Intelligent mineralization prediction method based on deep learning

Abstract

The invention discloses an intelligent prospecting prediction method based on deep learning, and the method comprises the following steps: collecting and sorting related information of geological information, geophysical information, geochemical information and remote sensing information of a research region, constructing a uranium resource exploration multivariate information database, and extracting information features to establish a uranium resource data set; constructing a uranium resource machine learning model, and training the uranium resource machine learning model by using the uranium resource data set until the output result of the uranium resource machine learning model meets the expectation; the uranium resource machine learning model outputs the similarity between the position of the central point of the prediction point of the research area and the known ore deposit / ore deposit, the uranium resource machine learning model is used for conducting favorable area prediction on the research area, and a uranium mineralization favorable area of the research area is determined; according to the method, the metallogenic condition can be analyzed from a more comprehensive angle, and the limitations of single data source and incomplete information in the prior art are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of mineralization prediction, and in particular to an intelligent mineralization prediction method based on deep learning. Background Art

[0002] In recent years, with the explosive growth of geological data and the rapid development of computer science and technology, new artificial intelligence methods and technologies represented by machine learning, especially deep learning, have gradually been applied to the field of geological research. At present, the application research of artificial intelligence methods in geochemistry and geophysics is relatively extensive. In the field of mineral resource exploration and evaluation, artificial intelligence and big data technologies are also gradually being used. Artificial intelligence methods can assist geologists in geological analysis, data interpretation, and modeling research, and even realize automatic rapid modeling and simulation testing after giving "expert" opinions based on artificial intelligence. This functional rapid iteration work mode that combines artificial intelligence with professional analysis software has greatly improved work efficiency.

[0003] At present, uranium geology has a certain accumulation in terms of data foundation. When existing deep learning-based mineralization prediction work processes complex geological data, traditional methods have problems such as limited data dimensions and difficulty in feature selection. This may lead to the inability to fully mine all important information related to mineralization, thereby affecting the accuracy of the prediction. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent mineralization prediction method based on deep learning to solve the technical problem that the existing technology cannot fully mine all important information related to mineralization, thereby affecting the accuracy of prediction.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0006] An intelligent mineralization prediction method based on deep learning, comprising the following steps:

[0007] Step 100: Collect and organize relevant geological information, geophysical information, geochemical information, and remote sensing information of the study area, construct a multi-information database for uranium resource exploration, classify, design, and constrain various attribute items of the multi-information database for uranium resource exploration, and extract information features to establish a uranium resource dataset;

[0008] Step 200: construct a uranium resource machine learning model, and train the uranium resource machine learning model using the uranium resource dataset until the output result of the uranium resource machine learning model meets expectations;

[0009] Step 300: The uranium resource machine learning model outputs the similarity between the center point position of the predicted point in the study area and the known ore deposit / ore point, and uses the uranium resource machine learning model to predict the favorable area of ​​the study area to determine the favorable area for uranium mineralization in the study area.

[0010] As a preferred solution of the present invention, in step 100, the relevant information of the multi-source data of geological information, geophysical information, geochemical information and remote sensing characteristic information of the study area, the relevant information of the multi-source data specifically includes drilling database information, documents and maps.

[0011] As a preferred solution of the present invention, in step 100, the method for establishing the uranium resource dataset is as follows:

[0012] Classify, design and constrain the various attribute items of the uranium resource exploration multi-information database, and complete the construction of various attribute items of the uranium resource dataset in the study area based on the data items and subordinate terms of the uranium resource exploration multi-information database;

[0013] Establish identification items based on classification applications and perform sample labeling on each data in the uranium resource dataset;

[0014] After the uranium resource dataset is constructed, the uranium resource dataset is tested, modified, and improved until each data in the uranium resource dataset is adjusted to standardized data.

[0015] As a preferred embodiment of the present invention, when classifying, designing and constraining various attribute items of the uranium resource exploration multivariate information database, the classified attribute items include geographical location, lithology description, mineral composition, geochemical information and geophysical information;

[0016] Perform data cleaning on the borehole database to ensure integrity, and perform secondary classification and feature design on lithology description, mineral composition, geochemical information, geophysical information, and geographic location to form sample features corresponding to each borehole label;

[0017] Construct feature vectors, standardize the numerical data in the drilling database, and label the data samples in the drilling database to distinguish between mineralized areas and non-mineralized areas.

[0018] As a preferred solution of the present invention, in step 200, the uranium resource dataset is divided into a training set and a test set, and the method for training the uranium resource machine learning model using the training set is as follows:

[0019] Inputting sample features corresponding to the borehole labels into an input end of the uranium resource machine learning model, where the sample features corresponding to the borehole labels include geological features, geophysical features, geochemical features, and remote sensing features;

[0020] The uranium resource machine learning model outputs a mineralization prediction result for the study area based on the sample characteristics corresponding to the drill hole labels. The mineralization prediction result for the study area includes mineralization potential evaluation, uranium deposit type classification or uranium leaching amount prediction.

[0021] As a preferred solution of the present invention, when judging whether the mineralization prediction result of the study area output by the uranium resource machine learning model meets the expectation, the specific implementation method is: measuring the classification performance of the uranium resource machine learning model by evaluating the accuracy, recall rate, ROC curve and AUC value indicators of mineralization potential evaluation, uranium deposit type classification or uranium leaching amount prediction, or by using the root mean square error and R 2 The value is used to evaluate the prediction accuracy of the uranium resource machine learning model for mineralization potential evaluation, uranium deposit type classification or uranium leaching amount.

[0022] As a preferred solution of the present invention, in step 300, the method for predicting favorable areas using the trained uranium resource machine learning model is as follows:

[0023] First, the geological data, geophysical information, and geochemical information of known mineral deposits in the study area are input into the uranium resource machine learning model. The difference between the output data of the trained uranium resource machine learning model and the corresponding matched and known characteristic data of the known mineral deposits is determined to verify the reliability of the uranium resource machine learning model.

[0024] Then, regular prediction points are generated in the study area, key features of the geographical location, lithological description, mineral composition, geochemistry and geophysical data of the regular prediction points are extracted, and data cleaning work is performed on the key features;

[0025] The key feature input values ​​are input into the uranium resource machine learning model trained, the uranium resource machine learning model is used to calculate the mineralization probability, and the mineralization probability distribution map is generated by interpolation, and finally the favorable uranium mineralization area is delineated.

[0026] As a preferred embodiment of the present invention, in step 300, geological data, geophysical information, and geochemical information of known mineral deposits in the study area are collected, and the mineralization characteristics corresponding to the geological data, geophysical information, and geochemical information of the known mineral deposits are analyzed;

[0027] Using the hierarchical analysis method or the characteristic analysis method, the metallogenic characteristics of the known ore deposits are used as the input parameters of the uranium resource machine learning model. The similarity between the output data of the uranium resource machine learning model and the corresponding matching feature data of the known ore deposits is calculated and studied. The reliability of the uranium resource machine learning model is verified by comparing the output data of the uranium resource machine learning model with the corresponding matching feature data of the known mineral points.

[0028] As a preferred solution of the present invention, when calculating the similarity between the center of the predicted point in the study area and the known mineral deposits / ore points, the geological characteristics, geochemical anomalies, geophysical characteristics and spatial position relationships are formed into multi-source data;

[0029] The multi-source data are weighted for comprehensive evaluation through the hierarchical analysis method, or the probability value of the predicted point around the known mineral deposit is calculated using the Gaussian distribution method, so as to quantify the degree of similarity and provide a basis for uranium mine prediction.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The uranium resource machine learning model provided by the present invention can analyze mineralization conditions from a more comprehensive perspective (such as multi-source data formed by the geological characteristics, geophysical characteristics, geochemical characteristics and remote sensing characteristics of the study area), overcoming the limitations of the existing technology of single data source and incomplete information.

[0032] Furthermore, this implementation utilizes a deep learning model with powerful feature extraction and learning capabilities, enabling it to automatically mine potential mineralization patterns and correlations from large amounts of complex geological data. Compared to traditional manual feature extraction methods, this approach is more efficient and accurate, enabling it to uncover more valuable information hidden within the data. Compared to traditional manual analysis and modeling methods, this approach can process large amounts of data more quickly, improving work efficiency and shortening exploration cycles. This allows geologists to obtain mineralization predictions more promptly, providing support for exploration decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0034] Figure 1 Schematic diagram of the process of the intelligent mineralization prediction method according to an embodiment of the present invention;

[0035] Figure 2A schematic diagram of a favorable mineralization area predicted by a uranium resource machine learning model according to an embodiment of the present invention; DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] like Figure 1 As shown, the present invention provides an intelligent mineralization prediction method based on deep learning, comprising the following steps:

[0038] Step 100: Collect and organize geological information, geophysical information, geochemical information, and remote sensing information of the study area, construct a multi-dimensional information database for uranium resource exploration, classify, design, and constrain various attribute items of the multi-dimensional information database for uranium resource exploration, and extract information features to establish a uranium resource dataset;

[0039] Step 200: Build a uranium resource machine learning model, and use the uranium resource dataset to train the uranium resource machine learning model until the output result of the uranium resource machine learning model meets expectations;

[0040] Step 300: The uranium resource machine learning model outputs the similarity between the center point of the predicted point in the study area and the known ore deposits / ore points, and uses the uranium resource machine learning model to predict the favorable area of ​​the study area to determine the favorable area for uranium mineralization in the study area.

[0041] The uranium resource machine learning model in this embodiment can analyze mineralization conditions from a more comprehensive perspective (such as multi-source data formed by the geological characteristics, geophysical characteristics, geochemical characteristics and remote sensing characteristics of the study area), overcoming the limitations of the existing technology of single data source and incomplete information.

[0042] Furthermore, this implementation utilizes a deep learning model with powerful feature extraction and learning capabilities, enabling it to automatically mine potential mineralization patterns and correlations from large amounts of complex geological data. Compared to traditional manual feature extraction methods, this approach is more efficient and accurate, enabling it to uncover more valuable information hidden within the data. Compared to traditional manual analysis and modeling methods, this approach can process large amounts of data more quickly, improving work efficiency and shortening exploration cycles. This allows geologists to obtain mineralization predictions more promptly, providing support for exploration decisions.

[0043] In step 100, the geological information, geophysical information, geochemical information and remote sensing information of the study area form relevant information of multi-source data, and the relevant information of the multi-source data specifically includes borehole database information, documents and maps.

[0044] The specific parameters of the uranium resource exploration multivariate information database are shown in Table 1 below.

[0045] Table 1 Multivariate information database of uranium resource exploration of Yahewan uranium deposit

[0046]

[0047] By classifying and processing the multivariate information database of uranium resource exploration, a uranium resource dataset is established. The specific implementation method is as follows:

[0048] The attribute items of the uranium resource exploration multi-information database are classified, designed and constrained. According to the data items and subordinate terms of the uranium resource exploration multi-information database, the construction of various attribute items of the geophysical, chemical, remote sensing, mineral and environmental attributes of the uranium resource data set in the study area is completed.

[0049] Identification items are established based on classification applications, and sample labeling is performed on each data in the uranium resource dataset.

[0050] After the uranium resource dataset is completed, it will be tested, modified, and improved until all data in the uranium resource dataset are adjusted to standardized data.

[0051] Specifically, when classifying, designing and constraining various attribute items of the uranium resource exploration multivariate information database, the classified attribute items include geographical location, lithology description, mineral composition, geochemistry and geophysical data;

[0052] The drilling database data is cleaned to ensure integrity, and secondary classification and feature design are performed on lithology description, mineral composition, geochemical information, geophysical information, and geographic location to form sample features corresponding to each drilling label;

[0053] Construct feature vectors, standardize the numerical data in the borehole database, label the data samples in the borehole database, and determine the sample characteristics corresponding to each borehole in the study area. Based on the sample characteristics of each borehole and the location of each borehole, distinguish between mineralized areas and non-mineralized areas.

[0054] The parameters of the uranium resource dataset are shown in Table 2 below.

[0055] Table 2 Uranium resource dataset of the Yahewan uranium deposit

[0056]

[0057] In step 200, the uranium resource dataset is divided into a training set and a test set. The method for training the uranium resource machine learning model using the training set is as follows:

[0058] Input the sample features corresponding to each drillhole label into the input of the uranium resource machine learning model. The sample features include geological features, geophysical features, geochemical features, and remote sensing features.

[0059] The uranium resource machine learning model outputs mineralization prediction results for the study area based on the sample characteristics corresponding to the drill hole labels. The mineralization prediction results for the study area include mineralization potential evaluation, uranium deposit type classification or uranium leaching amount prediction.

[0060] When judging whether the mineralization prediction results of the uranium resource machine learning model for the study area meet expectations, the specific implementation method is as follows:

[0061] The classification performance of the uranium resource machine learning model is measured by evaluating the accuracy, recall, ROC curve and AUC value of mineralization potential evaluation, uranium deposit type classification or uranium leaching amount prediction, or by the root mean square error and R 2 The value is used to evaluate the prediction accuracy of uranium resource machine learning models for mineralization potential evaluation, uranium deposit type classification or uranium leaching amount.

[0062] In step 400, the method for predicting favorable areas using the trained uranium resource machine learning model is as follows:

[0063] First, the reliability of the uranium resource machine learning model is verified by determining the difference between the output data of the trained uranium resource machine learning model and the corresponding matching feature data of the known deposits in the study area through geological data, geophysical information, and geochemical information.

[0064] Then, regular prediction points are generated in the study area, key features of the geographical location, lithological description, mineral composition, geochemistry and geophysical data of the regular prediction points are extracted, and data cleaning work is performed on the key features;

[0065] The key feature input values ​​are trained into the uranium resource machine learning model, the uranium resource machine learning model is used to calculate the probability of mineralization, and the mineralization probability distribution map is generated through interpolation, and finally the favorable uranium mineralization areas are delineated.

[0066] In step 400, geological data, geophysical information, and geochemical information of known mineral deposits in the study area are collected, and the mineralization characteristics corresponding to the geological data, geophysical information, and geochemical information of the known mineral deposits are analyzed;

[0067] Using the hierarchical analysis method or the characteristic analysis method, the metallogenic characteristics of known deposits are used as the input parameters of the uranium resource machine learning model. The similarity between the output data of the uranium resource machine learning model and the corresponding matching feature data of the known deposits is calculated. The reliability of the uranium resource machine learning model is verified by comparing the output data of the uranium resource machine learning model with the corresponding matching feature data of the known mineral points.

[0068] When calculating the similarity between the center of the predicted point in the study area and the known deposits / ore points, the geological characteristics, geochemical anomalies, geophysical characteristics and spatial position relationships are formed into multi-source data;

[0069] The multi-source data are weighted for comprehensive evaluation through the hierarchical analysis method, or the probability value of the predicted point around the known mineral deposit is calculated using the Gaussian distribution method, so as to quantify the degree of similarity and provide a basis for uranium mine prediction.

[0070] The uranium resource machine learning model outputs the similarity between the center point of the predicted point and the known ore deposit / ore point (the probability value of the predicted point). The value is interpolated to obtain the favorable uranium mineralization area in the demonstration area, such as Figure 2 As shown, the similarity value between the center point of the predicted point indicated by blue and the known ore deposit / ore point is small, and the similarity value between the center point of the predicted point indicated by red and the known ore deposit / ore point is large. Therefore, blue represents a small probability of having a mineral in the study area, and red represents a high probability of having a mineral in the study area. Therefore, the probability of having a mineral in the study area increases from small to large in order from left to right.

[0071] The uranium resource machine learning model in this embodiment can analyze mineralization conditions from a more comprehensive perspective (such as multi-source data formed by the geological characteristics, geophysical characteristics, geochemical characteristics and remote sensing characteristics of the study area), overcoming the limitations of the existing technology of single data source and incomplete information.

[0072] Furthermore, this implementation utilizes a deep learning model with powerful feature extraction and learning capabilities, enabling it to automatically mine potential mineralization patterns and correlations from large amounts of complex geological data. Compared to traditional manual feature extraction methods, this approach is more efficient and accurate, enabling it to uncover more valuable information hidden within the data. Compared to traditional manual analysis and modeling methods, this approach can process large amounts of data more quickly, improving work efficiency and shortening exploration cycles. This allows geologists to obtain mineralization predictions more promptly, providing support for exploration decisions.

[0073] Moreover, this flexibility and adaptability enables the mineralization prediction method to be continuously improved and optimized, maintaining a high level of prediction accuracy and reliability, and providing long-term and stable technical support for uranium resource exploration.

[0074] The above embodiments are merely exemplary embodiments of the present application and are not intended to limit the scope of the present application. The scope of protection of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and scope of protection of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present application.

Claims

1. An intelligent mineralization prediction method based on deep learning, characterized in that: The following steps are involved: Step 100: Collect and organize relevant geological information, geophysical information, geochemical information, and remote sensing information of the study area, construct a multi-dimensional information database for uranium resource exploration, classify, design, and constrain various attribute items of the multi-dimensional information database for uranium resource exploration, and extract information features to establish a uranium resource dataset; Step 200: construct a uranium resource machine learning model, and train the uranium resource machine learning model using the uranium resource dataset until the output result of the uranium resource machine learning model meets expectations; Step 300: The uranium resource machine learning model outputs the similarity between the center point position of the predicted point in the study area and the known ore deposit / ore point, and uses the uranium resource machine learning model to predict the favorable area of ​​the study area to determine the favorable area for uranium mineralization in the study area.

2. The intelligent mineralization prediction method based on deep learning according to claim 1, characterized in that: In step 100, the multi-source data related information of the geological information, geophysical information, geochemical information and remote sensing characteristic information of the study area, the multi-source data related information specifically includes drilling database information, documents and maps.

3. The intelligent mineralization prediction method based on deep learning according to claim 2 is characterized in that: In step 100, the method for establishing the uranium resource dataset is as follows: Classify, design and constrain the various attribute items of the uranium resource exploration multi-information database, and complete the construction of various attribute items of the geophysical, chemical, remote sensing, mineral and environmental of the uranium resource dataset in the study area according to the data items and subordinate terms of the uranium resource exploration multi-information database; Establish identification items based on classification applications and perform sample labeling on each data in the uranium resource dataset; After the uranium resource dataset is constructed, the uranium resource dataset is tested, modified, and improved until each data in the uranium resource dataset is adjusted to standardized data.

4. The intelligent mineralization prediction method based on deep learning according to claim 3 is characterized in that: When classifying, designing and constraining various attribute items of the uranium resource exploration multivariate information database, the classified attribute items include geographical location, lithologic description, mineral composition, geochemical information and geophysical information; Perform data cleaning on the borehole database to ensure integrity, and perform secondary classification and feature design on lithology description, mineral composition, geochemical information, geophysical information, and geographic location to form sample features corresponding to each borehole label; Construct feature vectors, standardize the numerical data in the drilling database, and label the data samples in the drilling database to distinguish between mineralized areas and non-mineralized areas.

5. The intelligent mineralization prediction method based on deep learning according to claim 4 is characterized in that: In step 200, the uranium resource dataset is divided into a training set and a test set. The method for training the uranium resource machine learning model using the training set is as follows: Inputting sample features corresponding to the borehole labels into an input end of the uranium resource machine learning model, where the sample features corresponding to the borehole labels include geological features, geophysical features, geochemical features, and remote sensing features; The uranium resource machine learning model outputs a mineralization prediction result for the study area based on the sample characteristics corresponding to the drill hole labels. The mineralization prediction result for the study area includes mineralization potential evaluation, uranium deposit type classification or uranium leaching amount prediction.

6. The intelligent mineralization prediction method based on deep learning according to claim 5 is characterized in that: When judging whether the mineralization prediction results of the study area output by the uranium resource machine learning model meet expectations, the specific implementation method is to measure the classification performance of the uranium resource machine learning model by evaluating the accuracy, recall rate, ROC curve and AUC value indicators of mineralization potential evaluation, uranium deposit type classification or uranium leaching amount prediction, or by using the root mean square error and R 2 The value is used to evaluate the prediction accuracy of the uranium resource machine learning model for mineralization potential evaluation, uranium deposit type classification or uranium leaching amount.

7. The intelligent mineralization prediction method based on deep learning according to claim 5 is characterized in that: In step 300, the method for predicting favorable areas using the trained uranium resource machine learning model is as follows: First, the geological data, geophysical information, and geochemical information of known mineral deposits in the study area are input into the uranium resource machine learning model. The difference between the output data of the trained uranium resource machine learning model and the corresponding matched and known characteristic data of the known mineral deposits is determined to verify the reliability of the uranium resource machine learning model. Then, regular prediction points are generated in the study area, key features of the geographical location, lithology description, mineral composition, geochemistry and geophysical data of the regular prediction points are extracted, and data cleaning work is performed on the key features; The key feature input values ​​are input into the uranium resource machine learning model trained, the uranium resource machine learning model is used to calculate the mineralization probability, and the mineralization probability distribution map is generated by interpolation, and finally the favorable uranium mineralization area is delineated.

8. The intelligent mineralization prediction method based on deep learning according to claim 4 is characterized in that: In step 300, geological data, geophysical information, and geochemical information of known mineral deposits in the study area are collected, and the mineralization characteristics corresponding to the geological data, geophysical information, and geochemical information of the known mineral deposits are analyzed; Using the hierarchical analysis method or the characteristic analysis method, the metallogenic characteristics of the known ore deposits are used as the input parameters of the uranium resource machine learning model. The similarity between the output data of the uranium resource machine learning model and the corresponding matching feature data of the known ore deposits is calculated and studied. The reliability of the uranium resource machine learning model is verified by comparing the output data of the uranium resource machine learning model with the corresponding matching feature data of the known mineral points.

9. The intelligent mineralization prediction method based on deep learning according to claim 8, characterized in that: When calculating the similarity between the center of the predicted point in the study area and the known deposits / ore points, the geological characteristics, geochemical anomalies, geophysical characteristics and spatial position relationships are formed into multi-source data; The multi-source data are weighted for comprehensive evaluation through the hierarchical analysis method, or the probability value of the predicted point around the known mineral deposit is calculated using the Gaussian distribution method, so as to quantify the degree of similarity and provide a basis for uranium mine prediction.

Images