A method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prediction

Abstract

The application discloses a method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prediction, comprising the following steps: S1, constructing ore-controlling factors based on three-dimensional geological modeling according to multi-source geoscience data; S2, determining the correlation of ore-controlling factors based on Bayesian-MCMC structure learning; and S3, calculating the conditional probability distribution table by using maximum likelihood estimation and then calculating the metallogenic posterior probability. Compared with the prior art, the Bayesian network used in the application can not only reflect the relationship between ore-controlling factors and known ore bodies, but also explore the complex correlation between ore-controlling factors, decompose the joint distribution into multiple simple probability distributions by using the variable conditional independent relationship, thereby reducing the model complexity and improving the reasoning efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional metallogenic prediction, in particular to a method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prediction. BACKGROUND

[0002] The formation of a deposit is a complex process coupled by multiple factors, and is not an isolated event. The formation process includes the source, migration, trapping, deposition and storage of metals and fluids, which leads to the fact that the ore-controlling factors in three-dimensional metallogenic prediction have a large amount of data, high data dimension, complex data pattern, and complex correlation and association between data. Therefore, it is challenging to mine the internal relationship between multiple source geological information.

[0003] Three-dimensional metallogenic prediction can be divided into knowledge-driven and data-driven methods. The knowledge-driven method needs to determine the correlation between factors manually, but this method has three defects: first, the determination of the correlation is greatly influenced by subjectivity, and different people may have different subjective judgments, which may lead to a large difference in the calculated post-mining probability; second, this method cannot clearly determine the spatial correlation between ore-controlling factors, that is, it cannot understand the spatial heterogeneity of the ore body and its spatial relationship with the ore-controlling factors; finally, in three-dimensional space, geological data not only has a large amount of data, high data dimension, complex data pattern, but also has strong correlation between data, and it is difficult for manual selection of ore-controlling factors to consider the complex correlation between data. A core task of the data-driven method is to mine implicit knowledge from observed data. This method is based on spatial statistical theory and based on a large amount of actual geological exploration data, so it has less human intervention and the obtained results are more objective.

[0004] The current three-dimensional metallogenic prediction method, whether knowledge-driven or data-driven, cannot accurately quantify the correlation between ore-controlling factors. The knowledge-driven method has defects in objectivity and reusability; although the data-driven method can automatically explore the correlation between ore-controlling factors and known ore bodies, it cannot reveal the quantitative correlation between factors, so the explainability of the three-dimensional metallogenic prediction process is poor. SUMMARY

[0005] In view of the problems in the related art, the present application provides a method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prediction to solve the problems in the background art.

[0006] To this end, the present application adopts the following specific technical solutions:

[0007] A method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prediction, comprising the following steps,

[0008] Step S1, constructing ore-forming factors based on three-dimensional geological modeling according to multi-source geosciences data;

[0009] Step S2, determining the correlation between ore-controlling factors based on Bayesian-MCMC structure learning;

[0010] Step S3, calculating the conditional probability distribution table by maximum likelihood estimation.

[0011] In one possible design, step S1 includes the following steps:

[0012] Step S11, obtaining modeling data;

[0013] Step S12, mapping data from two dimensions to three dimensions, mapping point data and line data from two-dimensional plane to three-dimensional space by stacking a digital elevation model (DEM), to obtain point set data and line set data in three-dimensional space;

[0014] Step S13, for geological data, constructing a corresponding three-dimensional geological structure model in the order of line, surface and volume by using explicit modeling, and finally constructing a three-dimensional field model by using three-dimensional spatial analysis, and for geological attribute data, constructing an attribute model by using spatial interpolation method;

[0015] Step S14, constructing a favorable ore-forming degree conceptual model, determining ore-forming favorable factors and constructing a three-dimensional ore-controlling factor model by analyzing the spatial relationship between known deposits and three-dimensional geological entity models and three-dimensional geological attribute models, and combining existing experience and knowledge;

[0016] Step S15, determining the threshold of ore-controlling factors by using the ore-forming conceptual model combined with a data-driven method, assigning a value of 1 to a region that meets the ore-controlling threshold and a value of 0 to a region that does not meet the ore-controlling threshold, to construct a binary ore-controlling factor model.

[0017] In one possible design, step S2 includes the following steps,

[0018] Step S21, constructing ore-controlling factor nodes;

[0019] Step S22, calculating the posterior probability of the correlation between ore-controlling factors;

[0020] Step S23, calculating the score of the topological structure between ore-controlling factors to screen the topological structure;

[0021] Step S24, searching and evaluating the topological structure to determine the correlation between ore-controlling factors;

[0022] Step S25, constructing the neighborhood space of the topological structure between factors;

[0023] Step S26, screening the topological structure;

[0024] Step S27, taking the conditional probability of each node as the estimated value to be solved in the maximum likelihood estimation, and determining the conditional probability of the node by calculating the extreme value.

[0025] In one possible design, in step S21, the Bayesian network used to describe the correlation between the ore-controlling factors and the known gold ore bodies is composed of two parts: a topological structure G and a conditional probability distribution table θ, where G is a DAG composed of V and an edge set E, θ is the joint probability distribution of each node calculated based on the conditional independence between different variables defined by the DAG, and V = {V1, V2, …, Vn} is the set of ore-controlling factors and known gold ore bodies, corresponding to the nodes in the DAG, and X = {X1, X2, …, Xn} is the set of geological variables, and each V is associated with a random geological variable X N . N i i i Each X is composed of binary discrete data derived from a binary evidence factor model.

[0026] In one possible design, in step S22, the topological structure G composed of ore-controlling factors and known gold ore body nodes is used in the iterative search process, and the posterior probability calculation formula of the topological structure G is as follows:

[0027]

[0028] Where p(G|X) is the posterior probability of the topological structure; p(G) is the prior based on all possible topological structures (G ∈ ζ n ), and a uniform distribution is used as the prior; p(X) is the probability of binary discrete data; p(X|G) is the marginal likelihood, which is the possibility of generating the corresponding combination of binary discrete data under the condition of determining the topological structure. In Bayesian inference, p(X) is an integral constant, and the calculation formula is as follows:

[0029]

[0030] Where ζ n is the set of all possible DAG structures.

[0031] In one possible design, in step S23, for each topological structure, a score is calculated to evaluate the fitting degree of the topological structure to the binary discrete data to screen the topological structure. Since the BN uses the independence assumption and the non-informative graph prior, the marginal likelihood function can be decomposed into the product of the local marginal probability of each factor X v (score v given the parent node).​​​

[0032]

[0033] wherein, score i (Pa G (X i ) is the score of node i, Pa G (X i ) is the parent node set of node i.

[0034] In a possible design, in step S11, the modeling data sources include but are not limited to planar geological maps, profile geological map extraction stratum boundary, fault boundary, apparent resistivity, magnetic anomaly and residual gravity density and other attribute sampling point data. The application further provides a three-dimensional metallogenic prediction control ore factor correlation determining device, including a memory, a control processor and a computer program stored on the memory and executable on the control processor, and the control processor executes the program to realize the three-dimensional metallogenic prediction control ore factor correlation determining method.

[0035] The application further provides a three-dimensional metallogenic prediction system, including the three-dimensional metallogenic prediction control ore factor correlation determining device.

[0036] The application further provides a computer readable storage medium, which stores computer executable instructions, and the computer executable instructions are used to make a computer execute the three-dimensional metallogenic prediction control ore factor correlation determining method.

[0037] Compared with the prior art, the application has the following beneficial effects:

[0038] 1. The Bayesian network used in the application can not only reflect the relationship between the ore-controlling factors and the known ore bodies, but also explore the complex correlation between the ore-controlling factors, decompose the joint distribution into multiple simple probability distributions by using the variable conditional independence relationship, thereby reducing the model complexity and improving the inference efficiency.

[0039] 2. The application uses the Bayesian network to mine multi-source geological information under uncertain environment, converts the geological problem into a mathematical model, and clearly defines the spatial variation of the three-dimensional geological ore-controlling elements, the correlation between the control ore factors in the three-dimensional metallogenic prediction and the strength thereof, can provide strong basis for explaining the complex metallogenic process, thereby improving the geological exploration efficiency, and therefore it is very important to determine and quantify the correlation between the ore-controlling factors.

[0040] 3. The purpose of the present application is to mine the correlation of ore-controlling factors based on Bayesian-MCMC structure learning in three-dimensional metallogenic prediction, which can explore the correlation between known ore bodies and ore-controlling factors and between ore-controlling factors, and at the same time, quantize the correlation between ore-controlling factors by using conditional probability distribution table, represent the ore-forming prediction influencing factors in three-dimensional space, overcome the difficulty of explaining the correlation between the prediction results and the geological ore-controlling factors in the existing three-dimensional metallogenic prediction method, and reveal the complex correlation between ore-controlling factors in three-dimensional space only by relying on input data.

[0041] 4. The present application takes the binary model of ore-controlling factors as input, constantly searches for new topological structures through transition operators and scoring functions, determines whether the state is transferred according to the acceptance probability in Monte Carlo sampling, finally obtains a stationary Markov chain, and the topological structure with the highest frequency in the chain is the final correlation between input ore-controlling factors, and then the conditional probability distribution table is calculated by using maximum likelihood estimation to calculate the ore-forming posterior probability.

[0042] 5. Compared with the existing method, the technical scheme proposed by the present application is a method for determining and quantizing the correlation between ore-controlling factors based on Bayesian-MCMC structure learning, which can directly learn the correlation between ore-controlling factors that fits the sample data according to the input discrete attribute data, quantize the correlation between ore-controlling factors in three-dimensional metallogenic prediction and its strength, and provide strong basis for explaining the complex ore-forming process. The process only relies on data and has no subjective experience interference.

[0043] 6. The present application performs statistical analysis on the discrete attribute values of the input ore-controlling factors, scores the Bayesian network composed of topological structures and conditional probability distribution table, calculates the acceptance probability of the topological structure with the highest score, and repeatedly performs the process until a stationary Markov chain is constructed, and the final factor correlation is obtained by statistically analyzing the structures in the chain. The present application can not only quantize the correlation between factors, but also generate a conditional probability distribution table according to the correlation as a basis for calculating the ore-forming posterior probability, which is more in line with the geological ore-forming regularity, has stronger explanatory ability, and is convenient for subsequent verification. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0045] Figure 1A flow chart of a method for determining the correlation between intermediate ore-controlling factors in three-dimensional metallogenic prediction based on Bayesian-MCMC structure learning in the method for determining the correlation between intermediate ore-controlling factors in three-dimensional metallogenic prediction is shown.

[0046] Figure 2 A metallogenic geological background map of a certain region in an embodiment of the application is shown, wherein: (a) a geological map; (b) a tectonic framework map.

[0047] Figure 3 A three-dimensional ore-controlling factor model diagram in the application is shown, wherein: (a) a fault distance field (D f ); (b) a paleo-crater distance field (D c ); (c) a Ganhe Formation and Early Carboniferous granite complex contact surface distance field (D i ); (d) a residual gravity density (D); (e) a magnetic anomaly (M); (g) a apparent resistivity distribution (R);

[0048] Figure 4 A binary ore-controlling factor model diagram in the application is shown, wherein: (a) a fault distance field (D f ); (b) a paleo-crater distance field (D c ); (c) a Ganhe Formation and Early Carboniferous granite complex contact surface distance field (D i ); (d) a residual gravity density (D); (e) a magnetic anomaly (M); (f) a apparent resistivity distribution (R);

[0049] Figure 5 A correlation diagram between ore-controlling factors in the application is shown.

[0050] Figure 6 A conditional probability distribution rose diagram between ore-controlling factors in the application based on Figure 4 , wherein: (a) a residual gravity density (D); (b) a Ganhe Formation and Early Carboniferous granite complex contact surface distance field (D i ); (c) a magnetic anomaly (M); (d) a paleo-crater distance field (D c ); (e) a apparent resistivity (R); (f) a gold mineralization (G). DETAILED DESCRIPTION

[0051] To further illustrate the embodiments, the application provides drawings which are part of the disclosure of the application and mainly serve to illustrate the embodiments and can be used to explain the operating principle of the embodiments in conjunction with the relevant description of the specification. Those of ordinary skill in the art can understand other possible implementations and advantages of the application by referring to these contents. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0052] The method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prediction provided in the present example comprises the following steps, as shown in Figure 1

[0053] Step S1, constructing an ore-forming factor based on three-dimensional geological modeling according to multi-source geoscience data; specifically comprising,

[0054] Step S11, obtaining modeling data, such as extracting stratum boundary, fault boundary, apparent resistivity, magnetic anomaly and residual gravity density attribute sampling point data, etc. from plane geological map and profile geological map;

[0055] Step S12, mapping data from two-dimensional to three-dimensional, realizing the mapping of point data and line data from two-dimensional plane to three-dimensional space through superimposing a digital elevation model (DEM), to obtain point set data and line set data in three-dimensional space;

[0056] Step S13, for geological data, constructing a corresponding three-dimensional geological structure model in the order of line, surface and body by using explicit modeling, and finally constructing a three-dimensional field model by using three-dimensional spatial analysis, and for geological attribute data, constructing an attribute model by using a spatial interpolation method;

[0057] Step S14, constructing an ore-forming favorable degree conceptual model, determining ore-forming favorable factors and constructing a three-dimensional ore-controlling factor model by analyzing the spatial relationship between known ore deposits and three-dimensional geological entity model and three-dimensional geological attribute model, and combining existing experience knowledge, a three-dimensional ore-controlling factor model example is shown in Figure 3

[0058] Step S15, determining the threshold of ore-controlling factors by using the ore-forming conceptual model combined with a data-driven method, assigning a value of 1 to a region meeting the ore-controlling threshold and a value of 0 to a region not meeting the ore-controlling threshold, to construct a binary ore-controlling factor model, a binary ore-controlling factor model example is shown in Figure 4

[0059] Step S2, learning and determining the correlation of ore-controlling factors based on Bayesian-MCMC structure; the specific steps are as follows,

[0060] ​​​Step S21: Construct ore-controlling factor nodes. A Bayesian Network (BN) is constructed to describe the relationships between ore-controlling factors and known gold ore bodies. It consists of a topological structure G and a conditional probability distribution table θ. G is a directed acyclic graph (DAG) composed of V and an edge set E. θ is the joint probability distribution of each node calculated based on the conditional independence of different variables defined in the DAG, serving to quantify the entire BN. In the DAG, nodes correspond to ore-controlling factors and known gold ore bodies, and directed edges between nodes correspond to the relationships between ore-controlling factor nodes. Let V = {V1, V2, ..., V...} N Let} be the set of ore-controlling factors and known gold ore bodies, corresponding to the nodes in the DAG. Let the set of geological variables be X = {X1, X2, ..., X...} N}, for each V i All are associated with a random geological variable X i Each X i All of them consist of binary discrete data derived from the binary evidence factor model;

[0061] Step S22: Calculate the posterior probability of the correlation between ore-controlling factors; the posterior probability of a topology G composed of ore-controlling factors and known gold orebody nodes is calculated using the following formula during the iterative search process:

[0062]

[0063] Where p(G|X) is the posterior probability of the topology; p(G) is calculated based on all possible topologies (G∈ζ). n The prior of ) is a uniform distribution; p(X) is the probability of binary discrete data; p(X|G) is the marginal likelihood, which is the probability of generating corresponding combinations of binary discrete data under a given topological structure.

[0064] In Bayesian inference, p(X) is the integration constant, and its calculation formula is as follows:

[0065]

[0066] Where ζ n It is the set of all possible DAG structures.

[0067] Step S23: Calculate the score of the topological structure among the ore-controlling factors to screen the topological structures; for each topological structure, calculate a score to evaluate the fit between the topological structure and the binary discrete data to screen the topological structures. Since BN employs the independence assumption and non-information graph prior, the marginal likelihood function can be decomposed into local marginal probabilities, i.e., for each factor X... vlocal score score v of the product:

[0068]

[0069] where score i (Pa G (X i )|X) is the score of node i, Pa G (X i ) is the parent set of node i.

[0070] The essence of the problem of topology selection is what kind of criteria to judge its pros and cons. The scoring function is used to judge whether a topology structure fits the binary discrete sample data or not. The commonly used scoring function is:

[0071] (1) BDe (Bayesian Dirichlet) scoring function: combine the prior knowledge of topology structure, and then use Bayesian method. In fact, it is to find the network structure with the maximum posterior probability by using data and prior, which is expressed as:

[0072] BIC(G|X) = ln(p(G)p(θ|G)) = ln p(G) + ln p(X|G) (4)

[0073] where G is the model structure, and X is the training data.

[0074] (2) BIC scoring function: Bayesian Information Criterion (Bayesian Information Criterion) is an approximation of the marginal likelihood function under the premise of large sample. Due to its intuitive meaning and convenient use, it is the most commonly used scoring function in practice. The basic idea is to expand the log-likelihood function near the maximum likelihood trajectory, and then convert the calculation to the integral of a multivariate normal distribution function in the neighborhood of the extreme point. The BIC scoring function is expressed as:

[0075]

[0076] where, for a node V i , the number of possible values is r i , and since this time the binary discrete data is used, the value is 2; m ij is the sample number of the jth value combination of the parent set of node V i ; m ijk is the sample number of the jth value combination of the parent set of node V iThe number of samples taking the kth value under the condition of parent node set value combination determination. The first item of BIC score function represents the log-likelihood of the model, which measures the fitting degree of the structure and the data, and the latter is the penalty term of the model complexity, which prevents overfitting. By setting the penalty term, BIC score function finds a balance between fitting degree (accuracy) and network structure complexity. Since BIC score is relatively simple and balanced between accuracy and complexity, it is often used in actual structure learning problems.

[0077] Generally, Bayesian-based score functions (such as BDe score function) can better distinguish Bayesian network structures with large training samples; for small sample training data, information theory-based score functions, especially BIC score functions, can learn better structures. In order to solve the problem of class imbalance, non-mineral data is thinned out, so it is more appropriate to use BIC score function.

[0078] Step S24, search and evaluate the topological structure to determine the correlation between ore-controlling factors; search the BN topological structure using a search algorithm, and then evaluate the topological structure using a score function to find the correlation between ore-controlling factors with the highest score. Since the data volume of ore-controlling factors in three-dimensional geological mineralization prediction is large, when the data volume is sufficient, the influence of the initial topological structure on the final result can be ignored, so the random topological structure is used as the starting point of the search.

[0079] In order to solve the problem that the derivation process of the topological structure score in the Bayesian structure learning process involves high-dimensional integration, MCMC method is used in the search process.

[0080] MCMC includes two parts: Markov chain and Monte Carlo sampling.

[0081] The purpose is to construct a stationary Markov chain, which finally converges to the posterior distribution of the given binary discrete data.

[0082] In Bayesian structure learning, MCMC method usually uses Metropolis-Hastings sampling algorithm, which takes the posterior distribution p(G|X) of the topological structure as the target distribution of the Markov chain. In each step of the Markov chain, a topological structure G0 is proposed, and the acceptance probability μ of G0 replacing the current correlation G is:

[0083] μ = min{1, R(G0, G)} (6);

[0084] Wherein:

[0085]

[0086] Here, Q(G|G0) is the forward move probability, i.e., the probability of transitioning from state G to state G0; Q(G0|G) is the reverse move probability, i.e., the probability of transitioning from state G0 to state G in the topology. The ratio of the two probabilities is called the Hastings factor, which corrects for any asymmetry that may exist during the proposed move.

[0087] Let nbg(G) be the neighborhood of G, that is, the set of topologies that can be reached from G in one step, including G itself. When When G0 ∈ nbg(G), Q(G|G0) is 0. When G0 ∈ nbg(G), Q(G|G0) should be greater than 0. This rule also applies to the probability of reverse movement. Therefore, the principle for constructing nbg(G) is that the transition matrix and the Markov chain are irreducible, that is, there is a positive probability of transitioning from any topology in the search space to other topologies.

[0088] Step S25: Construct the neighborhood space of the topological structure between factors; to achieve irreducibility, a movement operator is used, namely edge addition, deletion, and reversal, to realize irreducibility and accelerate convergence. In practical applications, the topological structure G utilizes a...

[0089] An n×n adjacency matrix A represents that if V j It is V i If the node is the parent of the node, then A(i,j) is 1; otherwise, it is 0. The rules for adding, deleting, and reversing will be explained below:

[0090] (a) Add V when A(j, i) is 0. j →V i It is valid, but when A(j, i) is 1, it means V i It is V j The parent node, and a directed path already exists between them. If V is added... j →V i This would create a directed circular path, making the addition invalid.

[0091] (b) Deletion: Deleting edges will not form any directed circular paths, so there are no restrictions on the deletion operation.

[0092] (c) Reverse: The first step is to delete the edge to be reversed; the second step is to add an edge in the opposite direction. Let V be an example. j →V i For the edges that need to be reversed, to prevent this step from introducing a loop path, V needs to be checked. i Is it V? j If it is not the parent node, then edge reversal can be performed.

[0093] Step S26, screen topology and calculate conditional probability distribution table Build the correlation between ore-controlling factors;After building a smooth Markov chain, the topology with the highest frequency of occurrence is the final determined topology G, Figure 4 For the correlation between ore-controlling factors obtained by training, different linear representations have different strengths of correlation. Then, the conditional probability distribution table G can be calculated by maximum likelihood estimation, Figure 6 For conditional probability distribution diagram instance, event X and event are opposite events, and the probability of their occurrence is 1. The probability of event occurrence is only listed in the figure, and the condition combination with event occurrence probability of 0 is excluded. The conditional probability distribution diagram can clearly reflect the influence relationship between geological elements, reflect the correlation between ore-controlling factors in three-dimensional metallogenic prediction and its strength, and provide strong basis for explaining complex ore-controlling action.

[0094] Step S3, calculate the conditional probability distribution table by maximum likelihood estimation;The specific steps are as follows,

[0095] Step S31 takes the node conditional probability as the target estimated value θ of maximum likelihood estimation;

[0096] Step S32 calculates the likelihood function L(θ);

[0097] Step S33 takes the logarithm lnL(θ) of the likelihood function;

[0098] Step S34 takes the derivative of the likelihood function;

[0099] Step S35 obtains the maximum value to obtain the conditional probability θ.

[0100] The application takes the binary model of ore-controlling factors as input, constantly searches for new topologies through transition operators and scoring functions, determines whether the state is transferred according to the acceptance probability in Monte Carlo sampling, finally obtains a smooth Markov chain, and the topology with the highest frequency of occurrence in the chain is the final correlation between input ore-controlling factors. After calculating the conditional probability distribution table by maximum likelihood estimation for metallogenic posterior probability calculation, the correlation between factors can be quantified, and the conditional probability distribution table generated according to the correlation can be used as the calculation basis of metallogenic posterior probability, which is more in line with the geological metallogenic regularity, has stronger explanation, and is convenient for subsequent verification.

[0101] The application also provides a device for determining the correlation between ore-controlling factors in three-dimensional metallogenic prediction, which comprises a memory, a control processor and a computer program stored in the memory and executable on the control processor, and the control processor executes the program to realize the method for determining the correlation between ore-controlling factors in three-dimensional metallogenic prediction.

[0102] The present invention also provides a computer-readable storage medium storing computer-executable instructions for causing a computer to execute the aforementioned method for determining the correlation of ore-controlling factors in three-dimensional mineralization prediction. Examples of computer-readable storage media include: read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-RLTH, etc. BD-RE, Blu-ray or optical disc storage, hard disk drive (HDD), solid-state drive (SSD), card storage (such as multimedia card, secure digital (SD) card, or ultra-fast digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state drive, and any other device configured to store computer programs and any associated data, data files, and data structures in a non-transitory manner and to provide the computer programs and any associated data, data files, and data structures to a processor or computer so that the processor or computer can execute the computer programs.

[0103] In one example, the computer program, along with any associated data, data files, and data structures, is distributed across a networked computer system, allowing the computer program, along with any associated data, data files, and data structures, to be stored, accessed, and executed in a distributed manner via one or more processors or computers.

[0104] Taking the main hydrothermal mineralization activities in a certain region as an example (e.g.) Figure 2 As shown, the mineralization factors include: 1) NNE and NW trending faults; 2) volcanic activity; and 3) tectonic control. The soil exhibits distinct gold and silver anomalous areas, primarily caused by the silicification of mylonite and alteration of pyrite along the faults. Volcanic activity mainly disrupts the original geological bodies and local gold enrichment, facilitating element migration during subsequent geological activities. Key tectonic controls on mineralization include: tectonic deformation, driving fluid flow, ore-trapping fluids, and / or the influence of metal deposition under fluctuating pressure conditions in fault zones on rock permeability. The binary ore-controlling factor data are input into the node dependencies and conditional probability distribution table obtained from Bayesian-MCMC structure learning.

[0105] Compared with other machine learning methods, the Bayesian network adopted by the application can not only reflect the relationship between ore-controlling factors and known ore bodies, but also explore the complex correlation between ore-controlling factors, decompose the joint distribution into multiple simple probability distributions by using variable conditional independence relationship, thereby reducing the model complexity and improving the reasoning efficiency.

[0106] Compared with the prior art, the technical scheme provided by the application is a method for determining and quantifying the correlation between ore-controlling factors based on Bayesian-MCMC structure learning, which can directly learn the correlation between ore-controlling factors that fits the sample data according to the input discrete attribute data, quantize the correlation between ore-controlling factors in three-dimensional metallogenic prediction and its strength, and provide a strong basis for explaining the complex metallogenic process. The process only relies on data and does not involve subjective experience.

[0107] The above merely describes preferred embodiments of the application and is not intended to limit the application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A method for determining the association relationship of ore-controlling factors in three-dimensional metallogenic prediction, characterized in that, The method comprises the following steps: Step S1, constructing an ore-forming factor based on three-dimensional geological modeling according to multi-source geosciences data; The method comprises the following steps: Step S11, obtaining modeling data; Step S12, mapping data from two dimensions to three dimensions, mapping point data and line data from two dimensions to three dimensions by superimposing a digital elevation model (DEM), and obtaining point set data and line set data in three dimensions; Step S13, for geological data, constructing a corresponding three-dimensional geological structure model in the order of line, surface and volume by using explicit modeling, and finally constructing a three-dimensional field model by using three-dimensional spatial analysis, and for geological attribute data, constructing an attribute model by using a spatial interpolation method; Step S14, constructing an ore-forming favorable degree conceptual model, determining ore-controlling factors by analyzing the spatial relationship between known ore deposits and three-dimensional geological entity models and three-dimensional geological attribute models, and constructing a three-dimensional ore-controlling factor model in combination with existing experience and knowledge; Step S15, determining a threshold value of the ore-controlling factor by using the ore-forming conceptual model in combination with a data-driven method, assigning a value of 1 to a region meeting the ore-controlling threshold value and a value of 0 to a region not meeting the ore-controlling threshold value, and constructing a binary ore-controlling factor model; Step S2, determining the association relationship between ore-controlling factors based on Bayesian-MCMC structure learning; The method comprises the following steps: Step S21, constructing an ore-controlling factor node; Step S22, calculating the posterior probability of the association relationship between ore-controlling factors; Step S23, calculating a score of the topological structure between ore-controlling factors to screen the topological structure; Step S24, searching and evaluating the topological structure to determine the association relationship between ore-controlling factors; Step S25, constructing a neighborhood space of the topological structure between factors; Step S26, screening the topological structure; Step S27, taking the conditional probability of each node as an estimated value to be solved in maximum likelihood estimation, and determining the conditional probability of the node by calculating the extreme value; Step S3, calculating a conditional probability distribution table by using maximum likelihood estimation; The method comprises the following steps: Step S31, taking the node conditional probability as a target estimated value θ of maximum likelihood estimation; Step S32, calculating a likelihood function L(θ); Step S33, taking a logarithm ln L(θ) of the likelihood function; Step S34, calculating a derivative of the likelihood function; Step S35, calculating a maximum value to obtain the conditional probability θ.

2. The method of claim 1, wherein the method further comprises: determining the association relationship between the ore-controlling factors in the three-dimensional metallogenic prediction. In step S21, the Bayesian network used to describe the correlation between the ore-controlling factors and the known gold ore bodies consists of two parts: the topological structure G and the conditional probability distribution table θ. G is a DAG composed of V and edge set E, and θ is the joint probability distribution of each node calculated based on the conditional independence between different variables defined by the DAG. Let V = {V1, V2,..., V N} be the set of ore-controlling factors and known gold ore bodies, corresponding to the nodes in the DAG, and let the set of geological variables be X = {X1, X2,..., X N}. For each V i , there is an associated random geological variable X i , and each X i is composed of binary discrete data derived from a binary evidence factor model.

3. The method of claim 2, wherein the method further comprises: determining the association relationship between the ore-controlling factors in the three-dimensional metallogenic prediction. In step S22, the topological structure GG composed of ore-controlling factors and known gold deposit nodes has the following posterior probability calculation formula in the iterative search process: where p(G|X) is the posterior probability of the topology; p(G) is the prior calculated based on all possible topologies, where G e z n , a uniform distribution is used as the prior; p(X) is the probability of binary discrete data; p(X|G) is the marginal likelihood, which is the likelihood of the corresponding combination of binary discrete data generated under the condition of determining the topology. In Bayesian inference, p(X) is the integral constant, and the calculation formula is as follows: where ζ n is the set of all possible DAG structures.

4. The method of claim 2 or 3, wherein, In step S23, for each topology, a score is calculated to evaluate the degree of fitting of the topology to the binary discrete data to screen the topology. Since the BN adopts the independence assumption and the non-informative graph prior, the marginal likelihood function can be decomposed into the product of local marginal probabilities, i.e. given the parent nodes, each factor X v has a local score score v : where score i (Pa G (X i )|X) is the score of node i, Pa G (X i ) is the parent set of node i.

5. The method of claim 4, wherein the method further comprises: determining the association relationship between the ore-controlling factors in the three-dimensional metallogenic prediction. In step S11, the modeling data sources include but are not limited to plane geological maps, profile geological maps, stratum boundary lines, fault boundary lines, apparent resistivity, magnetic anomaly and residual gravity density attribute sampling point data.

6. A device for determining the association relationship of ore-controlling factors in three-dimensional metallogenic prediction, characterized in that, A computer program product, comprising a memory, a control processor and a computer program stored on the memory and executable on the control processor, wherein the control processor executes the program to implement the method for determining the association relationship between ore-controlling factors in three-dimensional ore prediction according to any one of claims 1-5.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer executable instructions for causing a computer to execute the method for determining the correlation of ore-controlling factors in three-dimensional metallogenic prognosis according to any one of claims 1-5.

Images