Sample set construction method suitable for sandstone type uranium mine big data prediction

Abstract

The invention belongs to the technical field of sandstone type uranium ore mineralization prediction, and particularly relates to a sample set construction method suitable for sandstone type uranium ore big data prediction, which comprises the following steps: a system collects and arranges geological-geophysical prospecting-geochemical prospecting-remote sensing data of a research area, extracts the geological, geophysical prospecting, geochemical prospecting and remote sensing data, and pre-processes the data; performing feature selection of a sample set by analyzing factors such as construction, magma activity, stratum and lithofacies paleogeography, geophysics, geochemistry and the like; integrating the selected sample feature data, and marking features and labels of each sample to form complete single sample data; and integrating the formed single sample data together to construct a sample set, performing feature engineering on the data in the sample set, randomly selecting samples to construct a training set, and using the remaining samples to construct a test set. According to the method, the working process of constructing the sandstone type uranium mine big data prediction sample set is optimized, so that the uncertainty in the working process is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of sandstone-type uranium mineralization prediction, and in particular relates to a sample set construction method suitable for sandstone-type uranium deposit big data prediction. Background Art

[0002] Big data prediction technology is currently being gradually applied in the prediction and evaluation of sandstone-type uranium deposits. It is primarily used to explore mineralization patterns within heterogeneous and multi-faceted data, playing a vital role in information synthesis and integration, and achieving promising results. However, its application still faces numerous challenges. The primary issue lies in the "big" aspect of big data prediction. Applications require insight into patterns based on data, but in practice, this approach often operates solely from a data-driven perspective. When constructing big data prediction sample sets, the full set of data is blindly used for analysis and mining. These constructed sample sets lack the expertise of geologists and cannot be corroborated with earlier geological understandings. Consequently, the resulting prediction results are subject to ambiguity, uncertainty, and lack of practical significance. Summary of the Invention

[0003] The purpose of the present invention is to provide a sample set construction method suitable for sandstone-type uranium deposit big data prediction. The method provides a basis for prospecting prediction by analyzing geological background, geophysical and geochemical data in traditional sandstone-type uranium deposit prediction work, integrates the steps of determining sandstone-type uranium deposit prospecting signs into the process of constructing big data prediction sample set, and makes full use of the experience and knowledge of geological experts to optimize the workflow of constructing sandstone-type uranium deposit big data prediction sample set, thereby reducing uncertainty in the working process.

[0004] The technical solution for achieving the purpose of the present invention is as follows:

[0005] A method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits, the method comprising:

[0006] Step 1: Systematically collect and organize geological, geophysical, geochemical and remote sensing data of the study area, extract geological, geophysical, geochemical and remote sensing data, and pre-process the data;

[0007] Step 2: Feature selection of sample sets by analyzing factors such as structure, magmatic activity, stratigraphy and lithofacies paleogeography, geophysics, and geochemistry;

[0008] Step 3: Integrate the selected sample feature data, mark the features and labels of each sample, and form a complete single sample data;

[0009] Step 4: Integrate the formed individual sample data together to construct a sample set, perform feature engineering on the data in the sample set, randomly select samples to construct a training set, and use the remaining samples to construct a test set.

[0010] The step 1 comprises:

[0011] Step 1.1: Define the possible geological environment of typical sandstone-type uranium deposits in the study area;

[0012] Step 1.2: Collect geological, geophysical and geochemical original information and data;

[0013] Step 1.3: Classify and organize the collected raw data and data, and perform preliminary preprocessing on the raw data and data using data cleaning, missing value interpolation, and integration methods;

[0014] Step 1.4: Establish a multi-information database containing geological, geophysical, geochemical and other data.

[0015] The geological, geophysical and geochemical original data and information include field observation data, laboratory analysis results, geophysical measurement original data or interpretation results, and geochemical original data or results data.

[0016] The step 2 includes:

[0017] Step 2.1: Analyze the structural factors related to sandstone-type uranium mineralization in the study area. Analyze the ore-controlling characteristics and favorable mineralization locations of the structural form most closely related to sandstone-type uranium mineralization, including faults, folds, fissures, structural interfaces, and igneous rock structures. Determine the nature, scale, activity period, and ore-control mechanism of this structural form.

[0018] Step 2.2: Analyze the factors of magmatic activity related to sandstone-type uranium mineralization in the study area, analyze the spatial and temporal connections between magmatic rocks and uranium mineralization, the specificity of magmatic mineralization, the physicochemical conditions of magmatic activity, and the genetic connection between magmatic activity and uranium mineralization, and determine the characteristics of magmatic activity closely related to sandstone-type uranium mineralization;

[0019] Step 2.3: Analyze the stratigraphic and lithofacies paleogeographic factors related to sandstone-type uranium mineralization in the study area. Analyze the rock types, sedimentary environments, and paleogeographic factors of the stratigraphic units related to sandstone-type uranium mineralization, and determine the stratigraphic and lithofacies paleogeographic characteristics closely related to sandstone-type uranium mineralization.

[0020] Step 2.4: Analyze the geophysical factors related to sandstone-type uranium mineralization in the study area, use geophysical methods to detect the underground structure and physical properties of rocks related to sandstone-type uranium mineralization, and determine the geophysical field characteristics closely related to sandstone-type uranium mineralization;

[0021] Step 2.5: Analyze the geochemical factors related to sandstone-type uranium mineralization in the study area. Identify the element distribution patterns and abnormal areas related to sandstone-type uranium deposits through geochemical data, and determine the geochemical characteristics closely related to sandstone-type uranium deposits.

[0022] The step 3 includes:

[0023] Step 3.1: Take the existing drill holes in the study area as samples of the data set and classify them according to whether they are sandstone-type uranium industrial holes, mineralized holes, and other drill holes. Sandstone-type uranium industrial holes and mineralized holes are marked as 1, and other drill holes are marked as 0, which are used as labels for whether the samples have mines.

[0024] Step 3.2: Combine the features selected in step 2 and integrate the various geological, geophysical, and geochemical data corresponding to the locations of the drill holes in the study area to form a single sample with a one-to-one correspondence between features and labels.

[0025] The step 3.2 includes:

[0026] Step 3.2.1: Based on the structural features selected in Step 2.1, correlate the locations of the drill holes in the study area with the structural data: first, determine the structural type of the drill hole location, then determine the type of the structure, then determine the scale of the structure, then determine the activity period of the structure, and finally determine the distance between the favorable part of the structure for sandstone-type uranium mineralization and the drill hole location;

[0027] Step 3.2.2: Based on the igneous rock characteristics selected in Step 2.2, the locations of the drill holes in the study area are correlated with the igneous rock data: first, identify whether there are intermediate-acidic igneous rocks such as granite and volcanic rocks in the study area, then determine the distance between these igneous rocks and the drill holes, then determine the period of magmatic activity, and finally determine the distance between the favorable part of the igneous rock for sandstone-type uranium mineralization and the drill hole location;

[0028] Step 3.2.3: Based on the stratigraphic and lithofacies paleogeographic features selected in Step 2.3, correlate the location of the borehole in the study area with the stratigraphic and lithofacies paleogeographic data: first, determine the prospecting target layer at the borehole location; second, determine the sedimentary facies type at the borehole location; third, determine the paleoclimatic conditions at the borehole location; and finally, determine the sedimentary rock structure, marker minerals, typical paleontology, sediment grain size, composition, chemical element composition, and distance from the favorable sandstone-type uranium mineralization sedimentary facies at the borehole location.

[0029] Step 3.2.4: Based on the geophysical features selected in Step 2.4, correlate the borehole locations within the study area with the intensity, morphology, and occurrence of geophysical anomalies, as well as the distances to the borehole locations of anomalies similar to known sandstone-type uranium deposits. First, plot the geophysical survey data as a plan or cross-section, then project the borehole locations in the study area onto the plot. Then, record the geophysical survey data corresponding to the borehole locations as features in the sample.

[0030] Step 3.2.5: Based on the geochemical features selected in step 2.5, correlate the locations of the drill holes in the study area with the intensity, variation gradient, and distance from the concentration center of the geochemical element anomaly combinations closely related to sandstone-type uranium deposits determined after analysis: First, plot the geochemical measurement data as a plan or cross-section, then project the locations of the drill holes in the study area onto the map, and then record the geochemical measurement data corresponding to the drill hole locations as features in the sample.

[0031] The step 4 comprises:

[0032] Step 4.1: Gather the individual samples formed in step 3 to form a sample set;

[0033] Step 4.2: Clean, integrate, transform, and convert the data in the sample set;

[0034] Step 4.3: Randomly select samples from the sample set to construct the training set, and the remaining samples are used to construct the test set.

[0035] Said step 4.2 comprises:

[0036] Step 4.2.1: Clean the data in the sample set, mainly removing duplicate data in the sample set, identifying and correcting outliers or erroneous data, and deleting samples with missing data;

[0037] Step 4.2.2: Integrate the data in the sample set, mainly merging data from different sources to ensure consistency and integrity;

[0038] Step 4.2.3: Transform the data in the sample set. This mainly involves converting non-numeric features in the sample set into numeric features through one-hot encoding. This means further refining a feature into a specific category. All refined features are included in the sample and replace the original features. If a refined feature exists, it is marked as 1; if it does not exist, it is marked as 0.

[0039] Step 4.2.4: Convert the data in the sample set. The data obtained through the above-mentioned analysis of structural, igneous rocks, lithofacies paleogeography, geophysics, geochemistry and other characteristics have different dimensions. Direct use will cause the trained sandstone-type uranium deposit big data prediction model to be biased towards larger numerical weights, which cannot reflect the true sandstone-type uranium mineralization characteristics. Therefore, it is necessary to convert the data in the sample set into dimensionless data through normalization, and first screen out the maximum and minimum values ​​of a certain feature in the data set, then subtract the minimum value from each value of the feature, and then divide the difference by the difference between the maximum and minimum values ​​to convert the value into a number between 0 and 1.

[0040] The step 4.3 includes:

[0041] This step requires the use of the Python programming language.

[0042] Step 4.3.1: Use Python to load the Numpy and Pandas modules and call the train_test_split function in the model_selection toolkit in the sklearn library.

[0043] Step 4.3.2: Open the sample set created earlier, assign the features of each sample to X and the label to y, and convert them into Numpy arrays;

[0044] Step 4.3.3: Divide the features X and labels y into training and test sets and save them as .npy files.

[0045] The beneficial technical effects of the present invention are:

[0046] The present invention provides a sample set construction method suitable for sandstone-type uranium deposit big data prediction. The method provides a basis for prospecting prediction by analyzing geological, geophysical, geochemical and other data in traditional sandstone-type uranium deposit prediction work, integrates the steps of determining sandstone-type uranium deposit prospecting signs into the process of constructing big data prediction sample sets, and makes full use of the experience and knowledge of geological experts to optimize the workflow of constructing sandstone-type uranium deposit big data prediction sample sets, eliminates the influence of interference factors, and thus reduces the uncertainty in the process of constructing sandstone-type uranium deposit big data prediction sample sets by relying solely on full data. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a gravity anomaly distribution map of the Yin'e Basin in an embodiment of the present invention;

[0048] Figure 2 This is the magnetotelluric characteristic map of the Saierheina Depression in an embodiment of the present invention: Figure 2 a is the profile obtained by audio frequency magnetotelluric measurement, Figure 2 b is based on Figure 2 a Geological profile obtained by interpreting the audio magnetotelluric profile;

[0049] Figure 3 This is the rare earth element distribution diagram of the rock sample in the lower section of Suhongtu in Yin'e Basin in the embodiment of the present invention: Figure 3 a is the rare earth element distribution diagram of the mudstone / siltstone and sandstone samples in drilling ZKJ03. Figure 3 b is the rare earth element distribution diagram of the sandstone sample of drill hole ZKS01, and the rare earth element distribution diagram of the mudstone / siltstone and conglomerate samples of drill hole ZKS01. DETAILED DESCRIPTION

[0050] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0051] The present invention provides a method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits, which specifically includes the following steps:

[0052] Step 1: Systematically collect and organize geological, geophysical, geochemical and remote sensing data of the study area, extract geological, geophysical, geochemical and remote sensing data, and pre-process the data;

[0053] Step 1.1: Define the possible geological environment of typical sandstone-type uranium deposits in the study area;

[0054] Step 1.2: Collect original geological, geophysical, and geochemical data. Original geological, geophysical, and geochemical data include but are not limited to field observation data, laboratory analysis results, geophysical survey original data or interpretation results, and geochemical original data or results data.

[0055] Step 1.3: Classify and organize the collected raw data and data, and perform preliminary preprocessing on the raw data and data using data cleaning, missing value interpolation, and integration methods;

[0056] Step 1.4: Establish a multi-information database containing geological, geophysical, geochemical and other data.

[0057] Step 2: Feature selection of sample sets by analyzing factors such as structure, magmatic activity, stratigraphy and lithofacies paleogeography, geophysics, and geochemistry;

[0058] Step 2.1: Analyze the structural factors related to sandstone-type uranium mineralization in the study area. Analyze the ore-controlling characteristics and favorable mineralization locations of the structural form most closely related to sandstone-type uranium mineralization, including faults, folds, fissures, structural interfaces, and igneous rock structures. Determine the nature, scale, activity period, and ore-control mechanism of this structural form.

[0059] Step 2.2: Analyze the factors of magmatic activity related to sandstone-type uranium mineralization in the study area, analyze the spatial and temporal connections between magmatic rocks, especially acidic magmatic rocks, and uranium mineralization, the specificity of magmatic mineralization, the physicochemical conditions of magmatic activity, and the genetic connection between magmatic activity and uranium mineralization, and determine the characteristics of magmatic activity closely related to sandstone-type uranium mineralization;

[0060] Step 2.3: Analyze the stratigraphic and lithofacies paleogeographic factors related to sandstone-type uranium mineralization in the study area. Analyze the rock types, sedimentary environments, and paleogeographic factors of the stratigraphic units related to sandstone-type uranium mineralization, and determine the stratigraphic and lithofacies paleogeographic characteristics closely related to sandstone-type uranium mineralization.

[0061] Step 2.4: Analyze the geophysical factors related to sandstone-type uranium mineralization in the study area. Use geophysical methods (such as gravity, magnetism, and resistivity) to detect the underground structure and rock physical properties related to sandstone-type uranium mineralization, and determine the geophysical field characteristics closely related to sandstone-type uranium mineralization.

[0062] Step 2.5: Analyze the geochemical factors related to sandstone-type uranium mineralization in the study area. Identify the element distribution patterns and abnormal areas related to sandstone-type uranium deposits through geochemical data, and determine the geochemical characteristics closely related to sandstone-type uranium deposits.

[0063] Step 3: Integrate the selected sample feature data, mark the features and labels of each sample, and form a complete single sample data;

[0064] Step 3.1: Take the existing drill holes in the study area as samples of the data set and classify them according to whether they are sandstone-type uranium industrial holes, mineralized holes, and other drill holes. Sandstone-type uranium industrial holes and mineralized holes are marked as 1, and other drill holes are marked as 0, which are used as labels for whether the samples have mines.

[0065] Step 3.2: Combine the features selected in step 2 and integrate the various geological, geophysical, and geochemical data corresponding to the locations of the boreholes in the study area to form a single sample with one-to-one correspondence between features and labels;

[0066] Step 3.2.1: Based on the structural features selected in Step 2.1, the locations of the drill holes in the study area are correlated with the structural data: first, the structural type of the drill hole location is determined (one of the five structural forms: fracture, structural interface, and igneous rock structure). Secondly, the type of the structure is determined. Thirdly, the scale of the structure is determined. Finally, the activity period of the structure is determined. Finally, the distance between the favorable part of the structure for sandstone-type uranium mineralization and the drill hole location is determined.

[0067] Step 3.2.2: Based on the igneous rock characteristics selected in Step 2.2, the locations of the drill holes in the study area are correlated with the igneous rock data: first, identify whether there are intermediate-acidic igneous rocks such as granite and volcanic rocks in the study area, then determine the distance between these igneous rocks and the drill holes, then determine the period of magmatic activity, and finally determine the distance between the favorable part of the igneous rock for sandstone-type uranium mineralization and the drill hole location;

[0068] Step 3.2.3: Based on the stratigraphic and lithofacies paleogeographic features selected in Step 2.3, correlate the location of the borehole in the study area with the stratigraphic and lithofacies paleogeographic data: first, determine the prospecting target layer at the borehole location; second, determine the sedimentary facies type at the borehole location; third, determine the paleoclimatic conditions at the borehole location; and finally, determine the sedimentary rock structure, marker minerals, typical paleontology, sediment grain size, composition, chemical element composition, and distance from the favorable sandstone-type uranium mineralization sedimentary facies at the borehole location.

[0069] Step 3.2.4: Based on the geophysical features selected in Step 2.4, correlate the locations of the boreholes in the study area with the intensity, morphology, and occurrence of geophysical anomalies (gravity, magnetic, electrical, and seismic), as well as the distances of anomalies similar to known sandstone-type uranium deposits to the borehole locations: First, plot the geophysical survey data as a plan or cross-section, then project the borehole locations in the study area onto the plot, and then record the geophysical survey data corresponding to the borehole locations as features in the sample;

[0070] Step 3.2.5: Based on the geochemical features selected in step 2.5, correlate the locations of the drill holes in the study area with the intensity, variation gradient, and distance from the concentration center of the geochemical element anomaly combinations closely related to sandstone-type uranium deposits determined after analysis: First, plot the geochemical measurement data as a plan or cross-section, then project the locations of the drill holes in the study area onto the map, and then record the geochemical measurement data corresponding to the drill hole locations as features in the sample.

[0071] Step 4: Integrate the individual sample data to construct a sample set, perform feature engineering on the data in the sample set, randomly select samples to construct a training set, and use the remaining samples to construct a test set;

[0072] Step 4.1: Gather the individual samples formed in step 3 to form a sample set;

[0073] Step 4.2: Clean, integrate, transform, and convert the data in the sample set;

[0074] Step 4.2.1: Clean the data in the sample set, mainly removing duplicate data in the sample set, identifying and correcting outliers or erroneous data, and deleting samples with missing data;

[0075] Step 4.2.2: Integrate the data in the sample set, mainly merging data from different sources to ensure consistency and integrity;

[0076] Step 4.2.3: Transform the data in the sample set. This mainly involves converting non-numeric features in the sample set into numeric features through one-hot encoding. This means further refining a feature into a specific category. All refined features are included in the sample and replace the original features. If a refined feature exists, it is marked as 1; if it does not exist, it is marked as 0.

[0077] Step 4.2.4: Convert the data in the sample set. The data obtained through the aforementioned analysis of structural, magmatic, lithofacies, paleogeographic, geophysical, and geochemical characteristics have different dimensions. Direct use will cause the trained sandstone-type uranium deposit big data prediction model to favor larger weights and fail to reflect the true sandstone-type uranium mineralization characteristics. Therefore, it is necessary to convert the data in the sample set into dimensionless data through normalization. For a certain feature in the data set, the maximum and minimum values ​​are first screened out. The minimum value is then subtracted from each value of the feature. The difference is then divided by the difference between the maximum and minimum values, (x-min) / (max-min), to convert the value to a number between 0 and 1.

[0078] Step 4.3: Randomly select samples from the sample set to construct the training set, and the remaining samples are used to construct the test set;

[0079] Step 4.3.1: Use Python to load the Numpy and Pandas modules and call the train_test_split function in the model_selection toolkit in the sklearn library.

[0080] Step 4.3.2: Open the sample set created earlier, assign the features of each sample to X and the label to y, and convert them into Numpy arrays;

[0081] Step 4.3.3: Divide the features X and labels y into training and test sets and save them as .npy files.

[0082] Taking the Yin'e Basin as an example, the present invention provides a sample set construction method suitable for sandstone-type uranium deposit big data prediction, which specifically includes the following steps:

[0083] Step 1: Systematically collect and organize relevant geological, geophysical, geochemical, and remote sensing data of the study area, extract the corresponding geological, geophysical, geochemical, and remote sensing data, and preprocess the data;

[0084] By defining the possible geological environment of typical sandstone-type uranium deposits in the study area, relevant geological, geophysical and geochemical original data and information are collected in a targeted manner, the collected original data and information are classified and organized, and the original data and information are preliminarily preprocessed to establish a database.

[0085] For example, by analyzing the mineralization geological background of the Yin'e Basin, it is preliminarily determined that the sandstone-type uranium mineralization in the Yin'e Basin is mainly found in the upper part of the Bayingebi Formation, followed by the Suhongtu Formation and the Ulansuhai Formation; the mudstone-type uranium mineralization occurs in the upper part of the Bayingebi Formation and the mudstone of the Suhongtu Formation; the volcanic rock-type uranium mineralization is mainly basalt-type, concentrated in the southern edge of the Suhongtu Depression, etc., as shown in Table 1.

[0086] Table 1 Characteristic data of uranium mineralization in the Yin'e Basin

[0087]

[0088] Targetedly collect relevant geographical, physical, and chemical data, organize and pre-process them, and establish a multi-information database, as follows:

[0089] 1) The Yin'e Basin is located in the Xing'an-Mongol Orogenic Belt, the eastern section of the Central Asian Orogenic Belt, sandwiched between the North China Plate and the Siberian Plate. It is located at the intersection of the Paleo-Asian Tectonic Domain and the Pacific Tectonic Domain, and is adjacent to the Tarim Plate, Kazakhstan Plate, Siberian Plate, and North China Plate. During the Mesozoic, the northward compression of the Indian Plate caused compression of the Alxa Block, forming a large-scale strike-slip pull-apart fault zone. At the same time, large-scale basaltic magma activity occurred along the strike-slip fault zone, forming a large strike-slip pull-apart basin.

[0090] 2) After sorting out the characteristics of the volcanic rocks in the area, the data were obtained, as shown in Table 2.

[0091] Table 2 Characteristic data of magmatic rocks in the Yin'e Basin

[0092]

[0093] 3) After sorting out the stratigraphic characteristics of the area, the data were obtained, as shown in Table 3.

[0094] Table 3 Distribution data of Mesozoic and Cenozoic strata and mineral resources in depressions of Yin'e Basin

[0095]

[0096] 4) After sorting out the hydration characteristics in the area, the data were obtained, as shown in Table 4.

[0097] Table 4 Hydration characteristics of the Yin'e Basin

[0098]

[0099] 5) After sorting out the geophysical characteristics of the area, the gravity anomaly distribution map of the Yin'e Basin was obtained, such as Figure 1 shown.

[0100] 6) The drilling data in the study area were collected and sorted to construct a database, as shown in Table 5.

[0101] Table 5 Partial preview data of the database

[0102]

[0103] Step 2: Feature selection of sample sets is performed by analyzing factors such as structure, magmatic activity, stratigraphy and lithofacies paleogeography, geophysics, and geochemistry. The specific steps can be divided into:

[0104] Step 2.1: First, analyze the structural form that is most closely related to sandstone-type uranium mineralization, its ore-controlling characteristics and favorable mineralization locations, and determine the structural characteristics such as the nature, scale, activity period, and ore-control mechanism of the structural form.

[0105] For example, favorable sandstone-type uranium mineralization structures in the Yin'e Basin include tectonic slope zones, a wing of a wide and gentle fold, and a fault-uplift zone. The target strata in these areas are shallowly buried and easily uplifted, favoring the infiltration of surface oxygen- and uranium-bearing water. The Juyanhai Depression within the basin has experienced numerous major tectonic movements since the Jurassic, including the Yanshanian II and III movements and the Himalayan Orogeny. These movements have resulted in unconformities between the Middle and Lower Jurassic and Cretaceous strata, between the Bayingebi and Suhongtu Formations, and between the Lower Cretaceous and Neogene strata. Among them, the northern part of the Wujuer Sag on the northern edge of the area has always maintained a slope belt, and gradually shrunk from the Jurassic-Bayingebi Formation to the Suhongtu Formation, which is conducive to infiltration-type uranium mineralization; the Lujing-Tiancao-Jianguoying Sag was a dustpan-shaped sag with north fault and south overlap in the early Suhongtu period, and delta sand bodies developed on the southeast slope belt. After tectonic transformation, the burial depth of the southern slopes of the Lujing Sag, Tiancao Sag, Jianguoying Sag, and Jigeda Sag was less than 800m, which is conducive to infiltration and exudation superimposed uranium mineralization; the late Suhongtu period was a depression stage, with inter-structural paleochannel-type uranium mineralization developed. The Lujing South Uplift, Wujiajing Uplift and Baogeda Uplift were in low-lying areas in the late Suhongtu period, which is conducive to sandstone-type uranium mineralization.

[0106] Faults are well developed in the depression, most of which are normal faults. According to the fault strike, they can be divided into three groups: NE, NNE, and NW. The activity period and intensity of the faults have different effects on the degree of structural, sedimentary construction and transformation, and can be mainly divided into two categories: one is the syn-sedimentary fault at the depression boundary, which plays a controlling role in the formation and development of the depression. For example, the uplifted wall of the Tiancao No. 1 fault is the Lunan uplift, which strikes NNE and dips to the southeast. It extends 86 km in the Carboniferous-Permian fault zone, with a maximum horizontal fault throw of 1.5 km and a maximum vertical fault throw of 1.8 km, controlling the sedimentation of the Tiancao depression; the Tiancao No. 2 fault strikes NNE and dips to the northwest. It extends 80 km in the Carboniferous-Permian fault zone, with a maximum horizontal fault throw of 1 km and a maximum vertical fault throw of 1.5 km. Together with the Tiancao No. 1 fault, it controls the distribution of the Tiancao depression. Second, faults developed within the depression in all periods of the Early Cretaceous, with greater intensity of fault activity during the Bayingebi and Suhongtu periods. Secondary faults developed throughout the depression, controlling the structural morphology of the gentle slope.

[0107] Step 2.2: Analyze the spatial and temporal connections between the magmatic rocks, especially the acidic magmatic rocks, and uranium mineralization in the study area, the specificity of magmatic mineralization, the physicochemical conditions of magmatic activity, and the genetic connection between magmatic activity and uranium mineralization, and determine the characteristics of magmatic activity that are closely related to sandstone-type uranium mineralization.

[0108] For example, the late Variscan granite is the most widely distributed in the Yin'e Basin, mainly in the Zongnai-Shalazha Mountain uplift, Hongger Mountain, Beida Mountain, Yabulai Mountain, the southern part of the Langshan uplift, and the Bayan Ula Mountain uplift. It is mainly composed of gray-red medium-grained porphyritic biotite granite, monzogranite, plagiogranite, potassium-feldspar granite, and fine-medium-grained biotite plagiogranite. The uranium content in the rock mass is relatively high, reaching (2.3-5.5)×10 -6 , the ancient uranium abundance is generally (2.0~93.5)×10 -6 The Th / U ratio is relatively large, ranging from 5.2 to 11.2, indicating that the amount of uranium loss is large, providing an over-rich uranium source for the interior of the basin.

[0109] Step 2.3: Analyze the stratigraphic and lithofacies paleogeographic factors related to sandstone-type uranium mineralization in the study area, focusing on the rock types, sedimentary environments and paleogeographic factors of the stratigraphic units related to sandstone-type uranium mineralization, and determine the stratigraphic and lithofacies paleogeographic characteristics closely related to sandstone-type uranium mineralization.

[0110] For example, the Kuquangou Formation (N2k) is primarily distributed in the Juyanhai Depression in the western Yin'e Basin. Its lithology consists of light brownish-red and brownish-red mudstone, fine-medium sandstone, and light grayish-white fine sandstone. Gray and light grayish-green mudstone can be seen in the middle, and thin conglomerates are occasionally found at the base. The Kuquangou Formation (N2k) produces a gymnosperm pollen assemblage, including single and double bundles of pine pollen (Abietineae / Pinuspollenites), spruce pollen (Piceaepollenites), cedar pollen (Cedripites deodariformis), and large larch pollen (Laricoidites magnus). This formation reflects a semi-arid hot paleoclimate and is composed of a predominantly variegated clastic rock formation.

[0111] During the deposition of the Kuquangou Formation, the Juyan Depression entered a period of differential subsidence, developing fluvial sediments. The basin is dominated by fine-grained floodplain sediments. The residual thickness of this formation ranges from 15 to 260 meters, with the greatest residual thickness occurring in the Tiancao-Jianguoying Sag area, followed by the Lujing Sag in the west. The center of the residual thickness is located in the basin center, thinning toward the basin margins until it disappears. This sand body is primarily distributed on the southern side of the Juyan Depression. The lithology is primarily light brown to light grayish-white fine and medium sandstone, with a thickness ranging from 25 to 150 meters, reaching 150 to 200 meters in some areas. The sand bodies are continuous and stable, with a moderate thickness, gradually thinning from south to north. Preliminary estimates suggest the sand bodies extend for approximately 50 kilometers along strike.

[0112] The Kuquangou Formation (N2k) was deposited in an arid, oxidizing environment. The strata are generally light brown-red in color, with generally loose rocks and good permeability. A phenomenon of "two red muds sandwiched between one gray sand" can be observed in the river channel. Uranium mineralization is controlled by gray sandstone, representing a paleo-channel-type, "seepage-type" composite uranium mineralization. For example, the anomalous section in drill hole ZKJ05 is a gray fine sandstone sandwiched between brown-red mudstone and siltstone. Microscopic analysis of the uranium mineralization in this layer reveals a significant presence of biogenic oolitic sandstone. Uranium enrichment is likely related to adsorption by biological debris.

[0113] Step 2.4: Analyze the geophysical factors related to sandstone-type uranium mineralization in the study area, mainly using geophysical methods (such as gravity, magnetism, resistivity, etc.) to detect the underground structure and physical properties of rocks related to sandstone-type uranium mineralization, and determine the geophysical field characteristics closely related to sandstone-type uranium mineralization.

[0114] For example: using magnetotelluric data to explain the deep sand body development characteristics of Saierhaina Depression in the northwest of Guaizihu Depression in Yin'e Basin, showing the location of sand bodies inferred by geophysical methods (electrical methods), showing the use of geophysical methods (electrical methods) to detect underground structures and physical properties of rocks related to sandstone-type uranium mineralization, and determining the geophysical field characteristics closely related to sandstone-type uranium mineralization. Figure 2As shown in the cross-section, it can be seen that dolomitic mudstone and uranium mineralization are developed at the top of the upper section of the Bayingebi Formation, and 20-60 meter sand bodies are developed in the lower part; on the plane, alluvial fan-fan deltas are developed around the depression, and the fan deltas in the west and north are large in scale, about 5-12 km wide and 3-5 km long, which is conducive to uranium mineralization.

[0115] Step 2.5: Analyze the geochemical factors related to sandstone-type uranium mineralization in the study area. Mainly through geochemical data, identify the element distribution patterns and abnormal areas related to sandstone-type uranium deposits, and determine the geochemical characteristics closely related to sandstone-type uranium mineralization.

[0116] For example, the major element analysis results of rock samples from the lower section of Suhongtu in the Yin'e Basin showed that the SiO2 content of mudstone samples was 13.7% to 65.6%, with an average of 39.7%; the Al2O3 content was 4.1% to 16.2%, with an average of 12%; the FeO content was 0.97% to 4.17%, with an average of 1.93%; the CaO content was 2.3% to 39.2%, with an average of 12.5%; the Na2O content was 0.4% to 4.4%, with an average of 2.8%; and the K2O content was 0.7% to 4.2%, with an average of 2.5%. The SiO2 content of sandstone samples is 30.3% to 69%, with an average of 58.4%; the Al2O3 content is 6.8% to 15.9%, with an average of 12.3%; the FeO content is 0.7% to 7.6%, with an average of 2.6%; the CaO content is 0.4% to 22.9%, with an average of 6.4%; the Na2O content is 1.15% to 3.15%, with an average of 2.0%; and the K2O content is 1.05% to 3.84%, with an average of 2.47%.

[0117] Samples with less CaO content than Na2O content were selected for calculation of the chemical weathering index (CIA) value, with a result of 60-74 (average of 692 mudstone / siltstone samples and 10 sandstone samples). A CIA value as low as 60-80 is generally considered to be moderate chemical weathering. The calculated results show that the samples have undergone moderate to weak chemical weathering. The composition variation index (ICV) value is 0.6-1.2, with an average of 1.0, while the ICV values ​​of most samples are between 0.6 and 0.8, which may indicate that the sediments have undergone a certain degree of re-cycling.

[0118] The results of trace element analysis showed that the V content was 27.8×10 -6 ~285×10 -6 , an average of 101×10 -6 ; Cr content 15.5×10 -6 ~271×10 -6 , an average of 57.5×10 -6 ; Ni content 8.11×10 -6 ~89.6×10-6 , with an average of 29.0×10 -6 ; Cu content 7.47×10 -6 ~156×10 -6 , with an average of 32.3×10 -6 ; Ba content 121×10 -6 ~912×10 -6 , an average of 416×10 -6 ; U content 1.13×10 -6 ~209×10 -6 , an average of 9.2×10 -6 .

[0119] The rare earth element analysis results showed that the ∑REE content was 42.2×10 -6 ~291.1×10 -6 , with an average of 133.9×10 -6 ; LREE / HREE values ​​are 3.8-10.4, with an average of 7.2; La / Yb values ​​are 4.8-22.9, with an average of 10.1, indicating that light and heavy rare earth elements are clearly differentiated, and light rare earth elements are relatively enriched; δEu values ​​are 0.5-0.8, with an average of 0.7, indicating that Eu is a weak negative anomaly; δCe values ​​are 0.8-1.1, with an average of 1.0, indicating that Ce is a weak negative anomaly to a slight positive anomaly. Three rock samples obtained from two boreholes (ZKJ03 and ZKS01) in the Yin'e Basin were analyzed for rare earth elements, and the ratio of their rare earth element content to that of chondrites was used to produce a rare earth element distribution pattern diagram, showing that the element distribution pattern related to sandstone-type uranium deposits can be identified through geochemical data, and the geochemical characteristics closely related to the mineralization of sandstone-type uranium deposits can be determined. Figure 3 As shown, the curve is right-leaning as a whole, the light rare earth element curve is right-shaped, and the heavy rare earth element curve is basically flat.

[0120] Step 3: Integrate the selected sample feature data, mark the features and labels of each sample, and form a complete single sample data. The specific steps can be divided into:

[0121] Step 3.1: Take the existing drill holes in the study area as samples of the data set and classify them according to whether they are sandstone-type uranium industrial holes, mineralized holes, and other drill holes. Sandstone-type uranium industrial holes and mineralized holes are marked as 1, and other drill holes are marked as 0 to determine whether the samples have a mine label.

[0122] For example, after classifying the drill holes in the study area into sandstone-type uranium industrial holes, mineralized holes, and other drill holes, the drill hole sample label data is obtained, as shown in Table 6.

[0123] Table 6 Label data of some borehole samples in the Yin'e Basin

[0124] Drilling number Coordinate Y Coordinate X Coordinate Z Correction hole depth Drilling type Label ZKL1-2 4624878 17574023 975 162.6 No ore hole 0 ZKT0-2 4623170 17614039 982 453.39 Uranium anomaly holes 0 ZKL0-3 4634406 17564153 963 550.07 Uranium mineralization holes 1 ZKY0-3 4634567 17654321 987 500.4 Uranium industry mine holes 1

[0125] Step 3.2: Integrate the various geological, geophysical, and geochemical data corresponding to the locations of the boreholes in the study area to form a single sample with one-to-one correspondence between features and labels;

[0126] For example, the structural, igneous rock, stratigraphic and lithofacies paleogeographic, geophysical, and geochemical characteristics selected in step 2 are integrated together and correspond to the location of the borehole to obtain a complete data sample, as shown in Table 7.

[0127] Table 7 Sample prediction data of single sandstone-type uranium deposits in the Yin'e Basin

[0128] Drilling number Is it in a slope area? Distance from fracture Distance from rock mass Destination layer Sedimentary environment Sedimentary environment Magnetotellurics soil radon Distance from soil radon concentration center Label ZKY0-3 1 5 8 N2k River channel phase drought oxidation 7 30000 0.5 1

[0129] Step 4: Integrate the individual sample data generated in step 3 to construct a sample set, perform feature engineering on the data in the sample set, and divide the sample set into a training set and a test set. The specific steps can be divided into:

[0130] Step 4.1: Gather the individual samples formed in step 3 to form a sample set;

[0131] For example, after integrating the samples from the Yin'e Basin, we get the sample set shown in Table 8.

[0132] Table 8. Sample set of big data prediction of sandstone-type uranium deposits in some Yin'e Basins

[0133] Drilling number Is it in a slope area? Distance from fracture Distance from rock mass Destination layer Sedimentary environment Sedimentary environment Magnetotellurics soil radon Distance from soil radon concentration center Label ZKY0-3 1 5 8 N2k River channel phase drought oxidation 7 30000 0.5 1 ZK6-6 0 4.3 10.5 K1b delta plain drought oxidation 12.4 16600 1.8 0 ZKJ1-1 0 3.2 9.4 K2s delta front Drought reduction 8.6 9876 2.2 0 ZKL0-2 1 2.7 6.6 K1b River channel phase drought oxidation 6.5 28430 0.3 1 ZKL1-1 0 5.8 12.3 K2s Alluvial fan Drought reduction 11.7 17200 1.4 0

[0134] Step 4.2: Perform feature engineering on the data in the sample set to ensure that the sample set can be used by the sandstone-type uranium deposit big data prediction model;

[0135] For example, the Yin'e Basin sample set needs to normalize all data greater than 1 and convert them into dimensionless values ​​between 0 and 1. One-hot encoding is used to convert all non-numeric data into numeric values ​​of 0 or 1. After feature engineering, the normalized sample set is obtained, as shown in Table 9.

[0136] Table 9 Normalized sample set of sandstone-type uranium deposit prediction big data in the Yin'e Basin

[0137] Drilling number Is it in a slope area? Distance from fracture Distance from rock mass N2k K1b K2s River channel phase delta plain delta front Alluvial fan drought oxidation Drought reduction Magnetotellurics soil radon Distance from soil radon concentration center Label ZKY0-3 1 0.74 0.25 1 0 0 1 0 0 0 1 0 0.08 1 0.11 1 ZK6-6 0 0.52 0.68 0 1 0 0 1 0 0 1 0 1 0.33 0.79 0 ZKJ1-1 0 0.16 0.49 0 0 1 0 0 1 0 0 1 0.36 0 1 0 ZKL0-2 1 0 0 0 1 0 1 0 0 0 1 0 0 0.92 0 1 ZKL1-1 0 1 1 0 0 1 0 0 0 1 0 1 0.88 0.36 0.58 0

[0138] Step 4.3: Randomly select a certain number of samples from the sample set to construct a training set for training the sandstone-type uranium deposit big data prediction model. The remaining samples are used to evaluate the prediction performance of the sandstone-type uranium deposit big data prediction model.

[0139] For example, to split the Yin'e Basin sample collection into training and test sets, first load the relevant modules in the Python programming environment; then define two empty lists, X and y, to store sample features and labels, respectively. Read the sample data line by line, extract the comma-delimited sample data one by one, save all data except the last value of each sample as features, and save the last value of each sample as label. Add the extracted features and label to arrays X and y. Then convert lists X and y to numpy arrays. Use the train_test_split function to split arrays X and y into training sets X_train and y_train, and test sets X_test and y_test. Finally, save the training and test set feature and label data to files X_train.npy, X_test.npy, y_train.npy, and y_test.npy, respectively.

[0140] The present invention has been described in detail above with reference to the accompanying drawings and embodiments. However, the present invention is not limited to the above embodiments. Various modifications can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention. Any content not described in detail in the present invention may be adapted from existing technologies.

Claims

1. A method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits, characterized in that: The method comprises: Step 1: Systematically collect and organize geological, geophysical, geochemical and remote sensing data of the study area, extract geological, geophysical, geochemical and remote sensing data, and pre-process the data; Step 2: Feature selection of sample sets by analyzing factors such as structure, magmatic activity, stratigraphy and lithofacies paleogeography, geophysics, and geochemistry; Step 3: Integrate the selected sample feature data, mark the features and labels of each sample, and form a complete single sample data; Step 4: Integrate the formed individual sample data together to construct a sample set, perform feature engineering on the data in the sample set, randomly select samples to construct a training set, and use the remaining samples to construct a test set.

2. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 1, characterized in that: The step 1 comprises: Step 1.1: Define the possible geological environment of typical sandstone-type uranium deposits in the study area; Step 1.2: Collect geological, geophysical and geochemical original information and data; Step 1.3: Classify and organize the collected raw data and data, and perform preliminary preprocessing on the raw data and data using data cleaning, missing value interpolation, and integration methods; Step 1.4: Establish a multi-information database containing geological, geophysical, geochemical and other data.

3. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 1, characterized in that: The geological, geophysical and geochemical original data and information include field observation data, laboratory analysis results, geophysical measurement original data or interpretation results, and geochemical original data or results data.

4. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 1, characterized in that: The step 2 includes: Step 2.1: Analyze the structural factors related to sandstone-type uranium mineralization in the study area. Analyze the ore-controlling characteristics and favorable mineralization locations of the structural form most closely related to sandstone-type uranium mineralization, including faults, folds, fissures, structural interfaces, and igneous rock structures. Determine the nature, scale, activity period, and ore-control mechanism of this structural form. Step 2.2: Analyze the factors of magmatic activity related to sandstone-type uranium mineralization in the study area, analyze the spatial and temporal connections between magmatic rocks and uranium mineralization, the specificity of magmatic mineralization, the physicochemical conditions of magmatic activity, and the genetic connection between magmatic activity and uranium mineralization, and determine the characteristics of magmatic activity closely related to sandstone-type uranium mineralization; Step 2.3: Analyze the stratigraphic and lithofacies paleogeographic factors related to sandstone-type uranium mineralization in the study area. Analyze the rock types, sedimentary environments, and paleogeographic factors of the stratigraphic units related to sandstone-type uranium mineralization, and determine the stratigraphic and lithofacies paleogeographic characteristics closely related to sandstone-type uranium mineralization. Step 2.4: Analyze the geophysical factors related to sandstone-type uranium mineralization in the study area, use geophysical methods to detect the underground structure and physical properties of rocks related to sandstone-type uranium mineralization, and determine the geophysical field characteristics closely related to sandstone-type uranium mineralization; Step 2.5: Analyze the geochemical factors related to sandstone-type uranium mineralization in the study area. Identify the element distribution patterns and abnormal areas related to sandstone-type uranium deposits through geochemical data, and determine the geochemical characteristics closely related to sandstone-type uranium deposits.

5. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 4, characterized in that: The step 3 includes: Step 3.1: Take the existing drill holes in the study area as samples of the data set and classify them according to whether they are sandstone-type uranium industrial holes, mineralized holes, and other drill holes. Sandstone-type uranium industrial holes and mineralized holes are marked as 1, and other drill holes are marked as 0, which are used as labels for whether the samples have mines. Step 3.2: Combine the features selected in step 2 and integrate the various geological, geophysical, and geochemical data corresponding to the locations of the drill holes in the study area to form a single sample with a one-to-one correspondence between features and labels.

6. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 5, characterized in that: The step 3.2 includes: Step 3.2.1: Based on the structural features selected in Step 2.1, correlate the locations of the drill holes in the study area with the structural data: first, determine the structural type of the drill hole location, then determine the type of the structure, then determine the scale of the structure, then determine the activity period of the structure, and finally determine the distance between the favorable part of the structure for sandstone-type uranium mineralization and the drill hole location; Step 3.2.2: Based on the igneous rock characteristics selected in Step 2.2, the locations of the drill holes in the study area are correlated with the igneous rock data: first, identify whether there are intermediate-acidic igneous rocks such as granite and volcanic rocks in the study area, then determine the distance between these igneous rocks and the drill holes, then determine the period of magmatic activity, and finally determine the distance between the favorable part of the igneous rock for sandstone-type uranium mineralization and the drill hole location; Step 3.2.3: Based on the stratigraphic and lithofacies paleogeographic features selected in Step 2.3, correlate the borehole locations within the study area with stratigraphic and lithofacies paleogeographic data: first, determine the prospecting target stratum at the borehole location; second, determine the sedimentary facies type at the borehole location; third, determine the paleoclimatic conditions at the borehole location; and finally, determine the sedimentary rock structure, marker minerals, typical paleontology, sediment grain size, composition, chemical element composition, and distance from favorable sandstone-type uranium mineralization sedimentary facies at the borehole location. Step 3.2.4: Based on the geophysical features selected in Step 2.4, correlate the borehole locations within the study area with the intensity, morphology, and occurrence of geophysical anomalies, as well as the distances to the borehole locations of anomalies similar to known sandstone-type uranium deposits. First, plot the geophysical survey data as a plan or cross-section, then project the borehole locations in the study area onto the plot. Then, record the geophysical survey data corresponding to the borehole locations as features in the sample. Step 3.2.5: Based on the geochemical features selected in step 2.5, correlate the locations of the drill holes in the study area with the intensity, variation gradient, and distance from the concentration center of the geochemical element anomaly combinations closely related to sandstone-type uranium deposits determined after analysis: First, plot the geochemical measurement data as a plan or cross-section, then project the locations of the drill holes in the study area onto the map, and then record the geochemical measurement data corresponding to the drill hole locations as features in the sample.

7. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 1, characterized in that: The step 4 comprises: Step 4.1: Gather the individual samples formed in step 3 to form a sample set; Step 4.2: Clean, integrate, transform, and convert the data in the sample set; Step 4.3: Randomly select samples from the sample set to construct the training set, and the remaining samples are used to construct the test set.

8. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 7, characterized in that: Said step 4.2 comprises: Step 4.2.1: Clean the data in the sample set, mainly removing duplicate data in the sample set, identifying and correcting outliers or erroneous data, and deleting samples with missing data; Step 4.2.2: Integrate the data in the sample set, mainly merging data from different sources to ensure consistency and integrity; Step 4.2.3: Transform the data in the sample set. This mainly involves converting non-numeric features in the sample set into numeric features through one-hot encoding. This means further refining a feature into a specific category. All refined features are included in the sample and replace the original features. If a refined feature exists, it is marked as 1; if it does not exist, it is marked as 0. Step 4.2.4: Convert the data in the sample set. The data obtained through the above-mentioned analysis of structural, igneous rocks, lithofacies paleogeography, geophysics, geochemistry and other characteristics have different dimensions. Direct use will cause the trained sandstone-type uranium deposit big data prediction model to be biased towards larger numerical weights, which cannot reflect the true sandstone-type uranium mineralization characteristics. Therefore, it is necessary to convert the data in the sample set into dimensionless data through normalization, and first screen out the maximum and minimum values ​​of a certain feature in the data set, then subtract the minimum value from each value of the feature, and then divide the difference by the difference between the maximum and minimum values ​​to convert the value into a number between 0 and 1.

9. The method for constructing a sample set suitable for big data prediction of sandstone-type uranium deposits according to claim 7, characterized in that: Said step 4.3 comprises: This step requires the use of the Python programming language. Step 4.3.1: Use Python to load the Numpy and Pandas modules and call the train_test_split function in the model_selection toolkit in the sklearn library. Step 4.3.2: Open the sample set created earlier, assign the features of each sample to X and the label to y, and convert them into Numpy arrays; Step 4.3.3: Divide the features X and labels y into training and test sets and save them as .npy files.

Images