarizability and resistivity values, which are used to characterize the physical field characteristics of deep geological bodies.

[0024] The preset mineralization periods are the main tectonic-magmatic activity periods pre-divided by geological experts based on regional geological maps, rock mass isotopic age data, and the period classification of known mineral deposits, and based on human experience. Each preset mineralization period corresponds to a specific tectonic movement type, igneous rock type, and known mineralization type.

[0025] For example, the multiple preset mineralization periods may include period T1 (Late Jurassic, 160-150 Ma), period T2 (Early Early Cretaceous, 140-130 Ma), and period T3 (Late Early Cretaceous, 120-110 Ma), where period T1 corresponds to the intrusion of intermediate-acidic rock bodies under a compressional setting, accompanied by high-temperature W-Mo mineralization; period T2 corresponds to the intrusion of granite porphyry under an extensional setting, accompanied by intermediate-temperature Au-Ag mineralization; and period T3 corresponds to the intrusion of diorite porphyry dikes under a extensional setting, accompanied by epithermal Au mineralization.

[0026] The ore-controlling elements of a pre-defined mineralization period are the set of geological ore-controlling elements corresponding to each pre-defined mineralization period. The ore-controlling elements of a pre-defined mineralization period include fault identification, ore-controlling rock mass type, favorable ore-bearing strata number, and ore-forming element assemblage. Specifically, the fault identification uniquely identifies the ore-controlling fault corresponding to the pre-defined mineralization period; the ore-controlling rock mass type refers to the typical magmatic rock type associated with the pre-defined mineralization period; the favorable ore-bearing strata number refers to the strata number preferred by the mineralization processes in the pre-defined mineralization period; and the ore-forming element assemblage refers to the typical element combination enriched by the ore-forming fluids in the pre-defined mineralization period, for example, a tungsten-molybdenum-bismuth assemblage for high-temperature periods and a gold-silver-arsenic-antimony assemblage for medium-temperature periods.

[0027] Known mineral deposits refer to discovered gold deposits. The mineralization period of a known mineral deposit refers to the pre-defined mineralization period to which the known mineral deposit belongs. The location of a known mineral deposit refers to its three-dimensional coordinates within the study area.

[0028] In one alternative implementation, multi-source geological data of the study area can be acquired first. After coordinate unification and preprocessing of the multi-source geological data, a three-dimensional geological entity model of the study area can be constructed. The three-dimensional geological entity model includes stratigraphic entities, fault surfaces, and rock mass entities. Then, the three-dimensional geological entity model is discretized into multiple grid units. Under normal scales, the unit size can be set to 10 meters by 10 meters by 10 meters. For key exploration areas, it can be densified to 5 meters by 5 meters by 5 meters by 5 meters. The multi-source geological data is mapped to the corresponding grid units to obtain the geological attribute information of each grid unit.

[0029] The multi-source geological data includes basic geological data, geochemical data, geophysical data, and data on known mineral deposits. Basic geological data includes regional geological maps, borehole logging data, fault measurement data, and rock mass isotope dating data; geochemical data includes elemental analysis data of rocks, soils, and borehole cores; and geophysical data includes high-precision magnetic survey data, induced polarization gradient measurement data, and resistivity measurement data.

[0030] It should be understood that coordinate unification refers to converting all data into a unified coordinate system, such as the CGCS2000 coordinate system, and the elevation datum adopts the 1985 National Elevation Datum. Preprocessing includes performing logarithmic transformation on geochemical data to eliminate the range effect of element content, extending geophysical data upward to eliminate shallow human interference, and using ordinary kriging interpolation to complete missing attribute values.

[0031] S102. Based on the geological attribute information of each grid unit and the ore-controlling elements of each preset mineralization period, determine the associated geological data of each grid unit under each preset mineralization period.

[0032] It should be understood that associated geological data refers to the set of data extracted from the geological attribute information of each grid unit for each preset metallogenic period, which is related to the metallogenic mechanism of that preset metallogenic period. Associated geological data is used to characterize the metallogenic geological environment characteristics of each grid unit under each preset metallogenic period.

[0033] In one optional implementation, the element combination score of each grid cell under each preset metallogenic period can be determined based on the content values ​​of various mineralization-related elements included in the geological attribute information of each grid cell and the combination of mineralization elements under each preset metallogenic period; the rock mass period affiliation of each grid cell under each preset metallogenic period can be determined based on the rock mass type included in the grid cell and the ore-controlling rock mass type under each preset metallogenic period; the stratigraphic period affiliation of each grid cell under each preset metallogenic period can be determined based on the stratigraphic number of the strata within the grid cell and the favorable ore-bearing strata number under each preset metallogenic period; the fault distance of each grid cell from each preset metallogenic period can be determined based on the distance of each grid cell from each fault and the fault identifier under each preset metallogenic period; and the element combination score, rock mass period affiliation, stratigraphic period affiliation, and fault distance of each grid cell under each preset metallogenic period can be determined as the associated geological data of each grid cell under each preset metallogenic period.

[0034] It should be understood that the element combination score refers to a quantitative indicator that reflects the geochemical characteristics of ore-forming fluids in a specific ore-forming period, extracted from the content values ​​of multiple ore-forming related elements through multivariate statistical methods.

[0035] Optionally, the content values ​​of multiple mineralization-related elements in all grid cells can be standardized first. Then, principal component analysis is performed to reduce the dimensionality based on the content values ​​of multiple mineralization-related elements in all grid cells to obtain the scores of multiple principal components. Each principal component corresponds to a combination of mineralization elements. Finally, based on the combination of mineralization elements in each preset mineralization period, the score of each grid cell on the corresponding principal component is determined as the element combination score of that grid cell in that preset mineralization period.

[0036] Understandably, the higher the element combination score, the closer the geochemical characteristics of the grid cell are to the typical ore-forming fluid characteristics of the preset ore-forming period.

[0037] Optionally, the standardization process can adopt the Z-score standardization method, which is to calculate the difference between the content value of a certain element in a certain grid cell and the average content value of the same element in all grid cells, and divide it by the standard deviation of the content value of the element to obtain the standardized element content value.

[0038] Principal component analysis (PCA) is a multivariate statistical method that transforms multiple correlated variables into a few uncorrelated composite variables through orthogonal transformations. These composite variables are the principal components. Specifically, the covariance matrix of a global dataset composed of the content values ​​of various mineralization-related elements in all grid units within the study area can be calculated. The eigenvalues ​​and eigenvectors of the covariance matrix are then solved, with each eigenvalue corresponding to a principal component. The larger the eigenvalue, the stronger the explanatory power of the principal component for the variation in the original data. The eigenvalues ​​are arranged in descending order, and multiple principal components with a cumulative variance contribution rate greater than a preset threshold are extracted. For each extracted principal component, the content value of the standardized elements in each grid unit is multiplied by the eigenvector corresponding to that principal component to obtain the score of each grid unit on that principal component. This score reflects the degree of agreement between the various mineralization-related elements in that grid unit and the combination of mineralization elements represented by that principal component.

[0039] After obtaining the score values ​​of each grid cell on each principal component, for each preset mineralization period combination of ore-forming elements, the sum of the absolute values ​​of the loading coefficients of each element contained in the feature vector of each extracted principal component is calculated, and the principal component with the largest sum of absolute values ​​is determined as the principal component corresponding to the preset mineralization period.

[0040] It should be noted that the cumulative variance contribution rate refers to the proportion of the sum of the eigenvalues ​​of the multiple principal components to the total sum of the eigenvalues ​​of all principal components. It is used to measure the degree to which the multiple principal components retain the overall variation information of the original element data.

[0041] The preset threshold can be taken as an empirical value of 85% to ensure that the extracted principal components can retain the main variation information of the original element data, while achieving effective dimensionality reduction. The number of these principal components can be determined based on the actual data. For example, when the cumulative variance contribution rate of the first to third principal components has reached 87%, the number of these principal components is determined to be 3.

[0042] Optionally, the distance of each grid cell from the fault of each preset mineralization period can be determined as the distance of each grid cell from the fault of each preset mineralization period.

[0043] It should be understood that the smaller the fault distance, the closer the grid cell is to the ore-controlling structure of the preset mineralization period in space, and the greater the possibility that the ore-forming fluid will migrate along the fault and affect the grid cell.

[0044] Optionally, the rock mass types included in the grid cell can be compared with the ore-controlling rock mass types of each preset mineralization period. If the rock mass type in the grid cell is consistent with the ore-controlling rock mass type of a preset mineralization period, the rock mass period of the grid cell under the preset mineralization period is determined to be in the assigned state; if they are inconsistent, the rock mass period of the grid cell under the preset mineralization period is determined to be in the unassigned state.

[0045] It should be understood that the classification of rock mass phases reflects the associated relationship between mineralization and concurrent magmatic activity, and the grid units belonging to the ore-controlling rock mass of that phase are more likely to be affected by the mineralization of that phase.

[0046] Optionally, the stratigraphic number of the strata in the grid unit can be compared with the favorable mineralized strata number of each preset mineralization period. If the stratigraphic number of the strata in the grid unit is consistent with the favorable mineralized strata number of a preset mineralization period, then the stratigraphic period of the grid unit under that preset mineralization period is determined to be favorable; if they are inconsistent, then the stratigraphic period of the grid unit under that preset mineralization period is determined to be unfavorable.

[0047] It should be understood that the stratigraphic sequence reflects the control of stratigraphic lithology on mineralization, and grid units located in favorable ore-bearing strata are more likely to become mineralization sites for that sequence.

[0048] In this embodiment of the application, the associated geological data is a collection of the above four types of phase-related geological features, which is used to comprehensively characterize the metallogenic geological environment of each grid unit under each preset metallogenic phase.

[0049] For example, suppose the ore-controlling factors for the preset mineralization period T1 are as follows: the ore-controlling rock mass type is intermediate-acidic rock mass, the ore-forming element combination is W-Mo-Bi, the fault identifier is F1, and the favorable ore-bearing strata number is D1; ​​the geological attribute information of grid cell A is as follows: the element content includes gold 3.2ppm, silver 1.5ppm, tungsten 12.0ppm, molybdenum 4.8ppm, bismuth 2.1ppm, arsenic 8.5ppm, and antimony 3.2ppm, and the element combination score of the ore-forming element combination W-Mo-Bi is 1.85; the distance from the F1 fault is 280 meters, the distance from the F2 fault is 120 meters, and the distance from the F3 fault is 450 meters; the rock mass type is granite porphyry; and the strata number is D2. The associated geological data of grid cell A under the preset mineralization period T1 can be as follows: the element combination score is 1.85; the fault distance from the F1 fault corresponding to period T1 is 280 meters; the rock mass period is assigned to a non-assigned state; and the stratigraphic period is assigned to a non-favorable state.

[0050] In the above method, by matching geological attribute information with the ore-controlling elements of the preset mineralization period, the undifferentiated original geological data is transformed into associated geological data closely related to the mineralization mechanism of each period. This allows for independent analysis based on the unique mineralization regularities of each period, avoiding the problem of period signal confusion caused by direct input of full data.

[0051] S103. Based on the locations of multiple known mineral deposits and the associated geological data of each grid unit under each preset mineralization period, determine the single-period mineralization probability of each grid unit under each preset mineralization period.

[0052] It should be understood that the single-period mineralization probability refers to the posterior probability of gold mineralization occurring in each grid cell under a single preset mineralization period, with a value ranging from 0 to 1.

[0053] S104. Based on the period affiliation of multiple known mineral deposits, the single-period mineralization probability of each grid unit under each preset mineralization period is weighted and fused to obtain the comprehensive mineralization probability of each grid unit.

[0054] It should be understood that the overall mineralization probability refers to the mineralization potential of each grid unit under the combined effect of all preset mineralization periods.

[0055] In one optional implementation, the number of known mineral deposits in each preset mineralization period can be determined based on the period affiliation of multiple known mineral deposits; the proportion of the number of known mineral deposits in each preset mineralization period to the total number of known mineral deposits in all preset mineralization periods can be determined as the period weight of each preset mineralization period; based on the period weight of each preset mineralization period, the single-period mineralization probability of each grid cell under each preset mineralization period is weighted and summed to obtain the comprehensive mineralization probability of each grid cell.

[0056] For example, suppose there are 5 known mineral deposits in the study area. The known mineral deposit 1 belongs to period T1, the known mineral deposit 2 belongs to period T1, the known mineral deposit 3 belongs to period T2, the known mineral deposit 4 belongs to period T3, and the known mineral deposit 5 belongs to period T3. Then the number of known mineral deposits for each preset mineralization period is 2, 1, and 2, respectively.

[0057] It should be understood that the phase weight is used to characterize the relative importance of each preset mineralization phase in the overall mineralization process of the study area, and the sum of the phase weights of each preset mineralization phase is 1.

[0058] Alternatively, the overall mineralization probability satisfies the following formula:

[0059] in, Represents grid cells The overall probability of mineralization Indicates the preset mineralization period Period weighting, Represents grid cells In the pre-set mineralization period The probability of single-phase mineralization is given. This represents the total number of pre-set mineralization periods.

[0060] In the above method, the number of known mineral deposits in each preset mineralization period is counted according to the period of known mineral deposits. Then, the period weight is obtained by calculating the proportion of the number of known mineral deposits in each period to the total number of mineral deposits. This results in the primary mineralization period (the period with more mineral deposits) receiving a higher weight in the comprehensive mineralization probability, while the secondary mineralization period (the period with fewer mineral deposits) receives a lower weight. This ensures that the comprehensive mineralization probability can truly reflect the superposition effect of multiple mineralization periods in the study area.

[0061] S105. Determine the gold mineralization prediction results for the study area based on the comprehensive mineralization probability of each grid unit.

[0062] In this embodiment of the application, the gold mineralization prediction result refers to the distribution of mineralization potential levels of each grid unit within the study area.

[0063] Optionally, each grid cell can be divided into multiple potential target areas based on a preset probability grading threshold.

[0064] For example, grid cells with a comprehensive mineralization probability greater than 0.7 are classified as high-potential target areas, grid cells with a comprehensive mineralization probability between 0.4 and 0.7 are classified as medium-potential target areas, and grid cells with a comprehensive mineralization probability less than 0.4 are classified as low-potential target areas.

[0065] The methods provided in S101-S105 establish a deep correlation between geological attribute information and the metallogenic regularity of preset metallogenic periods by acquiring geological attribute information of grid cells, ore-controlling elements of preset metallogenic periods, and the period affiliation and location of known mineral deposits. This transforms undifferentiated geological attribute information into associated geological data closely related to each preset metallogenic period, providing period-specific feature inputs for independent phased prediction. Based on this, by combining the location of known mineral deposits with associated geological data, the single-phase metallogenic probability is determined for each preset metallogenic period, enabling the metallogenic regularity of different preset metallogenic periods to operate within an independent prediction framework, avoiding the problems of mutual interference and superposition cancellation of signals from multiple periods. Finally, by weighting and fusing the single-phase metallogenic probability according to the period affiliation of known mineral deposits, the comprehensive metallogenic probability can reflect the true contribution ratio of each preset metallogenic period in the actual metallogenic process, thereby ensuring that the prediction results are highly consistent with the regional geological metallogenic regularity and significantly improving the accuracy and reliability of deep gold mineralization prediction results.

[0066] Figure 2This application illustrates a method flow for a three-dimensional prediction method for gold mineralization based on multi-source data fusion, according to an embodiment of this application. Figure 2 .

[0067] Combination Figure 1 ,like Figure 2 As shown, the above S103 can be implemented through S201-S202.

[0068] S201. Based on the associated geological data of each grid unit under each preset mineralization period, determine the mineralization contribution of each preset mineralization period to each grid unit.

[0069] Among them, the mineralization contribution degree is used to characterize the degree of influence of the mineralization of each grid cell on the corresponding preset mineralization period. The higher the mineralization contribution degree, the stronger the influence of the mineralization of that period on the grid cell.

[0070] In one alternative implementation, a feature matrix can be constructed with grid cells as rows and the associated geological data of each grid cell under each preset mineralization period as columns; non-negative matrix decomposition is performed on the feature matrix to obtain the mineralization contribution of each preset mineralization period to each grid cell.

[0071] It is understandable that the number of rows in the feature matrix is ​​equal to the total number of grid cells in the study area, and the number of columns is equal to the total number of features, which is the product of the number of each preset mineralization period and the number of four types of features (i.e., element combination score, fault distance from each preset mineralization period, rock mass period assignment, and stratigraphic period assignment).

[0072] Specifically, when constructing the feature matrix, it is necessary to convert the rock mass period assignment and stratigraphic period assignment into numerical codes. For example, the assignment status and favorable status are encoded as 1, and the non-assigned status and non-favorable status are encoded as 0. The element combination score and the fault distance from each preset mineralization period are subjected to maximum and minimum normalization processing to map the values ​​to the interval [0, 1]. Then, the feature columns corresponding to all preset mineralization periods are concatenated in order to form a complete feature matrix. This feature matrix is ​​a non-negative feature matrix.

[0073] For example, assuming there are three preset mineralization periods T1, T2, and T3, the columns of the feature matrix are as follows: T1 element combination score, T1 fault distance, T1 rock mass period assignment, T1 stratigraphic period assignment, T2 element combination score, T2 fault distance, T2 rock mass period assignment, T2 stratigraphic period assignment, T3 element combination score, T3 fault distance, T3 rock mass period assignment, and T3 stratigraphic period assignment.

[0074] It should be understood that nonnegative matrix decomposition is a mathematical method that decomposes a nonnegative data matrix into the product of two nonnegative matrices, and is suitable for separating mixed signals into several independent basic components. In the embodiments of this application, nonnegative matrix decomposition is used to separate multi-period data mixed in the feature matrix into independent contribution components corresponding to each preset mineralization period.

[0075] In one alternative implementation, the feature matrix can be decomposed into the product of the basis matrix and the coefficient matrix. Then, a loss function is constructed based on the element values ​​in the feature matrix, the element values ​​in the basis matrix, the element values ​​in the coefficient matrix, the fault distance of each grid cell from each preset mineralization period, and the rock mass period attribution of each grid cell under each preset mineralization period. The loss function is iteratively solved with the goal of minimizing it, and the mineralization contribution of each preset mineralization period to each grid cell is determined based on the element values ​​of each row in the solved basis matrix.

[0076] In this matrix, the rows of the basis matrix are the same as the rows of the feature matrix, that is, each row corresponds to a grid cell. The columns of the basis matrix are each preset mineralization period, and each column corresponds to the contribution distribution of a preset mineralization period. An element value in the basis matrix represents the contribution component of a preset mineralization period to a grid cell.

[0077] The coefficient matrix consists of preset mineralization periods, with each row corresponding to a feature pattern of a preset mineralization period. The columns of the coefficient matrix are the same as the columns of the feature matrix, with each column corresponding to a feature dimension.

[0078] It should be understood that the loss function is used to measure the degree of difference between the non-negative matrix decomposition result and the original feature matrix. In order to avoid the problem that the decomposition result does not conform to geological laws due to the lack of constraints in traditional non-negative matrix decomposition, in the embodiments of this application, two geological regularization constraints are introduced when constructing the loss function to ensure that the decomposition result conforms to geological laws. The loss function includes a matrix reconstruction error term, a fault distance regularization term, and a rock mass period regularization term.

[0079] Among them, the matrix reconstruction error term is used to measure how closely the product of the basis matrix and the coefficient matrix approximates the feature matrix. The smaller the value of this term, the more accurately the decomposition result can restore the geological feature distribution pattern in the original data.

[0080] Optionally, the matrix reconstruction error term is determined based on the element values ​​in the characteristic matrix, the element values ​​in the basis matrix, and the element values ​​in the coefficient matrix. It is obtained by calculating the product between the basis matrix and the coefficient matrix, and then calculating the square of the Frobenius norm between the characteristic matrix and the product.

[0081] The fault distance regularization term increases the penalty in the loss function for the contribution component of a grid cell corresponding to a preset mineralization period as the distance between the grid cell and the fault increases. Specifically, when the distance between a grid cell and the fault of a preset mineralization period is large, the value of the fault distance regularization term increases accordingly, thereby penalizing the decomposition result and making the decomposition result conform to the geological law of ore-forming fluid migration along faults.

[0082] The rock mass phase regularization term is used to increase the penalty in the loss function for the contribution component of a grid cell corresponding to a preset mineralization phase when the rock mass phase of a grid cell is inconsistent with that phase. Specifically, when the rock mass type of a grid cell does not belong to the ore-controlling rock mass type of a preset mineralization phase, the value of the rock mass phase regularization term is increased accordingly, thereby penalizing the decomposition result and making the decomposition result conform to the geological law of mineralization and concurrent magmatic activity.

[0083] Optionally, the loss function satisfies the following formula:

[0084] in, The value of the loss function. For the characteristic matrix, As a basis matrix, It is a coefficient matrix; The square of the Frobenius norm. This is the distance regularization coefficient, with an empirical value range of 0.1 to 0.3; This is the regularization coefficient for rock mass phases, with an empirical value range of 0.1 to 0.3; This represents the total number of grid cells; This is the preset total number of mineralization periods; Indicates the preset mineralization period For grid cells Contribution weight, Represents grid cells Distance from the preset mineralization period The normalized value of the fault distance ranges from 0 to 1. Represents grid cells The types of rock masses included and the corresponding periods; Period indicator function, when and The value is 0 when the state is consistent (i.e., when the state is affiliated), and 1 when the state is inconsistent (i.e., when the state is not affiliated).

[0085] In this formula, For the matrix reconstruction error term, For distance regularization, This represents the rock mass period regularization term. The sum of these three terms is the loss function value. By minimizing the loss function, both mathematically optimal decomposition and reasonable geological constraints can be achieved simultaneously.

[0086] This formula obtains the contribution components of each preset mineralization stage to each grid cell by finding the basis matrix that minimizes the loss function.

[0087] In one alternative implementation, the gradient descent method can be used to iteratively update the basis matrix and coefficient matrix with the goal of minimizing the loss function. When the rate of change of the loss function value in consecutive iterations is less than a preset rate of change threshold, the mineralization contribution of each preset mineralization period to each grid cell can be determined based on the element values ​​of each row in the basis matrix obtained in the current iteration.

[0088] It should be understood that gradient descent is an optimization algorithm that gradually approaches the minimum value of the loss function by updating parameters in the opposite direction of the gradient. In each iteration, the partial derivatives of the loss function with respect to each parameter are calculated based on the current basis matrix and the current coefficient matrix. The parameters are then updated in the opposite direction of the gradient with a certain step size, which gradually reduces the value of the loss function.

[0089] The rate of change threshold is used to determine whether the iterative optimization has converged. The rate of change can be calculated by the ratio of the difference between the loss function values ​​of two adjacent iterations to the previous loss function value. When the rate of change of the loss function value of multiple consecutive iterations is less than the preset rate of change threshold, it indicates that the iteration has converged. At this time, the update can be stopped and the current basis matrix can be output as the final decomposition result.

[0090] For example, the preset rate of change threshold can be set to 0.1%, and the number of consecutive iterations can be 10.

[0091] Understandably, due to the complexity of deep mineralization systems and local noise interference in the raw data, the contribution components obtained above may exhibit discontinuities or isolated anomalies in spatial distribution. That is, the contribution components of some grid cells may be significantly higher than those of their surrounding grid cells, forming isolated high values ​​that are inconsistent with geological patterns. However, from a geological perspective, mineralization within the same period should be spatially continuous. Ore-forming fluids migrate along faults and precipitate in favorable geological environments, and their influence range typically exhibits a continuous and gradual change, rather than abruptly jumping isolated points. Therefore, it is necessary to spatially smooth the contribution components of each preset mineralization period obtained from the base matrix to each grid cell, making the smoothed results more consistent with actual mineralization patterns.

[0092] In one alternative implementation, for each preset mineralization period, the mineralization favorability coefficient of each grid cell under the preset mineralization period can be determined. Based on the mineralization favorability coefficient, the contribution components of all grid cells within the three-dimensional sliding window of each grid cell can be spatially weighted and averaged to obtain the mineralization contribution of each grid cell under the preset mineralization period.

[0093] It should be understood that the mineralization favorability coefficient is used to characterize the favorable geological environment in which the grid cell is located.

[0094] Optionally, if the strata of a grid cell are located in the favorable ore-bearing strata of the preset mineralization period, the fault distance from the preset mineralization period is less than a preset distance threshold (e.g., 200 meters), and the rock mass type of the grid cell is consistent with the ore-controlling rock mass type of the preset mineralization period, the mineralization favorableness coefficient of the grid cell under the preset mineralization period can be determined to be 1; otherwise, the mineralization favorableness coefficient is 0.2.

[0095] Optionally, the size of the three-dimensional sliding window can be set to 3 x 3 x 3 grid cells. For a certain grid cell to be smoothed, a grid cell is extended to its three-dimensional neighborhood from the center of the grid cell to form a cube window containing 27 grid cells.

[0096] Optionally, the mineralization contribution satisfies the following formula:

[0097] in, Indicates the preset mineralization period For grid cells The contribution of mineralization; Indicates the preset mineralization period Mesh cells within a 3D sliding window Contribution component; Represents grid cells The mineralization favorableness coefficient; Represents grid cells The set of all mesh cells within a 3D sliding window.

[0098] In one optional implementation, the mineralization contribution of each grid cell under each preset mineralization period can be judged as an anomaly. Grid cells whose mineralization contribution is greater than the contribution threshold and whose proportion of the mineralization contribution greater than the average mineralization contribution in their three-dimensional sliding window is lower than a preset proportion threshold are marked as isolated anomaly cells. The mineralization contribution of isolated anomaly cells is reset to the average value of the mineralization contribution of all grid cells in their three-dimensional sliding window; the mineralization contribution of non-isolated anomaly cells remains unchanged.

[0099] Optionally, a preset mineralization period corresponds to a contribution threshold and a mean contribution. For each preset mineralization period, the mean and standard deviation of the mineralization contribution of all grid cells under that preset mineralization period are calculated. The mean contribution is determined as the mean contribution, and the sum of the mean contribution and the standard deviation is determined as the contribution threshold.

[0100] It is understandable that when the proportion of grid cells with a mineralization contribution greater than the average mineralization contribution within the three-dimensional sliding window of a grid cell is lower than the preset proportion threshold, it indicates that the mineralization contribution of the vast majority of grid cells around that grid cell is lower than the average level, which does not conform to the geological law of continuous spatial distribution of mineralization.

[0101] Optionally, the preset ratio threshold can be set to an empirical value of 80%.

[0102] In the above method, by constructing a feature matrix and performing non-negative matrix decomposition, the mixed multi-stage signals in each grid cell are separated into independent contribution components for each preset mineralization stage. During the non-negative matrix decomposition process, a loss function containing reconstruction error terms, distance regularization terms, and rock mass stage regularization terms is constructed to ensure both mathematical accuracy of the decomposition and that the separation results satisfy the constraints of objective geological laws. The resulting mineralization contribution accurately reflects the true spatial influence range of mineralization for each preset mineralization stage.

[0103] S202. Based on the mineralization contribution of each preset mineralization period to each grid unit, the associated geological data of each grid unit under each preset mineralization period, and the location of multiple known mineral deposits, determine the single-period mineralization probability of each grid unit under each preset mineralization period.

[0104] In one alternative implementation, the grid cell to which each known mineral deposit belongs can be determined based on the location of multiple known mineral deposits. The ratio between the number of grid cells containing known mineral deposits and the total number of all grid cells is determined as the a priori mineralization probability of the study area. The data and mineralization contribution of each grid cell in the associated geological data under each preset mineralization period are used as evidence factors. The existence status of each evidence factor is determined according to the preset mineralization conditions. The evidence weight of each evidence factor is determined according to its existence status. The single-period mineralization probability of each grid cell under each preset mineralization period is determined based on the evidence weight of each evidence factor under each preset mineralization period and the a priori mineralization probability of the study area.

[0105] Optionally, the three-dimensional coordinates of a known mineral deposit can be projected onto a three-dimensional geological entity model to determine the grid cell to which each known mineral deposit belongs.

[0106] It should be understood that the a priori mineralization probability refers to the background probability of gold mineralization occurring in any grid cell within the study area without considering any evidence factors.

[0107] Evidence factors refer to geological variables used to infer the probability of mineralization, including element combination scores, fault distances, rock mass period assignments, stratigraphic period assignments, and mineralization contributions in associated geological data. Existence status refers to whether each evidence factor meets the preset mineralization conditions, and existence status includes existence or non-existence.

[0108] It should be understood that since the current calculation requires the probability of mineralization in a single period, each preset mineralization period should be analyzed separately to avoid interference between the mineralization patterns of different periods.

[0109] Optionally, the preset mineralization conditions may include: determining the existence status of the evidence factor as present when the element combination score is greater than or equal to the element combination score threshold, otherwise as absent; determining the existence status of the evidence factor as present when the fault distance is less than or equal to the fault distance threshold, otherwise as absent; determining the existence status as present when the rock mass period is assigned to a given state, otherwise as absent; determining the existence status as present when the stratigraphic period is assigned to a favorable state, otherwise as absent; and determining the existence status as present when the mineralization contribution is greater than or equal to the contribution threshold, otherwise as absent.

[0110] Optionally, the element combination score threshold can be the median of the element combination scores of all grid units under the preset mineralization period in the study area, the fault distance threshold can be an empirical value of 200 meters, and the contribution threshold can be the average mineralization contribution of all grid units under the preset mineralization period in the study area.

[0111] It should be understood that the weight of evidence refers to the influence of each evidence factor on the probability of mineralization in a single period under different states of existence.

[0112] In one optional implementation, grid cells belonging to known mineral deposits can be marked as positive samples, and grid cells whose spatial distance from known mineral deposits is greater than a preset distance threshold and which do not contain known mineral deposits can be marked as negative samples. The proportion of grid cells in the positive samples that are in the state of existence of the same type of evidence factor under the same preset mineralization period is determined as the first conditional probability of the same type of evidence factor under the same preset mineralization period when it is in the state of existence. The proportion of grid cells in the negative samples that are in the state of existence of the same type of evidence factor under the same preset mineralization period is determined as the second conditional probability of the same type of evidence factor under the same preset mineralization period when it is in the state of existence. The evidence weight of the same type of evidence factor under the same preset mineralization period when it is in the state of existence is determined based on the ratio of the first conditional probability to the second conditional probability.

[0113] It should be understood that the preset distance threshold is used to ensure that there is sufficient spatial separation between negative samples and known mineral points, so as to avoid misjudging units within the mineralization influence range as unmineralized.

[0114] For example, the preset distance threshold can be set to 500 meters, that is, grid cells that are more than 500 meters away from a known mining site and do not contain a known mining site are selected as negative samples.

[0115] It is understandable that the number of negative samples is usually greater than the number of positive samples. In order to ensure the balance of training samples, the same number of grid cells as the number of positive samples can be selected from the negative samples as the final negative samples. For example, random sampling or sampling according to spatial uniform distribution can be used to ensure that the number of positive samples and negative samples is the same.

[0116] It should be understood that the first conditional probability refers to the conditional probability of mineralization occurring in the grid cell under the presence of the evidence factor, and the second conditional probability refers to the conditional probability of mineralization not occurring in the grid cell under the presence of the evidence factor. The ratio of the first conditional probability to the second conditional probability is used to measure the relative change in the probability of mineralization relative to the probability of non-mineralization under the presence of the evidence factor.

[0117] Optionally, the natural logarithm of the ratio can be taken to transform the probability ratio into a symmetric weight value, which can be used as the evidence weight of the evidence factor when the existence state is existence.

[0118] Understandably, when the first conditional probability is higher than the second conditional probability, the ratio is greater than 1, and the natural logarithm yields a positive value, indicating that the existence of this evidence factor is conducive to mineralization. Furthermore, the larger the ratio, the higher the evidence weight, indicating that the evidence factor has a stronger indicative significance for mineralization. When the first conditional probability is lower than the second conditional probability, the ratio is less than 1, and the natural logarithm yields a negative value, indicating that the existence of this evidence factor is not conducive to mineralization. When the two are equal, the ratio is equal to 1, and the natural logarithm yields a zero value, indicating that the existence of this evidence factor has no effect on the probability of mineralization.

[0119] It should be noted that when all grid cells that do not contain the evidence factor are not mineralized, the existence of the evidence factor is strongly correlated with mineralization. In this case, the evidence weight of the existence state tends to positive infinity, and it can be directly assigned a reasonable upper limit value, such as 5, to avoid numerical overflow. When all grid cells that contain the evidence factor are not mineralized, the evidence weight of the existence state tends to negative infinity, and it can be directly assigned a reasonable lower limit value, such as -5.

[0120] In one optional implementation, the proportion of grid cells in the positive sample where the existence state of the same type of evidence factor under the same preset mineralization period is non-existent can be determined as the third conditional probability when the existence state of the same type of evidence factor under the same preset mineralization period is non-existent; the proportion of grid cells in the negative sample where the existence state of the same type of evidence factor under the same preset mineralization period is non-existent can be determined as the fourth conditional probability when the existence state of the same type of evidence factor under the same preset mineralization period is non-existent; and the evidence weight when the existence state of the same type of evidence factor under the same preset mineralization period is non-existent can be determined based on the ratio of the third conditional probability to the fourth conditional probability.

[0121] It should be understood that the third conditional probability refers to the conditional probability of mineralization occurring in a grid cell when the evidence factor is absent, and the fourth conditional probability refers to the conditional probability of mineralization not occurring in a grid cell when the evidence factor is absent. The ratio of the third conditional probability to the fourth conditional probability is used to measure the relative change in the probability of mineralization relative to the probability of non-mineralization when the evidence factor is absent.

[0122] Optionally, the natural logarithm of the ratio can be taken to transform the probability ratio into a symmetric weight value, which can be used as the evidence weight of the evidence factor when the existence state is non-existence.

[0123] It is understandable that when the third conditional probability is lower than the fourth conditional probability, it indicates that the absence of the evidence factor is positively correlated with gold mineralization. In this case, taking the natural logarithm yields a positive value, indicating that the absence of the evidence factor itself is not conducive to mineralization, that is, the presence of the evidence factor is beneficial; otherwise, it is a negative value.

[0124] It should be noted that when all grid cells containing the evidence factor are mineralized, i.e., when the evidence factor exists, the probability of non-mineralization is zero. In this case, the evidence weight for non-mineralization tends towards negative infinity, and can be directly assigned a reasonable lower limit value, such as -5. When all grid cells not containing the evidence factor are mineralized, i.e., when the evidence factor does not exist, the probability of non-mineralization is zero. In this case, the evidence weight for non-mineralization tends towards positive infinity, and can be directly assigned a reasonable upper limit value, such as 5, to avoid numerical overflow.

[0125] Alternatively, the probability of mineralization in a single phase may satisfy the following formula:

[0126] in, Represents grid cells In the pre-set mineralization period The probability of single-phase mineralization is given. This represents the prior mineralization probability of the study area. This represents the prior probability that the mineral will not form; Represents grid cells In the pre-set mineralization period The total number of evidence factors It is a natural exponential function; It is the natural logarithm function; For grid cells In the pre-set mineralization period Next The evidence weights of each evidence factor are based on the grid cells. In the pre-set mineralization period Next The existence status of each evidence factor is determined.

[0127] In this formula, the numerator term Characterizing mesh cells In the pre-set mineralization period The posterior probability of mineralization. This parameter characterizes the overall metallogenic background benchmark of the study area. If there are many known mineral occurrences in the study area, this parameter will be positive, which will increase the overall probability of single-period mineralization, and vice versa. The comprehensive correction factor characterizing the probability of mineralization in a single period; when a grid cell possesses favorable evidence factors, the corresponding... A positive value increases the term upon accumulation, while a negative value decreases the term due to unfavorable evidence. (Denominator term) Used for molecular normalization to ensure It falls within the range of 0 to 1. When the numerator is larger, the denominator is approximately equal to the numerator, and the probability of mineralization in a single period approaches 1. When the numerator is smaller, the denominator approaches 1, and the probability of mineralization in a single period approaches 0.

[0128] The final single-period mineralization probability characterizes the probability that the grid unit will actually undergo gold mineralization under the influence of the preset mineralization period, based on the known mineralization background of the study area and after comprehensively considering all evidence factors of the grid unit under the preset mineralization period.

[0129] The methods provided in S201-S202 above decompose the associated geological data of each grid unit, so that the mineralization contribution of each grid unit under each preset mineralization period can truly reflect the influence intensity of the preset mineralization period on the grid unit. On this basis, the single-period mineralization probability is calculated for the exclusive data of each preset mineralization period, which fundamentally avoids the problem of mutual interference and superposition cancellation of mineralization patterns of different preset mineralization periods, and significantly improves the geological interpretability and accuracy of single-period mineralization probability prediction.

[0130] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0131] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A three-dimensional prediction method for gold mineralization based on multi-source data fusion, characterized in that, include: The study aims to acquire geological attribute information of multiple grid units within the study area, ore-controlling elements of multiple preset mineralization periods, and the period affiliation and location of multiple known mineral deposits. Based on the geological attribute information of each grid unit and the ore-controlling elements of each preset metallogenic period, the associated geological data of each grid unit under each preset metallogenic period are determined. Based on the locations of multiple known mineral deposits and the associated geological data of each grid unit under each preset mineralization period, the single-period mineralization probability of each grid unit under each preset mineralization period is determined. Based on the period affiliation of multiple known mineral deposits, the single-period mineralization probability of each grid unit under each preset mineralization period is weighted and fused to obtain the comprehensive mineralization probability of each grid unit. The gold mineralization prediction results for the study area are determined based on the comprehensive mineralization probability of each grid cell.

2. The method according to claim 1, characterized in that, The geological attribute information of a grid cell includes the content values ​​of various ore-forming elements, the distance from each fault, the rock mass type, and the stratigraphic number. The ore-controlling elements of a preset mineralization period include fault identification, ore-controlling rock mass type, favorable ore-bearing strata number, and ore-forming element combination. The process of determining the associated geological data of each grid cell under each preset mineralization period based on the geological attribute information of each grid cell and the ore-controlling elements of each preset mineralization period includes: Based on the content values ​​of various mineralization-related elements included in the geological attribute information of each grid cell and the combination of mineralization elements in each preset mineralization period, the element combination score of each grid cell in each preset mineralization period is determined. Based on the rock mass types included in the grid unit and the ore-controlling rock mass types of each preset mineralization period, the rock mass period affiliation of each grid unit under each preset mineralization period is determined; Based on the stratigraphic number of the strata within the grid unit and the favorable ore-bearing strata number of each preset metallogenic period, the stratigraphic period affiliation of each grid unit under each preset metallogenic period is determined. Based on the distance of each grid unit from each fault and the fault markers of each preset mineralization period, the distance of each grid unit from each preset mineralization period is determined. The element combination score, rock mass period assignment, stratigraphic period assignment, and fault distance from each preset metallogenic period for each grid cell are determined as the associated geological data for each grid cell under each preset metallogenic period.

3. The method according to claim 2, characterized in that, The step of determining the single-phase mineralization probability of each grid unit under each preset mineralization phase, based on the locations of multiple known mineral deposits and the associated geological data of each grid unit under each preset mineralization phase, includes: Based on the associated geological data of each grid unit under each preset metallogenic period, the metallogenic contribution of each preset metallogenic period to each grid unit is determined. Based on the mineralization contribution of each preset mineralization period to each grid cell, the associated geological data of each grid cell under each preset mineralization period, and the location of multiple known mineral deposits, the single-period mineralization probability of each grid cell under each preset mineralization period is determined.

4. The method according to claim 3, characterized in that, The step of determining the mineralization contribution of each preset mineralization period to each grid unit based on the associated geological data of each grid unit under each preset mineralization period includes: A feature matrix is ​​constructed with grid cells as rows and the associated geological data of each grid cell under each preset mineralization period as columns. Perform nonnegative matrix decomposition on the feature matrix to obtain the mineralization contribution of each preset mineralization period to each grid cell.

5. The method according to claim 4, characterized in that, The nonnegative matrix decomposition of the feature matrix to obtain the mineralization contribution of each preset mineralization stage to each grid cell includes: The feature matrix is ​​decomposed into the product of a basis matrix and a coefficient matrix. The rows of the basis matrix are the same as the rows of the feature matrix, the columns of the basis matrix are each preset mineralization period, the rows of the coefficient matrix are each preset mineralization period, and the columns of the coefficient matrix are the same as the columns of the feature matrix. A loss function is constructed based on the element values ​​in the feature matrix, the element values ​​in the basis matrix, the element values ​​in the coefficient matrix, the fault distance of each grid cell from each preset mineralization period, and the rock mass period attribution of each grid cell under each preset mineralization period. The solution is iteratively obtained by minimizing the loss function, and the contribution of each preset mineralization period to the mineralization of each grid cell is determined based on the element values ​​of each row in the obtained basis matrix.

6. The method according to claim 5, characterized in that, The iterative solution, with the goal of minimizing the loss function, determines the element values ​​of each row in the obtained basis matrix as the mineralization contribution of each preset mineralization stage to each grid cell, including: With the goal of minimizing the loss function, the gradient descent method is used to iteratively update the basis matrix and coefficient matrix; When the rate of change of the loss function value in multiple consecutive iterations is less than a preset rate of change threshold, the mineralization contribution of each preset mineralization period to each grid cell is determined based on the element values ​​of each row in the basis matrix obtained in the current iteration.

7. The method according to claim 3, characterized in that, The determination of the single-phase mineralization probability of each grid unit under each preset mineralization phase, based on the mineralization contribution of each preset mineralization phase to each grid unit, the associated geological data of each grid unit under each preset mineralization phase, and the location of multiple known mineral deposits, includes: The grid cell to which each known mineral deposit belongs is determined based on the location of multiple known mineral deposits; The ratio between the number of grid cells with known mineral deposits and the total number of all grid cells is determined as the a priori mineralization probability of the study area. The data and mineralization contribution of each grid cell in the associated geological data of each preset mineralization period are used as evidence factors. The existence status of each evidence factor is determined according to the preset mineralization conditions. The evidence weight of each evidence factor is determined based on its existence status. Based on the evidence weights of each evidence factor in each grid unit under each preset mineralization period and the prior mineralization probability of the study area, the single-period mineralization probability of each grid unit under each preset mineralization period is determined.

8. The method according to claim 7, characterized in that, The determination of the evidence weight of each evidence factor based on its existence status includes: The grid cells to which the known mining point belongs are marked as positive samples, and the grid cells that are more than a preset distance threshold from the known mining point and do not contain the known mining point are marked as negative samples. The proportion of grid cells in the positive sample that are in the state of existence of the same type of evidence factor under the same preset mineralization period is determined as the first conditional probability of the same type of evidence factor under the same preset mineralization period when it is in the state of existence. The proportion of grid cells in the negative sample that are in the state of existence of the same type of evidence factor under the same preset mineralization period is determined as the second conditional probability of the same type of evidence factor in the state of existence under the same preset mineralization period. Based on the ratio of the first conditional probability to the second conditional probability, the evidence weight of the same type of evidence factor under the same preset mineralization period is determined when the existence state is "existence".

9. The method according to claim 8, characterized in that, The determination of the evidence weight of each evidence factor based on its existence status includes: The proportion of grid cells in the positive sample where the state of existence of the same type of evidence factor under the same preset mineralization period is non-existent is determined as the third conditional probability of the same type of evidence factor under the same preset mineralization period when the state of existence is non-existent. The proportion of grid cells in the negative sample where the state of existence of the same type of evidence factor under the same preset mineralization period is non-existent is determined as the fourth conditional probability of the same type of evidence factor under the same preset mineralization period when the state of existence is non-existent. Based on the ratio of the third conditional probability to the fourth conditional probability, the evidence weight of the same type of evidence factor under the same preset mineralization period is determined when the existence state is non-existent.

10. The method according to claim 1, characterized in that, The step of weightedly fusing the single-period mineralization probability of each grid unit under each preset mineralization period based on the period affiliation of multiple known mineral deposits to obtain the comprehensive mineralization probability of each grid unit includes: The number of known mineral deposits for each preset mineralization period is determined based on the period affiliation of multiple known mineral deposits. The weight of each preset mineralization period is determined based on the proportion of the number of known mineral deposits in each preset mineralization period to the total number of known mineral deposits in all preset mineralization periods. Based on the period weights of each preset mineralization period, the single-period mineralization probability of each grid unit under each preset mineralization period is weighted and summed to obtain the comprehensive mineralization probability of each grid unit.

Images